Sci. Aging Knowl. Environ., 1 June 2005
Vol. 2005, Issue 22, p. pe15
[DOI: 10.1126/sageke.2005.22.pe15]


Misdirection on the Road to Shangri-La

S. Jay Olshansky, Bruce A. Carnes, Ronald Hershow, Doug Passaro, Jennifer Layden, Jacob Brody, Leonard Hayflick, Robert N. Butler, David B. Allison, and David S. Ludwig

The authors are in the Division of Epidemiology and Biostatistics at the University of Illinois at Chicago, Chicago, IL 60612, USA (S.J.O., R.H., D.P., J.L., and J.B.), the Reynolds Department of Geriatric Medicine at the University of Oklahoma Health Sciences Center, Oklahoma City, OK 73190, USA (B.C.), the Department of Anatomy at the University of California, San Francisco, School of Medicine, The Sea Ranch, CA 95497, USA (L.H.), the International Longevity Center, New York, NY 10028, USA (R.N.B.), the Department of Biostatistics at the University of Alabama at Birmingham, Birmingham, AL 35294, USA (D.B.A.), and the Department of Medicine at Children's Hospital, Boston, MA 02115, USA (D.S.L.). E-mail: sjayo{at} (S.J.O.)

Key Words: obesity • mortality • life expectancy • life span • type II diabetes


When scientific evidence fails to support a position based on advocacy, those who take such positions sometimes rely on a very old but well-tested strategy--the magician's use of misdirection. For those who may be unfamiliar with this strategy, misdirection is the magician's way of distracting the attention of the audience away from reality. We contend that, in a recent Perspective in SAGE KE on the future of mortality [entitled "Future Mortality: A Bumpy Road to Shangri-La?" (1)], Shripad Tuljapurkar uses misdirection to take readers on a detour away from the reality of U.S. mortality and life expectancy by advocating, as he has in the past (2), that life expectancy in the United States should continue to rise rapidly in the future because it has done so elsewhere in the past.

In our article, recently published in the New England Journal of Medicine (NEJM) (3), we predicted that the steady rise in life expectancy in the United States in this century may soon come to an end because younger people today are far different from previous generations of younger people. We estimated that if obesity did not exist in the United States today, life expectancy at birth would be higher by approximately 4 to 9 months. We further estimated that if the observed relation between obesity and mortality remains constant, then the rapid rise of childhood obesity in the past two decades will, by the middle of this century, lead to a reduction in life expectancy of anywhere between 2 and 5 years. This turned out to be a very conservative assumption, because the prevalence of childhood obesity has risen since the time of the survey upon which our study was based.

Our calculations were not based on estimates of the number of deaths attributable to obesity (which remains the subject of great debate) or on forecasts of trends in obesity in the future but, rather, on the observed relation between body mass index (BMI) and the risk of death among people in the United States between the ages of 20 and 85 that was reported in health survey data (4).

In our NEJM article, we made the important point that the negative consequences for health and life expectancy that can be expected from rising levels of childhood obesity have yet to be realized. We envision these problems occurring in three phases (5). The first phase, which has been ongoing for some time now, is the rising prevalence of obesity among young children. The second phase, which was rare in the past but is emerging now, is the manifestation of obesity-induced health conditions among the young. For example, the rise of type II diabetes in children was unheard of just 30 years ago, but it is now occurring among obese children with an alarming frequency. The third phase, obesity-induced mortality, will emerge over the next 50 years as these children survive to the adult ages when obesity begins to kill. It is during this third phase that we suggest life expectancy in the United States could begin to decline.

The rise of childhood obesity in the United States can be equated to the introduction of a behavior that has a harmful but delayed effect on morbidity and mortality. Consider a hypothetical nation of mostly nonsmokers. A 30 percent increase in the prevalence of smoking among children over three decades would have minimal short-term health effects, but the long-term effects would be devastating. This is the scenario that confronts us with today's childhood obesity epidemic. The most detrimental health and longevity effects will not be seen for decades--a phenomenon that cannot be detected by current methods used to forecast life expectancy or estimate the number of deaths currently attributable to obesity. The reduction in life expectancy attributed to obesity is not trivial. Obesity already has a negative effect on life expectancy that is larger than all accidental deaths combined (for example, those caused by accidents, homicide, and suicide), and it has the capacity (discussed in our NEJM article) to soon approach or exceed the reduction in life expectancy caused by ischemic heart disease or cancer.

The Misdirected Road to Shangri-La

In our NEJM paper, the focus was exclusively on the future of life expectancy in the United States (3). However, Tuljapurkar invokes mortality trends only from countries that conform to his line of reasoning about where he thinks life expectancy is headed--namely, Sweden and Japan. Our NEJM article was not about the future of human life expectancy in general or mortality in Sweden or Japan. Instead, we described trends in mortality and life expectancy for the U.S. population that are exactly as reported in our paper.

