Are Modern Environments Really Bad for Us Revisiting the Demographic and Epidemiologic Transitions.код для вставкиСкачать
YEARBOOK OF PHYSICAL ANTHROPOLOGY 48:96–117 (2005) Are Modern Environments Really Bad for Us?: Revisiting the Demographic and Epidemiologic Transitions Timothy B. Gage* Department of Anthropology and Department of Epidemiology, University at Albany-SUNY, Albany, New York 12222, and Department of Genetics, Southwest Foundation for Biomedical Research, San Antonio, Texas 78284 KEY WORDS demographic transition; epidemiologic transition; mortality; cause of death ABSTRACT It is a common assumption that agriculture and modernization have been detrimental for human health. The theoretical argument is that humans are adapted to hunter-gatherer lifestyles, and that the agricultural and \modern" environments are novel and hence likely to be detrimental. In particular, changes in nutrition, and population size and distribution with the adoption of agriculture, are considered to increase the risk of infectious disease mortality. Similarly, changes due to modern lifestyles, notably changes in nutrition, smoking, exercise, and stress, are thought to be associated with an increased risk of degenerative disease mortality in the industrial environment. This paper reviews the available literature on the history and prehistory of total mortality (the demographic transition) and cause of death (the epidemiologic transition), and ﬁnds that neither agriculture nor modernization is associated with increases in mortality, i.e., declines in health. First, mortality does not appear to have increased during the transition to agriculture, or during the early phases of the industrial revolution. Clearly, infectious diseases have declined with modernization. Second, the empirical data, when uncorrected for INTRODUCTION Evolutionary medicine was ﬁrst proposed as a uniﬁed discipline in the early 1990s (Nesse and Williams, 1994). A basic tenet of the discipline is that both infectious agents and human hosts are subject to evolutionary processes, which have largely been ignored in the mainstream literature. A case in point is the classic work of McKeown (1976), concerning the decline in infectious disease mortality over the past 150 years. He dismissed any role that the evolution of virulence might have played in this decline, largely on the basis that there were few data to demonstrate such changes (McKeown, 1976). Evolutionary medicine has advanced on several fronts: advances in molecular biology have documented the phylogeny and age of some human pathogens (Holmes, 1999, 2004; Van Blerkom, 2003), advances in evolutionary ecology have revolutionized the concept of host/parasite coevolution and shown that the evolution of virulence is considerably more complex than previously thought (Antia et al., 2003; Bull, 1994; Ebert, 1999), and anthropology has contributed evidence concerning human evolution and changing lifestyles throughout human history (Cohen, 1989; Cohen and Armelagos, 1984; Eaton et al., 1988; Strassmann and Dunbar, 1999). One of the predominant themes of evolutionary thinking and evolutionary medicine in particular is that C 2005 V WILEY-LISS, INC. misclassiﬁcation of cause of death, do suggest an increase in degenerative disease mortality, at least until the mid 20th century, when these causes of death clearly began to decline. All studies that correct for misclassiﬁcation of cause of death, however, ﬁnd that the general decline in degenerative disease mortality began much earlier, perhaps as early as the 1850s in the developed countries. This is about the same time that infectious disease mortality began to decline in these countries. The exception is neoplasms, which increased with modernization until quite recently. Part of the increase in neoplasms may be attributable to increases in smoking during the course of modernization. Nevertheless, the overall risk of degenerative disease mortality appears to have declined with modernization. The fact that the decline in the risk of infectious disease mortality, and the decline in risk of degenerative disease mortality, are largely coordinated suggests that the causes of both declines may be related. Historical trends in morbidity, and potential causes of the decline in infectious and degenerative disease mortality, are brieﬂy considered. Yrbk Phys Anthropol 48:96–117, 2005. C V 2005 Wiley-Liss, Inc. when organisms are introduced to novel environments, the health of the organism declines (Nesse and Williams, 1999; Stearns, 1999). The view that novel environments are detrimental to health is widely held in biological anthropology, and pervades biological anthropology’s contributions to evolutionary medicine (Cohen, 1989; Cohen and Armelagos, 1984; Eaton and Eaton, 1999; Eaton et al., 1988; Strassmann and Dunbar, 1999). For example, it is argued that the development of agriculture, a novel environment for hunter-gatherers, negatively affected human health (Cohen, 1989; Cohen and Armelagos, 1984). The theoretical basis of this argument is twofold: ﬁrst, that agriculture changed the diet, leading to poorer nutrition (Eaton et al., 1988); and second, that increases in population density and distribution changed the ecology of infectious agents, leading to increased disease loads (Fenner, *Correspondence to: Timothy B. Gage, Department of Anthropology, AS 114, University at Albany-SUNY, Albany, NY 12222. E-mail: firstname.lastname@example.org DOI 10.1002/ajpa.20353 Published online in Wiley InterScience (www.interscience.wiley.com). ARE MODERN ENVIRONMENTS REALLY BAD FOR US? 1970). The empirical evidence, on the other hand, is contested. Some argued that the skeletal data indicate a deterioration in health (Cohen and Armelagos, 1984; Barrett et al., 1998). Others argued that the interpretation of the data by these investigators was ﬂawed and that the data could just as easily represent a period of improvement in health (Gage, 2000; Strassmann and Dunbar, 1999; Wood et al., 1992). Similarly, modern/industrialized/ Westernized/cosmopolitan environments (i.e., another novel human environment) are also considered to negatively inﬂuence health, although everyone agrees that longevity has dramatically increased due to the decline in infectious disease mortality. Again, the theoretical basis for this argument is changes in diet, and changes in exercise levels. These changes in diet and exercise are widely considered to \promote the development" (Eaton and Eaton, 1999) of deaths due to the \diseases of afﬂuence" (Eaton and Eaton, 1999), \diseases of civilization" (Eaton and Eaton, 1999), and \degenerative and manmade diseases" (Omran, 1977), particularly the \epidemic" (Barker, 1989, 1999; Maynard Smith et al., 1999) of cardiovascular disease. The empirical evidence, however, is again contested. There is increasing evidence that the degenerative diseases as a group, and cardiovascular disease in particular, have declined over the last century (Gage, 1993, 1994; Preston, 1976), despite the emergence of novel modern environments and lifestyles. Perhaps modernization is not as bad for health as some suggest. The aim of this paper is to review the empirical evidence concerning historical trends in mortality and cause of death. The trends with respect to the advent of agriculture are brieﬂy examined. However, the main focus of this paper concerns the last 100–150 years in the currently modernized/industrialized/Westernized/cosmopolitan countries. The primary intent is to provide an empirical basis for evaluating how modernization, industrialization, Westernization, and cosmopolitan lifestyles may have inﬂuenced health. Modernization, industrialization, Westernization, and cosmopolitan have not been well-deﬁned terms. For the purposes of this paper, these terms will be deﬁned loosely as the secular trends observed in countries that are generally agreed upon as modern, industrial, Western, cosmopolitan countries, e.g., Western Europe and United States. For brevity, the terms modern and modernization will be used when referring to these societies and processes. \Health" will be deﬁned as expectation of life or age-standardized death rates. As expectation of life increases and age-standardized death rates decline, \health" improves. Mortality is, of course, only one possible deﬁnition of health. Morbidity is also an important component of health, although considerably less is known about morbidity. The speciﬁc aims are to empirically document the demographic transition, i.e., trends in total mortality, and the epidemiologic transition, i.e., trends in cause of death, and to brieﬂy examine secular trends in morbidity. Finally, the paper concludes with a short review of the vast literature concerning why infectious causes of death declined, and a short discussion of some hypotheses as to why degenerative causes of death declined. SOURCES OF EVIDENCE Documenting secular trends in mortality and morbidity would seem to be a relatively simple task. However, this has not proven to be so. The earliest nation to begin collecting mortality data was Finland in 1722, followed shortly thereafter by the remaining Scandinavian coun- 97 tries. The other European countries did not follow suit until later, well after 1800 (Vallin, 1991). Data from the rest of the world’s nations are available much later or not at all. A second source of data is from family reconstitution studies. This consists of linking birth, marriage, and death records from liturgical or other records. These data are best developed for England and Wales (Wrigley and Schoﬁeld, 1981), and are available from 1541 or so. Data from prehistoric periods are only available from skeletal and other archaeological evidence. Consequently, the history of mortality is of necessity biased in favor of the Western European experience, and limited to a relatively small number of data sets. The mortality data to illustrate this paper, and the basis for many of the papers reviewed here, come primarily from four sources. The life tables for Sweden are available from the Human Mortality Database (2004; revised as of January 6, 2005) for each year from 1751–2003, and are unabridged. Information on age-standardized trends in cause of death in the United States from 1900–1998 is available from the Center for Disease Control (CDC)/ National Center for Health Statistics (NCHS) National Vital Statistics Systems. Life tables decremented by cause of death for an international sample are taken from Preston et al. (1972). Archaeological data concerning mortality are adapted from Gage (1988, 2000) and Lovejoy (1977). Perhaps the most important data set for the historic period is the sample by Preston et al. (1972). They compiled and examined cause of death in 165 populations of each sex, representing 43 countries. The vast majority of these countries were European or of European extraction, e.g., Canada and the United States. However, a few non-European countries with adequate cause-of-death data were included: Chile, Colombia, Costa Rica, El Salvador, Guatemala, Japan, Mexico, Panama, Taiwan, Trinidad and Tobago, and Venezuela. England and Wales represent the longest series of life tables by cause of death, and are available in Preston et al. (1972), from 1861–1964. England and Wales was the ﬁrst country to keep cause-of-death records. All causes of death in this data set are classiﬁed into 12 categories, based on the International Classiﬁcation of Diseases, (World Health Organization, 1955). The categories that are of concern here are deﬁned in Table 1. These causes of death can be loosely classiﬁed into ﬁve primarily infectious and three primarily degenerative disease categories, plus an unknown cause of death category. However, the standard classiﬁcation conventions are not designed to distinguish infectious from noninfectious disease, at least for the broader categories reported here. Consequently, the division into infectious and degenerative deaths is not completely accurate. For example, rheumatic fever is classiﬁed as a cardiovascular disease by the ICD (World Health Organization, 1955). Thus some infectious disease is included in the three categories assumed to represent degenerative causes of death. These problems are minor, and are not considered to detrimentally affect general conclusions (Preston, 1976). It is standard procedure with demography to question the observed empirical trends in demographic data to determine if the patterns observed could be a result of \bad data." The manner in which demographic data are collected, even by modern nations, is prone to error. National data collection systems and/or liturgical records may not collect information on the entire population (by age, region, economic, social, and/or religious status) consistently over 98 T.B. GAGE TABLE 1. Classiﬁcation of deaths by cause1 Title Infectious causes of death Respiratory tuberculosis Other infectious and parasitic diseases Deﬁnitions B1 B2–17 Respiratory tuberculosis Tuberculoses (other forms), syphilis, typhoid, cholera, dysentery, scarlet fever, diphtheria, whooping cough, meningococcal infections, plague, polio, smallpox, measles, typhus, malaria, and all other infectious diseases Inﬂuenza, pneumonia, and bronchitis Inﬂuenza, pneumonia, and bronchitis Diarrhea B30–32 Certain diseases of infancy B42–44 Degenerative causes of death Neoplasms Cardiovascular Certain degenerative diseases Miscellaneous causes of death Other and unknown causes of death 1 ICD7 codes B36 Gastritis, duodinitis, enteritis, and colitis (except diarrhea of newborn) Birth injuries, infections of newborn, and other diseases due to infancy and immaturity B18–19 B22, B24–29 A85, A86 Malignant and benign Vascular lesions, rheumatic fever and heart disease, arteriosclerosis, other diseases of heart, hypertension, and diseases of arteries and circulatory system Nephritis, nephrosis, cirrhosis of liver, ulcers of stomach and duodenum, and diabetes B20, B33, B37, B38 B46, B21, B23, B34, B35, B39, B41 All other diseases (except diseases of arteries, A85, and other diseases of circulatory system, A86), anemias, nonmeningococcal meningitis, appendicitis, intestinal obstruction and hernia, hyperplasia of prostate, and congenital malformations Adapted and abridged from Preston et al. (1972). time. Perhaps the most problematic issue for our purposes is that diagnoses of cause of death and morbidity have improved over time. Thus, observed trends in mortality could be due to improvements in data collection and diagnosis rather than true secular trends in mortality. There are also technical issues with respect to documenting trends in mortality. Crude death rates are generally not useful, because mortality varies with age, and hence changes in age structure inﬂuence mortality estimates. As a result, mortality trends are usually presented as expectations of life at birth, or as age-standardized death rates to control for changes/differences in the age structure of populations compared. Expectation of life at birth is an estimate of the average number of years a person just born can expect to live, based on a life table. It is the same as mean age at death in a stationary population. Age-standardized death rates are estimated by multiplying (for each age) the age-speciﬁc death rates estimated for a population by an arbitrary \standard" age structure (typically an observed age structure of one of the populations examined), and then summing the age-speciﬁc results. By holding the standard age structure constant in comparisons across populations and across time periods, changes in age structure are controlled. However, different standard age structures will still lead to slightly different dynamics in age-standardized death rates (due to heterogeneity in mortality among the ages). Consequently, agestandardized death rates are more difﬁcult to interpret than expectation of life. Thus, expectation of life is preferred for total mortality. Studies of cause of death have traditionally used age-standardized death rates. Methods of estimating the effects of a speciﬁc cause of death on expectation of life are available, but are more difﬁcult to calculate. Only one study reported below used the impact on expectation of life of speciﬁc causes of death. A third metric, cumulative (lifetime) hazard rates from 0–80 years of age, is also occasionally used. This is essentially a sum of age-speciﬁc death rates. It can be thought of as the total risk a person would live through if he or she survived to age 80. Alternatively, it can be thought of as an age-standardized death rate where the standard age structure has equal numbers of individuals at each age 0–80. This measure, like expectation of life, is consistent (because the standard age structure is always the same), and can also be easily applied to cause of death. THE DEMOGRAPHIC TRANSITION Thompson (1929) ﬁrst proposed the theory of the demographic transition. The \theory" of the demographic transition is a descriptive model of secular declines in mortality, fertility, and consequent population growth that occurred following the industrial revolution. Only the trends in mortality will be considered here. Like all models, the demographic transition is a simpliﬁcation of the facts. The transition is typically described as occurring in several discrete stages, but in fact, each phase trends into the next, so in reality these phases are not completely discrete. Further, the demographic transition did not occur throughout the world, within Europe, or even in geographically different areas of the same country simultaneously. Much has been written about rural and urban differentials in mortality (Woods, 1991, 2000; Wrigley and Schoﬁeld, 1981). In particular, urbanization may have slowed (or stalled) the rate of decline in mortality in England and Wales (Woods, 2000). Recently, large differentials (larger than rural-urban contrasts) were identiﬁed among rural areas of England and Wales (Dobson, 1997). Apparently, local conditions are important with respect to the details and exact timing of the demographic transition. Nevertheless, there are common characteristics and general trends with respect to secular changes in mortality stages in all of the various subdivisions of the population (Vallin, 1991). These general trends occurring over the course of modernization are presented here. ARE MODERN ENVIRONMENTS REALLY BAD FOR US? Fig. 1. Secular decline in mortality, as reﬂected by increasing life expectancy. Yearly estimates for 1751–2003 are from Sweden. Data source, Human Mortality Database, January 6, 2005. Observed secular trends in total mortality Stage I of the demographic transition is largely preindustrial, and is characterized by high mortality and fertility, with stationary or slow population growth. Detailed examinations of this period indicate that \normal" mortality was punctuated by large ﬂuctuations in mortality (crisis mortality) from one year to the next (Fig. 1). Prior to the second phase of the transition, which began ca. 1850 in Sweden (Fig. 1), crisis mortality began to decline in frequency and amplitude. However, crisis mortality continued into the 20th century, particularly with respect to inﬂuenza, which due to its biology has remained an epidemic disease (Potter, 2002). Perhaps the 1918–1919 inﬂuenza pandemic marked the end of the crisis mortality era. Early on, crisis mortality was not coordinated geographically and/or temporally (Woods, 2000). However, crises became more geographically and temporally synchronized after the turn of the 20th century (Fig. 3.2 in Vallin, 1991), ending with the aforementioned worldwide inﬂuenza pandemic of 1918–1919. Presumably, this globalization of the human infectious disease environment reﬂects improvements in transportation with modernization. Stages II and III of the demographic transition are characterized by declining \normal" mortality (Vallin, 1991). Stage III of the demographic transition involves the decline in fertility, which does not concern us here, and continued (stage II) declines in mortality. In any event, stage II–III declines in mortality began in some European countries such as Sweden and England and Wales in the mid 1800s (Figs. 1, 2). It began in less developed countries worldwide about 1920, e.g., Chile (Fig. 2). Stage II–III declines in mortality appear to have been completed around World War II in the developed nations, but continue in less developed countries. However, the process in less developed countries has been accelerated. Thus, worldwide mortality appears to be converging on low mortality rates (Fig. 2) (Wilson, 2001). Stage IV is characterized by the low mortality observed in developed countries after World War II (Vallin, 1991). Nevertheless, mortality continues to decline, albeit more slowly. The slowing of the decline is thought to be due to the fact that infant, childhood, and young adult mortality have reached a minimum and will not decline further. Most 99 of the recent declines have occurred among the elderly. Age-speciﬁc declines are discussed in more detail below. Prehistoric mortality tends to be similar to, or higher than that observed during the historic stage I period. Expectation of life derived from prehistoric life tables ranges from about 18–25 years (Gage, 2000), slightly below the underlying expectation of life of about 35 years experienced in Sweden in the late 1750s (Fig. 1). The highest recorded prehistoric expectations of life are 35 years, reported for the Mediterranean (Acsadi and Nemeskeri, 1970). It is possible that underestimation of age at death might cause an underestimation of expectation of life in these prehistoric life tables. Better methods of aging skeletal data are needed and are actively being researched (Hoppa and Vaupel, 2002). However, omission of infant skeletons due to poorer preservation and other taphonomic issues tend to overestimate expectation of life in prehistoric life tables. Thus there is no good evidence suggesting a large increase or decline of mortality just prior to or during the early industrial revolution. Similarly, there is no convincing empirical evidence that mortality increased with the agricultural revolution. Mean expectations of life for a sample of prehistoric life tables classiﬁed by mode of production indicate that hunter-gatherers