No matter what mortality trends occur elsewhere in the world, the fact is that life expectancy for females at age 65 in the United States has stagnated at about 19 years since 1983. Despite this evidence, the Social Security Administration (SSA) assumed, under the advice of other mathematical demographers who also invoke the mortality schedules of other countries when making assumptions about the United States, not only that the life expectancy for females at age 65 will continue to increase but also that the increase will be greater than that predicted just a few years earlier. The fact also remains that childhood obesity rose dramatically in the United States in recent years--an increase of approximately 50% per decade throughout the 1980s and 1990s (6)--and it is certain to lead to an elevated risk of death as these cohorts age.

When the SSA was found to have consistently underestimated the rise in life expectancy at older ages in their projections from 1935 to 1980 (see Figure 2 in our NEJM article), they eventually realized that their projections should be based on what the data were telling them, rather than relying on preconceived (and ill-advised) notions of an impending biological limit to life. The irony in this case is that the year in which the SSA decided to listen to the message contained within the data was coincident with the very moment that the rise in life expectancy at older ages began to level off (see Figure 3 in our NEJM article). The SSA has been repeatedly reminded by mathematical demographers of this error in past judgment and encouraged to listen to the message contained within the data. After all, observed trends are the basis for the extrapolation method of forecasting mortality and life expectancy. Yet, as current trends in life expectancy contradict their conviction that there are no biological forces that limit the duration of life, Tuljapurkar and others invoke misdirection to draw our attention away from the demographic facts before us, essentially encouraging the SSA to ignore the very data that they previously argued should be followed. If the historical record of mortality and life expectancy is going to be used to inform forecasts, it should not be invoked selectively to favor one particular viewpoint.

Specific Points Raised in the Bumpy Road

Tuljapurkar reminds the reader that large gains in life expectancy at birth have been made in all industrialized countries and have been followed recently by gains in life expectancy in most of the developing world. This is true to be sure, but he selectively fails to mention the notable deceleration in the rise in life expectancy among the longest lived subgroup of Americans--older women; he does not mention the extremely rapid recent declines in life expectancy in developing nations as a result of infectious diseases and other negative forces influencing life expectancy (7); and he may be surprised to learn that the prevalence of overweight and obesity among children (8, 9) and adolescents (10) has reached epidemic proportions (11) in Sweden in the past two decades--thus raising concerns about whether their life expectancy increases can be sustained much longer.

Tuljapurkar invokes the mortality schedule of Sweden from 1751 to 2003 (contained in his Figure 1) as evidence of the historic pattern of decline in mortality and then, based on this observation, states that "Olshansky et al. assert that this pattern necessarily limits increases in e0 [life expectancy at birth] over the coming decades." We make no such assertion--the focus of our discussion is on the mortality pattern of the United States, not Sweden or any other country.

Tuljapurkar states "When Olshansky et al. say that 'another quantum leap in life expectancy can only occur if the future is different from the past'..., they are decrying possible increases in e0 at the rate of the past 50 years, which was perhaps 0.22 years per calendar year." This is another instance of misdirection. Our statement has nothing to do with the rate of increase in life expectancy and has everything to do with how those gains will be achieved. In fact, we invite you to read the original quote from our NEJM article and judge for yourself how fairly it was portrayed by Tuljapurkar: "Given that past gains in life expectancy have largely been a product of saving the young, and since future gains must result from extending life among the old, another quantum leap in life expectancy can occur only if the future is different from the past" (p. 1139).

Tuljapurkar states that actuaries and demographers have advised the SSA to use long historical periods for their forecasts to avoid the myopia caused by looking too closely at a particular short period of time. He then goes on to state: "However, Olshansky et al. are firmly myopic when they choose to extrapolate--thus, they criticize the [SSA] actuaries for not being worried that the rise in U.S. life expectancy appeared to have stalled in the 1990s." The irony of this statement should not go unnoticed. Our extrapolation, as Tuljapurkar refers to it, is not based on extending historical trends in obesity into the future but rather on dramatic increases in childhood obesity that have been observed over the past 30 years. These children are already here (Phase I), and the health consequences of their obesity has begun to emerge (Phase II). Most important, the rise of mortality expected for these cohorts as they live into their third through sixth decades of life (Phase III) is invisible to the extrapolation methods (based on period measures of mortality) on which Tuljapurkar and others rely so heavily.