had a mean expectation of life of 21.6 years (standard deviation, 2.1), compared with horticulturalists with a mean of 21.2 years (standard deviation, 3.9) and agriculturalists with a mean of 24.9 (standard deviation, 8.5). Although agriculturalists have the higher mean expectation of life (lowest mortality), none of the differences are statistically signiﬁcant (Gage, 2000). The view that health deteriorated with the agricultural revolution is based on the ﬁnding that most longitudinal series of life tables from the same localities indicate a decline in expectation of life with agriculture, particularly the Neolithic. Further, there is an increase in skeletal lesions across the archaeological series (Barrett et al., 1998; Cohen and Armelagos, 1984). The problem with these observations is that if population growth accelerated with the agricultural revolution, as was also argued by Cohen and Armelagos (1984), then estimates of mean age at death and expectation of life are necessarily biased downward for agricultural populations compared to hunter-gatherers (Gage, 2000; Gage et al., 1989; Moore et al., 1975; Strassmann and Dunbar, 1999; Wood et al., 1992). Further, the increases in skeletal lesions can be interpreted as representing improvements in mortality (Wood et al., 1992) as justiﬁably as declines in health (Cohen and Armelagos, 1984). Overall, the empirical evidence does not support the argument that the agricultural revolution is associated with a decline in health (Gage, 2000; Strassmann and Dunbar, 1999; Wood et al., 1992). In fact, the prehistoric empirical evidence suggests high levels of mortality similar to, or slightly higher than mortality for stage 1 of the demographic transition (Gage, 2000). Observed secular trends in age-speciﬁc total mortality As total mortality declines during the historic period, risk declines at all ages. This is reﬂected in the construction of the model life tables of Coale et al. (1983). The age-speciﬁc mortality curves at lower expectations of life (higher levels of mortality) nest inside and above the age-speciﬁc mortality curves at higher expectations of life (lower levels of mortality) (Fig. 3A). Lee and Carter (1992) proposed a formal relationship between differences in level of mortality (expectation of life) and differen- 100 T.B. GAGE Fig. 2. Secular decline in mortality measured as expectation of life at 10-year intervals for developed nation (England and Wales) and developing nation (Chile). Data source is Preston et al. (1972). Figure reprinted from Gage (2000). ces in age-speciﬁc risk of death. They hypothesized that mortality declines are a constant function of the log of risk, i.e., age-speciﬁc mortality curves displayed on a log of risk scale should be equidistantly apart at all ages. Whether the hypothesis of Lee and Carter (1992) is precisely correct or not, secular declines in mortality have occurred at all ages, and the largest are for those ages with the highest risk (Figs. 3, 4). In Sweden, age-speciﬁc mortality was relatively constant until about 1851 (Fig. 3). It is possible that childhood mortality declined between 1801–1851. However, in other respects, the curves are remarkably similar until 1851. Based on absolute measures of risk (Fig. 3A), as opposed to log of risk, infant and childhood mortality declined moderately, and adult mortality dropped signiﬁcantly, between 1851–1901. From 1901–1951, infant mortality declined signiﬁcantly, while childhood through elderly mortality continued to decline. From 1951–2001, infant mortality continued to decline, but the largest declines occurred at ages of 50 years and above. Currently, there is little room for infant mortality to decline further. Future declines must occur among the elderly or not at all. Overall, infant and elderly mortality declined the most. The picture of the decline looks rather different when graphed as a log of risk, the standard convention for presenting age-speciﬁc mortality graphically (Fig. 3B). In this case, mortality appears to decline the most in the childhood and young adult years. Log of risk measures the decline in mortality in terms of relative decline in risk. For example, an age group that declined from a risk of 0.2 to 0.1, i.e., by a factor of 0.5, is considered to have declined by the same amount as an age group that declined from a risk of 0.8 to 0.4, also by a factor of 0.5. On the other hand, the decline in absolute risk is 0.1 and 0.4, respectively. The childhood ages experienced the largest relative decline. However, infants and the elderly experienced the largest absolute declines in mortality. With respect to the effects of modernization on health, absolute risk would appear to be the more informative metric. The age patterns of risk of mortality for earlier (prehistoric) periods depend on archaeological data. Care must be taken when interpreting these results. However, a comparison of the Libben life table (Lovejoy et al., 1977, as smoothed by Gage, 1988) with Sweden in 1751 indicates characteristic differences between archaeological life tables and national life tables (Gage, 1988, 2000) (Fig. 4A,B). These suggest similar levels of mortality until age 5 years, after which mortality is signiﬁcantly higher in the paleodemographic life table. The higher mortality at ages greater than 5 years, of course, reﬂects differences in expectation of life between the two life tables (approximately 38 years for Sweden, and 20 years for Libben). As mentioned above, the surprisingly low mortality in the Libben life table at the youngest ages is typically thought to be a result of poor preservation of infant remains. However, if this is true, then the expectation of life of 20 years is an overestimate, and the Libben mortality was high indeed, all other things being equal. The high mortalities at the oldest ages (rapid rate of aging) characteristic of paleodemographic life tables (Gage, 1988, 2000) are often attributed to misspeciﬁcation of age. Howell (1982) argued that the Libben life table is critically ﬂawed, because the implied age structure does not resemble observed age structures, even among contemporary hunter-gatherers. However, \high" mortality begins well before 20 years of age, at which age, estimation of age should be relatively accurate. Further, what is particularly interesting about Figure 4B is that above age 5 years, the difference between the Libben and Swedish life tables is relatively constant using the log of risk metric, i.e., the hypothesis of Lee and Carter (1992) appears to hold at ages above 5 years (Fig. 4B). Mortality may increase slightly more rapidly than in the hypothesis of Lee and Carter (1992) at older ages in the Libben life table (due to age misestimation?), but not radically. Thus the relationship between level of mortality and age-speciﬁc risk of death, thought to hold for contemporary populations, appears to hold for this ARE MODERN ENVIRONMENTS REALLY BAD FOR US? 101 Fig. 3. Age-speciﬁc mortality rates for Sweden, 1751, 1801, 1851, 1901, 1951, and 2001, indicating secular changes in contribution of different ages to decline in mortality. A: Data on absolute scale. B: Same data, based on log of mortality. Data source, Human Mortality Database, January 6, 2005. paleodemographic/early historic comparison. Perhaps the Libben life table is not as ﬂawed as is generally assumed. A second view of the association of historical declines in mortality with declines in age-speciﬁc risk is provided by Gage (1993, 1994). This method uses a competing hazard model (a kind of parametric mixture model) that divides mortality into three competing components, simi- lar to the named causes of death (Fig. 5). In particular, the Siler Model (1979) divides risk of death into immature mortality (which declines with age), residual mortality (which is constant with respect to age), and senescent mortality (which increases with age). The immature and senescent components represent \endogenous" mortality, while the residual component represents \exoge- 102 T.B. GAGE Fig. 4. Age-speciﬁc mortality rates for Sweden, 1751, and Libben, a North American Late Woodland population (hunter-gatherer) ca. 800–1100 AD. A: Data on absolute scale. B: Same data, based on log of mortality. Data sources, Sweden (Human Mortality Database 6 January 2005); Libben life table reported by Lovejoy et al. (1977), smoothed by Gage (1988). ARE MODERN ENVIRONMENTS REALLY BAD FOR US? 103 enza pandemic (Gage, 1993, and below). The decline in residual (exogenous) mortality is theoretically consistent with stage I declines in crisis mortality (epidemic mortality). From 1890–1920, both immature and senescent mortality declined rapidly. From 1920 onward, immature and residual (exogenous) mortality declined, although more slowly than during previous periods. Exogenous mortality became immeasurably small after 1951. Senescent mortality increased a little between 1921–1931, and a reasonable amount between 1931–1941, before declining. Thus stage II–III mortality declines include all three components, while stage IV declines are limited to immature and senescent risks. The point is that, in general, mortality has declined at all ages over the course of the demographic transition. Senescent mortality did not decline monotonically, but still declined substantially. Summary: demographic transition Fig. 5. Graphical depiction of model of age-speciﬁc mortality (Siler 1979). Model is sum of three components of mortality. Area \a" is immature component representing decline in mortality immediately following birth, area \b" is senescent component, representing increase in mortality with age, and bottommost area \c" is residual component representing age-independent, exogenous risks. From Gage (1991). nous" mortality, as deﬁned in the evolution of life-history literature (Stearns, 1992). Gage (1994) examined the cumulative (lifetime) risk of mortality at ages 0–80 with respect to expectation of life in the sample by Preston et al. (1972) of cause-speciﬁc life tables (Fig. 6). Cumulative or lifetime risk is the sum total (integral) of risks at each age experienced by someone who lives, in this case, to the age of 80. For example, cumulative immature risk is the area labeled \a" in Figure 5. Total cumulative risk would equal the sum of areas a–c in Figure 5. Altogether, about 50% of the decline in total cumulative risk of death across this sample is due to declines in senescent mortality. A little more than 25% of the decline in the cumulative hazard is due to declines in immature mortality. The remaining 25% declines as a constant across all ages. This suggests that the risks of senescent mortality have declined more than any other component of mortality. Of course, this decline is spread across a larger number of age categories than is immature mortality. The ﬁgures vary slightly by sex. Females tend to have slightly higher residual and slightly lower senescent mortality than males, given identical expectations of life. This difference is thought to be largely due to maternal mortality (Gage 1994), which occurs in the residual component of female life tables. Since expectation of life is ﬁxed (controlled for) in these analyses, senescent or immature mortality must compensate. In this case, it is senescent mortality that compensates. Immature mortality is relatively equal in male and female life tables of the same expectation of life. Keep in mind that this analysis gives equal weight to every age up to age 80, whereas conventional measures, such as expectation of life and age-standardized death rates, tend to give greater weight to younger than older ages, as described above. Gage (1993) presented a similar analysis of the historical trends in England and Wales from 1861–1964 (Fig. 7). Exogenous mortality is the ﬁrst component to decline, with rapid declines from 1871–1891. During this period, senescent mortality increases, probably due to the 1890 inﬂu- Does the historical decline in mortality mean that \modernization" has been good for human health? Clearly, total mortality has experienced a secular decline over the last 300 years in the developed nations. Mortality prior to the historic period is consistent with or perhaps slightly higher than mortality (lower expectation of life) during stage I of the demographic transition. There is no evidence that mortality increased during the early phases of the industrial period, only beginning to decline in the later phases of the demographic transition (Fig. 1). If this is the deﬁnition of improved health, then modernization and modern lifestyles have clearly beneﬁted health overall. On the other hand, no one argues that the decline in infectious disease mortality accompanied modernization, and that in this regard health has improved. The argument that modernization is bad for health typically concerns the secular trends in degenerative diseases, e.g., cancers, cardiovascular diseases, or diabetes. It is clear from the evidence presented above that as mortality declined, the risk of death declined at all ages, including among the elderly, where the degenerative diseases are concentrated. However, these declines among the elderly could be due to infectious diseases of the elderly such as inﬂuenza, pneumonia, and bronchitis. Consequently, the risk of degenerative disease mortality may have declined, remained the same, or even increased. It is the assumption that degenerative causes of death increased with modernization that has given rise to the concept that modern lifestyles are bad for health. Below, I review our knowledge of the secular trends in cause of death. THE EPIDEMIOLOGIC TRANSITION Trends in cause-speciﬁc mortality are even harder to document and interpret than trends in overall mortality. England and Wales was the ﬁrst country to formally record cause of death, beginning in 1830. Very little empirical information on cause of death is available prior to 1830 in any population. As with total mortality, there are difﬁculties in examining secular trends in cause of death. First, there have been changes in the conventions for reporting diseases, which affect the enumeration of cause of death. The International Classiﬁcation of Diseases (ICD) is currently in its tenth revision. The transition from one revision of the ICD to the next is often not completely consistent. For example, beginning with the eighth revision of the ICD, cardiovascular-renal disease was categorized with general cardiovascular deaths, whereas in previous revisions, cardiovascular-renal disease was 104 T.B. GAGE Fig. 6. Decline in component-speciﬁc mortality as total mortality declines in sample of Preston et al. (1972). Results presented here statistically control for period (i.e., calendar year) effects. From Gage (1994). not included in general cardiovascular deaths. Consequently, some data adjustment is usually necessary to compile consistent data sets across time. These, however, are intentional adjustments in classiﬁcation of cause of death. There has also been a secular improvement in classiﬁcation of cause of death, as medical doctors and examiners become better trained and more knowledgeable at diagnosing cause of death. It should be pointed out that classiﬁcation in the late 20th century was still far from perfect (Manton and Stallard, 1984), although earlier periods were far worse. Finally, application of the ICD classiﬁcation conventions is also thought to vary (culturally) among nations (Preston, 1976). However, even if the underlying data are accurate, there remain technical difﬁculties in interpretation. It is as important with cause-of-death data as with total mortality that methods account for age-speciﬁc risks of death. A larger proportion of individuals die of degenerative diseases today than in the past. This does not necessarily mean that the risk of degenerative death has increased. The reduction or elimination of one cause of death will, all other things being equal, increase the number and proportion of deaths attributed to the causes that remain. However, the age-speciﬁc risk of the remaining causes may not have increased. Individuals may simply survive longer and die at an older age. Originally, the epidemiologic transition was deﬁned in terms of the rank ordering of deaths by cause (Omran, 1977). Omran (1977) showed that degenerative diseases replaced infectious diseases as the most common causes of death over the last few centuries. The CDC still reports the rank order of common causes of death. Such information has important applications. For example, research dollars might be allocated on the basis of frequency of causes of death. On the other hand, to show that modernization is bad for health, i.e., that modernization exacerbates the degenerative causes of death, it is necessary to demonstrate that the risk of degenerative disease mortality increased with modernization. Consequently, the description of the epidemiologic transition presented below is based on changes in risk of degenerative disease mortality. In particular, it is argued that the empirical evidence suggests that the risk of degenerative death overall has declined with modernization. However, the decline in infectious disease mortality was larger than the decline in degenerative disease mortality. Hence degenerative diseases have become more common than infectious diseases as causes of death with modernization, even though both types of cause have declined. Epidemiologic transition: 1861–1964 The age-standardized observed secular trends in total infectious, total degenerative, and other and unknown causes of death for England and Wales 1861–1964 are shown in Figure 8. Overall, infectious diseases decline, while degenerative diseases increase slightly. The category of deaths named \other and unknown diseases" also declines. Although other and unknown diseases do include inborn errors of metabolism, this category consists predominantly of diseases assigned to causes of death such as \senility," which have no modern counterpart in the ICD. Hence this category represents predominantly misclassiﬁed deaths (Table 1). These misclassiﬁed ARE MODERN ENVIRONMENTS REALLY BAD FOR US? 105 Fig. 7. Decline in component-speciﬁc mortality of Siler (1979) in England and Wales, 1861–1964. Data source, Preston et al. (1972). From Gage (1993). Fig. 8. Decline in total infectious, degenerative, and other and unknown causes of mortality (as deﬁned in Table 1) for England and Wales, 1861–1964. Data source, Preston et al. (1972). deaths decline as classiﬁcation procedures improve, shifting deaths to other named categories. Thus the trends in the named causes of death are biased. The declines in other and unknown diseases are about half as large as the decline in named infectious diseases, and much larger than the observed increases in degenerative diseases. Thus the observed secular increase in degenerative disease mortality could be due to improvements in diagnosis of cause of death. Degenerative diseases could have increased, remained the same, or even declined. All of the infectious disease categories clearly decline (Fig. 9). However, the trends in degenerative diseases are less consistent (Fig. 10). Neoplasms ﬂuctuate, but appear to increase. Cardiovascular deaths ﬂuctuate over time, and may also increase slightly. Note that in England and Wales, neoplasms exceed cardiovascular mortality. This 106 T.B. GAGE Fig. 9. Decline in named infectious, and other and unknown vauses of mortality (as deﬁned in Table 1) for England and Wales, 1861–1964. Data source, Preston et al. (1972). Fig. 10. Decline in named degenerative, and other and unknown causes of mortality (as deﬁned in Table 1) for England and Wales, 1861–1964. Data source, Preston et al. (1972). differs from the United States, where cardiovascular deaths traditionally exceed neoplasms, possibly due to cultural differences in the application of the ICD classiﬁcation conventions (Preston, 1976). The remaining degenerative diseases increase and then decline. The highs and lows in neoplasms and cardiovascular deaths follow the same trends as the infectious-category inﬂuenza, pneumonia, and bronchitis (Fig. 9). Excess deaths due to other causes of death are often correlated with inﬂuenza epidemics (Azambuja and Duncan, 2002; Davenport, 1976; Lancaster, 1990). One possible explanation for the excess mortality is that the risk of neoplasm and cardiovascular disease mortality increases in individuals with inﬂuenza due to the added stress. Consequently, an excess of neoplasm and cardiovascular disease deaths is likely during an inﬂuenza epidemic, depleting the number of individuals particularly susceptible to neoplasms and cardiovascular disease deaths. This would then be followed by a period of lower-than-expected neoplasm and cardiovascular disease deaths, while the population of susceptibles increases once again. The high neoplasm and cardiovascular death rates for England and Wales in 1891 are associated directly with the inﬂuenza epidemic that spans the 1890s, while the relatively low neoplasm and cardiovascular death rates reported in 1921 occurred a year or so after the end of the 1918–1919 inﬂuenza pandemic. This might explain the ﬂuctuations in neoplasm and cardiovascular disease deaths from 1861–1964. Taken together, however, the observed trends in the named degenerative diseases tend to increase over time, at least until the mid-1950s. 107 ARE MODERN ENVIRONMENTS REALLY BAD FOR US? Based on observations such as in Figures 8–10, shifts in cause-of-death structure are described by a second descriptive model, the \epidemiologic transition" which accompanied the \demographic transition" (Omran, 1977). Omran (1977) called stage I of the demographic transition \the age of pestilence and famine." This period was characterized by high rates of infectious diseases occurring in periodic epidemics (exogenous or crisis mortality). This was followed by the \age of receding pandemics" (stages II–III of the demographic transition). During this period, infectious diseases changed from epidemic to endemic diseases (hence the reduction in exogenous or crisis mortality), and began to decline as causes of death, but remained endemic childhood diseases. The ﬁnal stage (IV) is called the \age of degenerative and manmade diseases." As infectious diseases declined in frequency, degenerative