Tuljapurkar appropriately notes that numerous factors influence the relative mortality risk of people within and across levels of BMI. He then goes on to state that "Olshansky et al. use obesity as a single factor to determine relative risk and assign the same optimal BMI (that is, the BMI with the lowest mortality risk) at all ages. How can we assess the likely error in such an approach?" Our estimates were based on the observed relation between BMI and age-sex-race-specific mortality rates documented in the National Health and Nutrition Examination Surveys (NHANES) data. As such, these estimates account for the numerous attributes that Tuljapurkar declares to have varying effects on mortality risk. Moreover, there are several counterarguments to his claim that obesity may signify an increase in prosperity that may counterbalance the detrimental effects of obesity. He makes this argument when he states, "the rise in obesity coincides with, indeed is made possible by, a remarkable rise in living standards that is known to contribute to the longer term reduction of absolute (as opposed to relative) mortality risk." This argument is rendered spurious by data that are based on the observed relation between BMI and death rates. Indeed, our observations about mortality trends should leave reason for concern because they have emerged during an era of increasing standard of living. Another related argument is the observation that obesity is increasing most rapidly in the most indigent segments of the population--minorities and other disenfranchised members of the population that may be disproportionately excluded from the standard of living improvements that have accompanied the recent 30-year rise in obesity prevalence.

The issue of health and obesity raised by Tuljapurkar is important for another reason. In an important paper recently published by scientists at the Centers for Disease Control and Prevention (CDC) (12), evidence indicates that in the time between NHANES I (1971-1975) and NHANES 1999-2002, the risk of death from cardiovascular diseases associated with obesity had declined. Although this report suggested that treatments for the complications associated with obesity may be somewhat effective in reducing death rates, it made no claims about how the health status of the surviving population might have changed during this time. In a twist of irony reminiscent of the expansion-of-morbidity hypothesis originally proposed by Greunberg (13) and Kramer (14) (that is, the argument that medical treatments for fatal and nonfatal diseases and disorders may extend life but that the price to be paid may be an extended period of frailty and disability), the authors of the CDC report concluded that "The net result of these phenomena may be a population that is, paradoxically, more obese, diabetic, arthritic, disabled, and medicated, but with a lower overall CVD [cardiovascular disease] risk" (p. 1873). In other words, the obese population may not die so quickly from CVD relative to earlier generations, but their overall health status might be much worse and the risk of death from other obesity-induced disorders could increase.

Tuljapurkar argues that a myriad of population characteristics (such as education, income, and place of residence) produce a multidimensional space of risk factors and that the distribution of relative mortality risk rides on a trajectory of changing average risk. He then states, "Olshansky et al. do not care about this secular drift (that is, drift taking place over a long period of time), focusing only on the distribution of individuals by obesity." Invoking data from Sweden once again, he concludes that "mortality decline is indeed a rising tide that lifts all boats." The secular drift upon which our NEJM article is based is the observed dramatic rise in the prevalence of obesity among children in the United States, particularly among minorities such as blacks and Hispanics. The Swedish data used by Tuljapurkar is misdirection in its most basic form. The rise of childhood obesity in the United States is a fact, it is continuing at an alarming pace, and its future impact on the mortality schedule of the United States will continue to remain obscure to those who insist on using data from other countries and time periods to make their case when data from the United States fail to do so.

Finally, in his section on implications, Tuljapurkar points out that "We have noted evidence that the relative mortality risk attributed to obesity actually declines with age, which greatly weakens their argument...." This is a remarkable conclusion. Tuljapurkar's point is irrelevant, because it is already taken into account by the fact that our estimates are based on the observed relation between BMI and age-sex-race-specific death rates. Moreover, we made the extremely conservative assumption that obesity had no negative effect on death rates for (i) the population under age 20 or over age 85, (ii) black males at ages 62 to 85 at a BMI of 30 and aged 67 to 85 at a BMI of 35, and (iii) black females aged 60 to 85 at a BMI of 30 and aged 67 to 85 at a BMI of 35.

The foundation of our NEJM article is not extrapolation, as suggested by Tuljapurkar--it is observation. Life expectancy gains in the United States have been stagnant for older females and slow but steady for males. Childhood obesity in the United States has climbed at an alarming rate over the past three decades, and this trend will substantially influence the future course of life expectancy in our country if something is not done about it. Extrapolation models such as those advocated by Tuljapurkar and others remain blind to these trends because they rely on historic patterns in period death rates that cannot capture the impending life-shortening effect of childhood obesity. The epitome of this model failure is the SSA's forecast of death rates from diabetes. Despite overwhelming evidence to the contrary, they project that death rates for diabetes will begin to decline in the year 2010 and sustain this decline for the remainder of the century (15).


The ongoing debate about the future of life expectancy in the United States is important because of its effects on public policy. There are many points of contention here, all of which deserve to be discussed openly in the scientific literature. Fortunately, SSA forecasts are redone annually, which means they can easily capture new developments--whether positive or negative--that might influence forecasts of life expectancy. It is also important to emphasize that the anticipated negative effect of childhood obesity on death rates in the coming decades need not occur if the public health community combats obesity as effectively as it has other behavior-driven health problems such as smoking and drunk driving. Although there is uncertainty in any forecast, the uncertainty in this case is not based on mortality trend data from other countries or less relevant historical trends in period data but on the hope that public health will deal with this issue as effectively as it has with others.

The rise of childhood obesity and its effect on mortality and life expectancy is a new trend that has not been detected by recent SSA forecasts, and it is our view (supported by experts in the field) (16; see Mizuno Review) that the alarming rise of childhood obesity will have a major impact on death rates in the future unless we intervene.

June 1, 2005
  1. S. Tuljapurkar, Future mortality: A bumpy road to Shangri-La? Sci. Aging Knowl. Environ. 2005 (14), pe9 (2005).[Abstract/Free Full Text]
  2. S. Tuljapurkar, N. Li, C. Boe, A universal pattern of mortality decline in G7 countries. Nature 405, 789-792 (2000).[CrossRef][Medline]
  3. S. J. Olshansky, D. Passaro, R. Hershow, J. Layden, B. A. Carnes, J. Brody, L. Hayflick, R. N. Butler, D. B. Allison, D. S. Ludwig, A potential decline in life expectancy in the United States in the 21st century. N. Engl. J. Med. 352, 1138-1145 (2005).[CrossRef][Medline]
  4. K. R. Fontaine, D. T. Redden, C. Wang, A. O. Westfall, D. B. Allison, Years of life lost due to obesity. JAMA 289, 187-93 (2003).[CrossRef][Medline]
  5. S. J. Olshansky, D. Passaro, R. Hershow, J. Layden, B. A. Carnes, J. Brody, L. Hayflick, R. N. Butler, D. B. Allison, D. S. Ludwig, Peering into the future of American longevity. Discovery Med. 5, 130-134 (2005).
  6. K. M. Flegal, M. D. Carroll, C. L. Ogden, C. L. Johnson, Prevalence and trends in obesity among US adults, 1999-2000. JAMA 288, 1723-1727 (2002).[CrossRef][Medline]
  7. The Impact of AIDS (United Nations Population Division, ESA/P/WP.185, New York, 2003).
  8. B. Littorin, L. Nystrom, B. Gullberg, L. Rastam, J. Ostman, H. J. Arnqvist, E. Bjork, G. Blohme, J. Bolinder, J. W. Eriksson et al., Increasing body mass index at diagnosis of diabetes in young adult people during 1983-1999 in the Diabetes Incidence Study in Sweden (DISS), J. Intern. Med. 254, 251-256 (2003).[CrossRef][Medline]
  9. S. Petersen, C. Brulin, E. Bergstrom, Increasing prevalence of overweight in young schoolchildren in Umea, Sweden, from 1986 to 2001, Acta Paediatr. 92, 848-853 (2003).[CrossRef][Medline]
  10. F. Rasmussen, M. Johansson, H. O. Hansen, Trends in overweight and obesity among 18-year-old males in Sweden between 1971 and 1995, Acta Paediatr. 88, 431-437 (1999).[CrossRef][Medline]
  11. K. Sundquist, J. Qvist, S. E. Johansson, J. Sundquist, Increasing trends of obesity in Sweden between 1996/97 and 2000/01. Int. J. Obes. Relat. Metab. Disord. 28, 254-261 (2004).[Medline]
  12. E. W. Gregg, J. C. Yiling, B. L. Cadwell, G. Imperatore, D. E. Williams, K. M. Flegal, K. M. V. Narayan, D. F. Williamson, Secular trends in cardiovascular disease risk factors according to body mass index in US Adults. JAMA 293, 1868-1874 (2005).[CrossRef][Medline]
  13. E. M. Greunberg, The failures of success. Milbank Mem. Fund Q. Health Soc. 55, 3-24 (1977).[CrossRef][Medline]
  14. M. Kramer, The rising pandemic of mental disorders and associated chronic diseases and disabilities. Acta Psychiatr. Scand. Suppl. 285, 382-397 (1980).
  15. F. C. Bell, M. L. Miller, Life tables for the United States Social Security Area 1900-2100 (Actuarial Study No. 116, Social Security Administration, Washington, DC, 2002).
  16. T. Mizuno, I. Shu, H. Makimura, C. Mobbs, Obesity over the life course. Sci. Aging Knowl. Environ. 2004 (24), re4 (2004). [Abstract/Free Full Text]
  17. We thank A. Viidik for insightful comments on earlier drafts of this manuscript.
Citation: S. J. Olshansky, B. A. Carnes, R. Hershow, D. Passaro, J. Layden, J. Brody, L. Hayflick, R. N. Butler, D. B. Allison, D. S. Ludwig, Misdirection on the Road to Shangri-La. Sci. Aging Knowl. Environ. 2005 (22), pe15 (2005).

Science of Aging Knowledge Environment. ISSN 1539-6150