diseases, particularly cardiovascular diseases and neoplasms, emerged as the most common causes of death. This occurred in the 1920s in England and Wales (subject to the age structure used to standardize the data; Fig. 8). Omran (1977) was careful not to argue that the risk of degenerative diseases increased, but only that the proportion of deaths due to degenerative causes eventually exceeded infectious causes (despite his use of the description \age of degenerative and manmade diseases"). Since other and unknown diseases declined more than degenerative diseases increased, it is not reasonable to conclude that the risk of degenerative diseases must have increased. The emergence of degenerative diseases as the most common cause of death could have occurred if the risk of degenerative diseases increased, declined, or remained the same, as long as infectious causes declined faster then degenerative causes. Epidemiologic transition 1861–1964: controlling for misclassiﬁcation Three attempts have been made to statistically correct for misclassiﬁcation of cause of death, and to clarify the secular trends in degenerative causes of death during the later part of the 19th and early part of the 20th centuries. The ﬁrst was a statistical analysis conducted by Preston (1976), based on the international life-table database published in Preston et al. (1972), as described above (Table 1). Gage (1994) reexamined these same data using the Siler Model to further focus the analysis. Finally, Gage (1993) studied the trends in England and Wales using the Siler Model (1979) and additional decremented life tables, but based on the sample by Preston et al. (1972). All three studies are consistent in concluding that the risk of cardiovascular disease, and of certain degenerative diseases, declined, while neoplasms increased with modernization (1861–1964). Total degenerative disease (the sum of cardiovascular disease, neoplasms, and certain degenerative diseases) appears to have declined. Preston (1976) was the ﬁrst to examine changes in the cause-of-death structure across the demographic transition statistically. He regressed all cause-standardized death rates on each cause-speciﬁc standardized death rate. Thus a positive regression coefﬁcient (slope) indicates that the ith cause declines as all-cause mortality declines. A negative coefﬁcient indicates that the ith cause increases as mortality declines. The regressions were conducted so that the slopes of the regression represent the proportion of total decline due to that cause. The results are summarized in Table 2. In general, the TABLE 2. Regression of speciﬁc causes of death on overall death rate, all standardized for age structure1 Slope (correcting for misclassiﬁcation) Causes of death Males Females Respiratory tuberculosis Other infections and parasitic diseases Neoplasms Cardiovascular 0.1188 0.1458 0.1059 0.1398 0.0569 0.0316 (0.2390) 0.2831 0.0245 0.0179 (0.2456) 0.2434 0.1050 0.0206 0.0447 0.1041 0.0165 0.0422 0.0197 0.0041 0.3307 0.9998 Inﬂuenza, pneumonia, and bronchitis Diarrhea Certain degenerative diseases Certain diseases of infancy Maternal Violence Other and unknown Sum 0.0232 0.3475 1.0002 1 Slopes may be interpreted as % change due to cause. Positive regression coefﬁcient (slope) indicates that ith cause declines as all-cause mortality declines. Negative coefﬁcient indicates that ith cause increases as mortality declines. Preston (1976) only reported actual value of corrected slope for cardiovascular deaths. Adapted from Preston (1976). coefﬁcients are positive, with several exceptions. The exceptions are neoplasms in both sexes and cardiovascular disease in males, indicating that these categories of death increased as mortality declined. The category of certain degenerative diseases, on the other hand, declines in both sexes. Inﬂuenza, pneumonia, and bronchitis are the largest named contributors to the decline in mortality (about 25% of the total decline). However, the category of other and unknown causes of death accounts for an even larger proportion of the decline (about 34%). It is unlikely that poorly diagnosed causes of death declined faster than named and (hopefully) properly diagnosed causes of death. More likely, diagnosis improved with the epidemiologic transition, and much of the decline in other and unknown causes of death is due to improvements in classifying these causes of death, rather than the elimination of deaths whose causes were not well-recognized at the time. Thus the rates of decline in the named categories may be underestimated, while the rates of increase may be overestimated. Preston (1976) reexamined the data by regressing allcause standardized death rates, and other and unknown standardized death rates, on each of the named causespeciﬁc standardized death rates. This controls for misclassiﬁcation of cause of death, provided that the negative correlations between the standardized death rate for a named cause and the standardized death rate for other and unknown causes of death across countries in the international sample are a result of misclassiﬁcation of cause of death. The corrected coefﬁcients for neoplasms indicate that this cause increases as overall mortality falls in both sexes. On the other hand, the corrected coefﬁcients for cardiovascular disease are positive, indicating that cardiovascular mortality declines in both sexes as mortality declines. In fact, the slopes suggest that about 24% of the decline in all-cause mortality is due to declines in cardiovascular disease mortality. If this is correct, than the decline in cardiovascular mortality over the course of the demographic and epidemiologic transitions rivals the impact of inﬂuenza, pneumonia, and 108 T.B. GAGE TABLE 3. Sign and signiﬁcance of regression coefﬁcients for regressions of component-and-cause-speciﬁc deaths on total deaths, all standardized for age structure, based on the Siler Model1 Trend in cause of death as mortality declines2 Cause of death Males All causes combined Immature Residual Senescent Respiratory tuberculosis Residual Senescent Other infectious and parasitic diseases Immature Residual Senescent Certain diseases of infancy Immature Residual Inﬂuenza, pneumonia, and bronchitis Immature Residual Senescent Diarrhea Immature Residual Senescent Neoplasms Immature + Senescent + Cardiovascular disease Immature 0 Residual Senescent Certain degenerative diseases Immature Senescent Females Misclassiﬁcation of cause of death Period effect R2 + 0 94.5% 64.3% 73.7% + 73.2% 47.1% 0 0 0 + 89.2% 83.5% 36.6% 0 0 0 0 71.9% 38.4% 0 0 & + 0 90.5% 72.9% 74.3% 0 0 0 + 0 0 79.4% 53.0% 59.6% 0 & 0 39.8% 58.6% 0 0 & 0 0 29.8% 17.7% 39.9% 0 & + 0 31.9% 10.1% 0 0 0 0 0 + + 1 , Cause of death declines as mortality declines or period increases; +, cause of death increases as mortality declines or period increases; 0, coefﬁcient not signiﬁcant; &, cause of death with signiﬁcant numbers of misclassiﬁed deaths. Adapted from Gage (1994). 2 Corrected for misclassiﬁcation and period effects. bronchitis as the single largest named cause of general decline in international mortality (Preston, 1976). Gage (1994) reanalyzed the same data using the Siler Model (Fig. 5), which divides deaths into three categories: immature, residual, and senescent. The sample by Preston et al. (1972) includes total life tables and cause-eliminated life tables, i.e., a series of life tables, one for each cause of death, from which that particular cause of death is mathematically eliminated from the population. Fitting the Siler Model to each of the total and cause-eliminated life tables in the database of Preston et al. (1972) allows the proportion of deaths of each cause, including other and unknown causes, to be subdivided into the three components. Gage (1994) then conducted regression analyses similar to the analysis of Preston (1976) reported above, but separately on each component of the Siler Model. There are two advantages to this procedure. First, the analysis is based on the cumulative risk of death, i.e., on the hazard, as opposed to age-standardized death rates, which vary depending on the standard. A second advantage of the component-speciﬁc methodology is that \immature" other and unknown causes are regressed on the named \immature" causes of death, and \senescent" other and unknown causes are regressed on the named \senescent" causes of death, etc. This focuses the use of other and unknown causes as a correction for misclassiﬁcation of deaths. The assumption is that senescent other and unknown causes are likely to be misclassiﬁed senescent deaths, i.e., one of the named causes of senescent death, as opposed to one of the named causes of immature or residual death. Consequently, the statistical correction for misclassiﬁcation of cause of death is likely to be more accurate. Finally, Gage (1994) included a third independent variable, period (the year the life table refers to), to control for period effects. The results presented in Table 3 are similar to those of Preston (1976). Not surprisingly, all of the infectious diseases decline. Only inﬂuenza, pneumonia, and bronchitis appear to be signiﬁcantly misclassiﬁed among infectious causes of death. Infectious diseases also show the only large period effects. In general, infectious disease mortality tends to decline in the residual component and increase in the immature component. In the senescent component, period increases in respiratory tuberculosis and other infectious and parasitic diseases are balanced by period declines in inﬂuenza, pneumonia, and bronchitis. These secular changes in component-speciﬁc infectious disease mortality support the hypothesis by Fenner (1970) that increased urbanization and population growth result in a shift from epidemic (exogenous, i.e., residual component causes) to endemic diseases affecting the very young and elderly. The analysis of degenerative diseases also supports the results of Preston (1976). All three degenerative disease categories are signiﬁcantly misclassiﬁed. Degenerative dis- ARE MODERN ENVIRONMENTS REALLY BAD FOR US? 109 Fig. 11. Decline in cause-speciﬁc total degenerative mortality as mortality declines in sample of Preston et al. (1972), while statistically controlling for misclassiﬁcation of causes of death and period effects. Decline in senescent degenerative mortality is very similar. Adapted from Gage (1994). Total degenerative mortality, represents the sum of cardiovascular, neoplasms, and certain degenerative deaths. eases occur in all three components of the Siler Model, but the vast majority of degenerative deaths occur in the senescent component, i.e., deaths in other components of the Siler Model are generally several orders of magnitude smaller (Gage, 1991, 1994). Consequently, the period effects on immature degenerative mortality are very small. Further, the period effects of the three degenerative cause-of-death categories tend to cancel each other out. Corrected for misclassiﬁcation, senescent and total neoplasms increase in both males and females as mortality declines. On the other hand, corrected senescent cardiovascular and certain degenerative disease mortality all decline (Fig. 11). Interestingly, males beneﬁted more from the decline in infectious disease mortality and less from degenerative diseases than females, although female mortality fell more, overall (Gage, 1994). In general, these results support those of Preston (1976). Corrected for misclassiﬁcation of cause of death, neoplasms increased, while cardiovascular and certain degenerative deaths declined. The declines in cardiovascular and certain degenerative diseases outweigh the increase in neoplasms, so total degenerative deaths decline as overall mortality declines. Gage (1993) also examined the decline in cause-speciﬁc mortality in England and Wales from 1861–1964, using the Siler Model. Here, statistical corrections for misclassiﬁcation of cause of death are not possible, because there is only a single longitudinal series (statistical correction for misclassiﬁcation of cause of death relies on comparisons across countries with different classiﬁcation efﬁciencies). The trends for total component-speciﬁc mortality were presented earlier (Fig. 7). Overall, senescent mortality declined between 1861–1964. However, some senescent deaths are likely to be due to the infectious diseases that occur in the senescent component (Gage, 1991), i.e., inﬂuenza, pneumonia, bronchitis, and respiratory tuberculosis. Figure 12 shows the trends in total senescent mortality and senescent mortality with inﬂuenza, pneumonia, bronchitis, and respiratory tuberculosis eliminated. In this latter case, the peaks in senescent mortality are notably reduced. Nevertheless, the remaining degenerative diseases increase until the 1890s and then decline. The increase in degenerative diseases until 1890 is at least partly due to the excess mortality resulting from the inﬂuenza epidemic of the 1890s (Azambuja and Duncan, 2002; Davenport, 1976; Lancaster, 1990). On the other hand, the 1921 low in degenerative disease mortality is probably due to excess degenerative diseases deaths associated with the 1918–1919 inﬂuenza pandemic, which occurred 2 years or so before 1921, as discussed above. If a line is ﬁtted to senescent degenerative deaths, mortality due to this cause declines. The total risk of senescent degenerative mortality (age-speciﬁc senescent risk summed or integrated across 0–80 years of age) in 1861 was 4.07, which declined to 3.79 in 1964. Assuming that all misclassiﬁed deaths in the senescent component are degenerative deaths (as 110 T.B. GAGE Fig. 12. Decline in senescent mortality in England and Wales 1861–1964, with and without infectious senescent causes of death decremented. Straight line is regression ﬁtted to senescent degenerative disease mortality (i.e., senescent mortality with senescent infectious diseases decremented). From Gage (1993). opposed to infectious causes of death) and that no infectious causes remain after decrementing inﬂuenza, pneumonia, bronchitis, and respiratory tuberculosis, the result suggests that degenerative diseases as a group have declined since the 1860s, and dramatically since the 1890s. Epidemiologic transition: last half of the 20th century While errors in classiﬁcation of cause of death are a serious problem for earlier periods, they are less likely to bias results during the latter half of the 20th century. Thus the observed trends are more likely to reﬂect real trends in mortality. In October 1978, the National Heart, Lung and Blood Institute of the National Institutes of Health sponsored a conference to consider the surprising ﬁnding that the risk of coronary heart disease mortality was declining (Havlik and Feinleib, 1979; National Center for Health Statistics (NCHS), 1978). The attendees were charged with the following question: \What preventative, curative, environmental, or other factors contributed to this recent decline, especially to the dramatic turnaround for ischemic heart disease?" Clearly the observation that the most common cause of death in the United States was declining caught the US public health services off guard. The observed trends of a number of degenerative diseases for the US during the 20th century are presented in Figure 13. These data clearly show the \rise and fall of ischemic heart disease" (Stallones, 1980), and of cardiovascular mortality more generally, which peaked in the US in the late 1940s or early 1950s. This phenomenon is now internationally established, e.g., in Poland (Zatonski et al., 1998), the Nordic countries (Martelin, 1987), most of Central Europe (Mesle, 2004), various other European countries (Vallin and Mesle, 2004), and Japan (Yanagishita and Guralnick, 1988). In fact, the trend is not limited to cardiovascular mortality. Barker (1989) discussed a number of diseases that rose and then fell in the 20th century. The etiology of some of these, like polio, is well-established (Fenner, 1970); others, like heart disease, are not well-established. In the late 1990s, even neoplasms began to decline in the US (Fig. 13). If the recent rates of decline in neoplasms and cardiovascular deaths continue at present rates, neoplasms will soon outrank cardiovascular deaths as the most common cause of death in the US. While Figure 13 only considers the most common degenerative diseases, it is clear, given the dramatic declines in heart disease, that overall degenerative disease mortality has declined since the 1940s. The decline in cancer began in Japan in the 1960s, much earlier then the decline in other regions of the world (Gersten and Wimoth, 2002). This is in part because of the unusual cause structure of cancers in Japan. Much of the decline in Japan was due to stomach and cervical cancers, which are largely attributed to infectious agents (Doll and Peto, 2001). Liver and lung cancers have increased and continue to increase, probably due to shifts in alcohol consumption and smoking. All of these trends appear to be heavily inﬂuenced by the involvement in World War II and subsequent economic development. However, the trends reﬂect a transition from cancers with a large infectious component of risk to cancers with a large behavioral component of risk (Gersten and Wimoth, 2002). Similar trends were observed in other industrialized countries (Becker, 1998; National Cancer Institute, 1999), although in general, the decline in infectious cancers must have been smaller than, and outweighed by, increases in those cancers associated with behavioral risk factors, since overall, cancers appear to have increased as a group at least until recently, e.g., the US (the trends in neoplasms, Fig. 13). How large an effect has the decline in cardiovascular disease had on the mortality of industrialized nations? White (1999) examined the two cause-of-death categories that changed the most during the 20th century in the US: respiratory tuberculosis and cardiovascular mortality. Respiratory tuberculosis was considered throughout the ARE MODERN ENVIRONMENTS REALLY BAD FOR US? 111 Fig. 13. \Rise" and fall of some degenerative diseases and inﬂuenza (cause responsible for largest decline in named infectious disease) in US male population from 1900–1998 (registration states, 1900–1932; US, 1933–1998). Age standardized to US 2000 age distribution. Data source: CDC/NCHS, National Vital Statistics System, Mortality, unpublished table HIST293. 20th century, while cardiovascular disease was examined from 1940 rather than 1900 due to the likelihood of errors in classiﬁcation in prior periods. In addition, White (1999) included cardiovascular-renal diseases as general cardiovascular deaths up until the Eighth Revision of the International Classiﬁcation of Diseases, which formalized this relationship in 1969. Measured as expectation of life, the decline in respiratory tuberculosis since 1900 has contributed an additional 5.3 years to expectation of life. On the other hand, since 1940, the decline in cardiovascular mortality has increased expectation of life by 6.4 years. This is also reﬂected in the number of deaths averted by the declines in these same diseases, estimated to be 26,003 for respiratory tuberculosis during 1900–2000, and 26,333 for cardiovascular disease during 1940–2000 (White, 1999). A similar comparison between the decline in inﬂuenza mortality (the cause that declines the most in the sample of Preston et al., 1972) and degenerative diseases for the US is presented in Figure 13. If cardiovascular mortality was declining before 1940, as suggested by the analyses of Preston (1976) and Gage (1993, 1994), then the decline in cardiovascular mortality of the last 100–150 years exceeds the decline in any other named cause of death, as deﬁned in Table 1. In any event, even if the declines in cardiovascular disease mortality before 1940 are ignored, cardiovascular mortality is the largest single cause of the historical decline in mortality in the US during the 20th century. Epidemiologic transition, age-speciﬁc: last half of the 20th century Salomon and Murray (2002) conducted an analysis of relative trends in the cause structure of mortality since 1950, using the Global Burden of Disease 1990 study sample. This consists of 58 countries, mostly European, and mostly developed for the period 1950–1990 (Murray and Lopez, 1996). However, a number of less developed countries are also included. For analysis, causes of death were divided into three groups: group 1, infectious diseases; group 2, noncommunicable diseases (including neoplasms, cardiovascular disease, and diabetes mellitus, i.e., degenerative diseases); and group 3, accidents and violence. Analysis differs from previous studies in that the transition is examined by age and sex and conﬁned to composition of cause of death (as opposed to absolute changes in level of mortality), as Omran (1977) originally intended. Unfortunately, the results are voluminous and cannot be presented here (Salomon and Murray, 2002). However, their conclusions are as follows. First, their analysis conﬁrms the general principles of the epidemiologic transition, a shift from predominantly infectious disease mortality to predominantly noninfectious and accidental deaths. Second, as infant mortality declines, infant mortality shifts from infectious to noninfectious causes of death, with few or no accidental deaths. Third, children over age 1 year experience shifts from predominantly infectious deaths to an equal mix of noninfectious and accidental deaths. Accidental deaths are more common in males than in females. Fourth, young adults (15–44 years) differ by sex. In males, declining mortality is associated with a shift from accidents to noncommunicable diseases, although this trend is ameliorated by an increased standard of living. As mortality declines among females, there is at ﬁrst a trend toward noncommunicable diseases, followed by a trend toward accidental deaths. Higher standards of living are again associated with higher accidental deaths. Finally, males and females over age 50 show almost no change in composition in cause of death as mortality declines during the latter half of the 20th century. Thus, since the 1950s, all causes have declined approximately equally in the population over 50 years of age. 112 T.B. GAGE Summary: epidemiologic transition The history of secular trends in cause of death depends on whether it is expressed as the proportional structure of cause of death (Omran, 1977) or as trends in the risk of cause of death. To answer the question of whether modernization is bad for health, it is the latter deﬁnition that must be employed. The studies presented above suggest that during the historical decline in mortality in England and Wales, the US, and other modernized countries, the risk of both infectious and degenerative causes of death declined. Clearly, infectious diseases began to decline in the 1850s. Corrected for misclassiﬁcation of cause of death, cardiovascular death rates and certain degenerative disease death rates may have begun to decline at approximately the same time as infectious diseases, but clearly declined after 1900. This decline is visible without correction for misclassiﬁcation of cause of death after the 1940s. Cardiovascular deaths probably account for more of the general decline in mortality than any other single cause of death. Neoplasms, corrected for misclassiﬁcation of cause of death, increased until late in the 20th century, and only recently began to decline. With the exception of neoplasms, these results suggest that we are in the second stage of a simple two-stage epidemiologic transition: 1) the age of pestilence and famine, and 2) the age of declining infectious and degenerative diseases. A future third stage might consist largely of accidental deaths (Rogers and Hackenberg, 1987), or possibly a return of infectious disease mortality (Barrett et al., 1998). Most other amendments to the epidemiologic transition involved adding additional stages to the model (Olshansky and Ault, 1986; Rogers and Hackenberg, 1987; Vallin and Mesle, 2004; WolleswinkelVan-Den-Bosch et al., 1997). These additions are due to the view that degenerative diseases increased prior to declining, as shown in Figure 13. However, as presented above, these increases appear to be largely due to misclassiﬁcation of cause of death. It is likely that the decline in degenerative diseases began prior to 1940. DEGENERATIVE DISEASE MORBIDITY If mortality, including overall degenerative disease mortality, has declined, why has morbidity increased? Or has morbidity increased? Mortality is only one way to deﬁne health. Health may also be deﬁned as the incidence or prevalence of morbidity. While incidence, prevalence, and mortality are all related with \health" and with each other, they need not be consistent. A relatively simple model of the relationship of morbidity and mortality due to degenerative diseases in the presence of competing causes of death, where morbidity is a clearly recognizable dichotomous trait, is presented in Figure 14. The model proposes that there is a pool of healthy individuals who are depleted in two ways. Individuals from the healthy pool become morbid due to degenerative cause A at a speciﬁc rate, which may or may not be dependent on age and or period effects. For example, blood pressure is thought to increase with age, so high blood pressure morbidity increases with age. The rate of transition from healthy to morbid state A is commonly called the incidence of the morbid state (number of new cases). There is then a rate at which morbid individuals die of cause A, e.g., the rate at which individuals with high blood pressure have cardiovascular accidents resulting in death. Of course, healthy individuals may also die Fig. 14. Graphical depiction of relationship of cause-speciﬁc morbidity and mortality. Prevalence of morbid condition A is function of rate at which healthy individuals become morbid (incidence), and rapidity at which individuals morbid with cause A die (of cause A or any other competing cause). of cause A without entering the morbid state, or enter the morbid state so brieﬂy that it is not clinically recognized. The prevalence of morbid state A (number of contemporary cases) is then a function of the incidence of morbidity and the mortality rate of morbid individuals due to cause A, and the mortality of morbid individuals due to other competing causes. Clearly, the higher the incidence and the longer the survival of morbid individuals (whatever the cause of death), the higher the prevalence of morbidity. A low prevalence of morbidity due to cause A does not necessarily mean that there is a low incidence or low mortality rate due to cause A. In fact, all other things being equal, the higher the mortality of morbid individuals, the lower the prevalence of morbidity! A second problem with respect to documenting incidence and prevalence concerns secular changes in the recognition of morbid conditions. Just as the diagnosis of cause of death has improved over time, the diagnosis of morbid conditions has also improved. With respect to morbid conditions, the issues are even more complex than when considering cause of death. Unlike death, morbidity may be more or less severe. Severity may inﬂuence diagnoses, and/or the deﬁnition of disease threshold may change, both of which could artiﬁcially inﬂuence secular trends in morbidity. It is also possible for a population to experience simultaneously a real increase in prevalence and a decrease in severity of morbidity, or vice versa. Below, I brieﬂy examine the literature concerning trends in morbidity, with particular reference to degenerative diseases. Morbidity during the last half of the 20th century Studies of the change in prevalence of various degenerative diseases over the last 20 or 30 years generally conclude that the prevalence of most degenerative conditions is increasing (Crimmins, 2004; Crimmins and Saito, 2000; Cutler and Richardson, 1997; Manton et al., 1995). These include the prevalence of neoplasms, cardiovascular diseases, diabetes, and arthritis. The increases in prevalence are due to the faster decline in degenerative ARE MODERN ENVIRONMENTS REALLY BAD FOR US? disease mortality than declines in incidence, at least in community-based studies where incidence can be estimated (Burke et al., 1989; Demirovic et al., 1993; Hunimk et al., 1997; McGovern et al., 1992, 1996). Stroke follows a similar pattern, but here incidence may have increased (Crimmins and Saito, 2000). The average number of morbid diseases reported by individuals also increased, because people survive diseases that once would have been fatal. Thus older people display more morbid diseases, but less disability than in the past. Severity of morbidity has declined (Crimmins, 2004). Morbidity during the ﬁrst half of the 20th century Information on degenerative morbidity prior to 1960 is not common, even in modern countries. The results presented below are largely taken from the Early Indicators Project, under the general direction of Robert Fogel (Costa, 2000; Fogel, 2005). Estimates of prevalence of degenerative morbidity are available from Costa (2000), who compared morbidity rates among Civil War veterans examined in 1900–1910 and men matched for age from National Health and Nutrition Examination Survey (NHANES) from 1971–1975. She reported declines in prevalence of a number of degenerative conditions including valvular heart disease, which declined from 19.2% to 1.7%, and arteriosclerosis, which declined from 1.9% to 0.9% among 50–64-year-olds. In older cohorts (60–74-year-olds), valvular heart disease declined from 26.9% to 3.6%, and arteriosclerosis from 8.2% to 2.3%. Costa (2000) also reported declines in a number of nonfatal conditions such as joint and back problems, irregular pulse, or heart murmur and trills. In general, the prevalence of morbid degenerative conditions may have declined from the turn of the 20th century to the early 1970s. Additionally, Costa (2000) estimated the survival (person-years surviving) of individuals with speciﬁc morbid conditions for the Civil War and NHANES (1971–1975) cohorts. The only signiﬁcantly shorter life spans for the Civil War vs. NHANES cohorts were reported for the total samples of both age groups and 60–74-year-olds with joint problems. Among the remaining morbid conditions examined, life spans tended to be shorter in the Civil War cohort, but not signiﬁcantly so. Thus survival times with a morbid condition may have increased slightly between 1900–1971, but have not increased signiﬁcantly. This would suggest (not surprisingly) that the declines in prevalence are not due to shorter survival of morbid individuals in the later cohorts. Finally, Fogel (2005) reported age at onset of several degenerative conditions for men born between 1830–1845 and 1918–1927. The age of onset of all conditions examined increased in the younger cohort. For example, average age at onset (diagnosis) for heart disease was 55.9 years for the 1830–1845 cohort, and 65.7 years for the 1918–1927 cohort. Similar results were reported for neoplasma, with age at onset of 59.0 years for the early cohort, and 66.6 years for the recent cohort. This suggests that the incidence of these causes of morbidity declined from the beginning to the end of the 20th century. Given that diagnoses of many of these morbid conditions were likely to be underestimated, particularly in the earlier cohort, these declines in incidence are likely to be underestimates. Morbidity during the prehistoric period It is commonly reported that degenerative mortality prevalence is low in traditional anthropological popula- 113 tions. The empirical basis for this conclusion is that degenerative disease risk factors (e.g., high blood pressure, high cholesterol, adverse lipoprotein proﬁles, and obesity) are absent from these populations (Eaton et al., 1988). As a result, it is often assumed that degenerative causes of death are low in these populations. However, as noted above, the prevalence of morbidity and mortality are, all other things being equal, negatively correlated. Cause of death is almost never available for anthropological populations, since the number of deaths observed by anthropologists in the ﬁeld is small, because anthropologists are not necessarily well-qualiﬁed to pronounce cause of death, and because the mere presence of an anthropologist suggests that the population’s mortality structure may be inﬂuenced by contemporary national populations. Consequently, the assumption that degenerative causes of death are absent in these populations is premature. The evidence that the prevalence of degenerative disease morbidity is low in traditional populations is better substantiated (Eaton et al., 1988). If it is assumed that these traditional populations have a low prevalence of degenerative disease, and that historical industrializing populations (ca. late 1800s) had a high prevalence, then either the incidence of degenerative morbidity increased at some time prior to the 1850s, the survival of individuals with degenerative disease morbidity improved, and/or general mortality declined and longevity improved. The only evidence currently available is that general mortality appears to have declined (Fig. 1). A better understanding of these issues requires additional information. Summary: morbidity A simple parsimonious explanation for trends in mortality, incidence, prevalence, and survival over the 20th century is that the incidence of degenerative morbid conditions declined throughout the ﬁrst half of the 20th century, and that survival with morbid conditions remained approximately constant until perhaps midcentury, and then increased in the latter part of the 20th century. As a result of these trends, prevalence declined until the mid20th century and then began to increase, as survival with morbidity improved. WHY DID INFECTIOUS DISEASE MORTALITY DECLINE? A large body of research has attempted to identify exactly why infectious diseases declined with modernization. Despite (or because of) this intensive research, the cause of the historical decline in infectious disease mortality remains controversial and largely unexplained (Hinde, 2003; Woods, 2000). In his classic exposé on the decline in infectious disease mortality, McKeown (1976) proposed four potential causes of the decline: 1) evolution of hostparasite interactions, 2) improvements in sanitation, 3) improvements in modern medicine (deﬁned as effective chemotherapy or vaccination), and 4) improvements in nutrition and standard of living. McKeown (1976) concluded, after eliminating \all other possible causes" (a–c), that the cause of the decline must be improved nutrition, despite any hard evidence for or against this conclusion. This analysis suffers from several problems (Hinde, 2003; Szreter, 1989; Woods, 2000). First, his arguments against host-parasite interactions were based on the notion that host and parasite evolve to accommodate one another, and that evolution is slow. Recent studies in the ecological literature argue that the evolution of host-parasite interac- 114 T.B. GAGE tions is considerably more complex, and that parasite evolution can occur more rapidly (Bull, 1994; Ebert, 1999) than McKeown (1976) assumed. Second, it was argued that sanitation may have been considerably more effective than McKeown (1976) supposed (Cutler and Miller, 2005; Szreter, 1989; Woods, 1991, 2000), although McKeown (1976) did acknowledge the contribution of sanitation. Third, the view of McKeown (1976) that improved nutrition (standard of living) was responsible for the decline is not supported by current data (Floud et al., 1990; Fogel et al., 1982; Livi-Bacci, 2000; Woods, 2000; Wrigley and Schoﬁeld, 1981). Only the argument by McKeown (1976) that effective chemotherapy was not responsible for the decline in infectious disease mortality remains largely uncontested. Finally, the argument by McKeown (1976) that these four causes represent all possible causes of the decline in infectious disease mortality is not convincing. Clearly, germ theory (medicine deﬁned more broadly), general education, and the implications of behavioral intervention (e.g., washing one’s hands) might have some effect on infectious disease mortality (Cutler and Miller, 2005; Ewald, 2000; Preston and Haines, 1991; Schoﬁeld and Reher, 1991; Woods, 2000). The problem with the current state of the critique of the hypothesis of McKeown (1976) is that if his standard-of-living argument is incorrect, then what explains the decline in airborne diseases (e.g., tuberculosis, inﬂuenza, or pneumonia), which were among the largest infectious disease categories contributing to the historical decline in mortality (Hinde, 2003)? My point here is not to argue for or against any of the various hypotheses concerning the decline in infectious disease mortality with modernization, but simply to point out that a convincing and comprehensive explanation of the decline in infectious disease mortality is still not available. WHY DID DEGENERATIVE DISEASE MORTALITY DECLINE? The fact that many of the degenerative diseases declined historically does not necessarily mean that the risk factors listed by Eaton and Eaton (1999), Eaton et al. (1988), and others do not contribute to the etiology of degenerative mortality. It is likely that cigarette smoking, which has increased with modernization, contributed to the rise of neoplasms (Crimmins, 2004; Gersten and Wimoth, 2002). Further, cigarette smoking, poor nutrition, lack of exercise, and obesity also appear to be risk factors for cardiovascular disease. However, if the historical decline in degenerative disease is correct, then what is it about modern environments and lifestyles that overcame these wellknown adverse risk factors of modern life and resulted in an overall decline in degenerative disease mortality? Listed below are some possible causes of the decline in degenerative mortality, along with some brief comments on each potential cause: 1. 2. 3. 4. Changes in lifestyles; Improvements in modern medicine; Direct interactions with infectious diseases; Indirect interactions with infectious disease mortality; and 5. Degenerative diseases are infectious diseases. The 1978 National Institutes of Health symposium concerning the unexpected declines in ischemic heart disease, observed post-1940, concluded that improvements in life- style (e.g., nutrition, exercise) and improvements in modern medicine reduced the risk or delayed the onset of cardiovascular disease (Havlik and Feinleib, 1979; National Center for Health Statistics (NCHS), 1978; Olshansky and Ault, 1986). However, it seems unlikely that the changes in lifestyles in question, which began to be recommended in the 1950s and 1960s, and/or modern medicine (e.g., blood pressure medicines, introduced even later) could explain the long-term decline in cardiovascular and certain degenerative disease mortality observed during the ﬁrst half of the 20th century. They might help explain the continuing declines in degenerative disease mortality in the last half of the 20th century. The infectious diseases pandemics, particularly respiratory diseases pandemics, are traditionally associated with excess degenerative disease mortality (Azambuja and Duncan, 2002; Davenport, 1976; Lancaster, 1990). Perhaps the additional stress of these infectious diseases contributed to degenerative mortality, which declined as infectious causes of death declined. This is possible, but the precise mechanisms remain to be determined. If respiratory diseases simply reduced the survival of individuals with degenerative disease morbidity (consistent with the data presented above), then the decline in respiratory diseases could potentially cause the prevalence of degenerative disease morbidity to increase. However, the prevalence of degenerative disease morbidity does not appear to have increased until late in the 20th century, well after the respiratory diseases declined as causes of death. Indirect interactions with infectious disease mortality could take several forms, such as early life effects (Fogel, 2005) or the fetal origins hypothesis (Barker, 1999; Godfrey and Barker, 2000). Costa (2000) and Fogel (2005) argued that exposure to infectious disease and/or the impact of infectious disease on nutritional status early in life could be a risk factor for developing degenerative disease later in life. Alternatively, infectious disease and/ or nutritional status of the mother could inﬂuence fetal development and predispose an infant to degenerative disease later in life (Barker, 1999; Godfrey and Barker, 2000). In either case, infectious disease has an indirect interaction with subsequent degenerative disease, i.e., the incidence of degenerative disease. Finally, there is the possibility that some degenerative diseases are in fact due to infectious agents (Ewald, 2000). There is currently clear evidence that some cancers are infectious (Doll and Peto, 2001), and both Chlamydia pneumoniae and Porphyromonas gingivalis are potentially implicated in the etiology of cardiovascular disease (Grayston, 2000; Patel et al., 1995). Like the historical decline in infectious disease mortality, the mechanisms and causes of the historical decline in degenerative causes of death remain to be determined. WHY HAS HUMAN MORTALITY DECLINED IN THE FACE OF NOVEL ENVIRONMENTS? Evolutionary theorists typically postulate that novel changes in the environment will be detrimental to organisms in these environments. Within biological anthropology, for example, it is argued that humans are \adapted" to the hunter-gatherer environment and not to the agricultural and \built" industrial environments which are \novel" (Eaton and Eaton, 1999; Eaton et al., 1988; Nesse and Williams, 1999). Clearly, however, humans have beneﬁted from the general changes in modern human environments by a general reduction in mortality. Further, it is ARE MODERN ENVIRONMENTS REALLY BAD FOR US? not even clear that mortality increased with the adoption of agriculture. Of course, some factors, such as smoking, probably have not contributed to this improvement in health, and there may be other negative factors as well (Eaton and Eaton, 1999; Eaton et al., 1988). On the whole, however, during the last several centuries mortality has declined, although, as discussed above, it is not exactly clear why. How can these improvements in health in the face of novel environments be reconciled with evolutionary theory? One solution is to consider the improved health an example of \genotype by environmental correlation." Genotype by environmental correlation occurs when and if organisms with speciﬁc genotypes seek out genotypically compatible environments (Stearns, 1992). Similarly, why cannot humans \build" environments that are on the whole compatible with human biology? After all, modernization includes the development of sanitation systems, public health systems, effective chemotherepy, etc. Of course, there are likely to be constraints on what can be \built," and consequently some aspects of the \built" environment may be detrimental, but the overall affect need not be negative. CONCLUSIONS The demographic history of the human population has proven difﬁcult to establish, because the available empirical evidence is incomplete. Since the evidence tends to improve with time, secular improvements in data collection have been interpreted as secular trends in mortality or health. It is the thesis of this paper that degenerative and infectious diseases have declined over the course of modernization. The widely held view that modern lifestyles are bad for human degenerative health is largely due to secular improvements in data collection rather then a true deterioration in health. 1. Barring occasional pandemics, total mortality has declined since the 1850s in the developed regions of the world, and since the 1920s in many of the less developed regions of the world. There is no indication that mortality prior to the beginning of the industrial revolution was signiﬁcantly lower than the mortality observed from the 1750s to 1850s in Sweden. Clearly, overall modernization/industrialization has been associated with improvements in human health. 2. There is no doubt that infectious diseases declined with modernization, despite increases in population size and urbanization, which exacerbate infectious causes of death (Fenner, 1970). Exactly what aspects of modern environments and modern lifestyles are responsible for the decline in infectious disease mortality remain controversial. 3. Degenerative diseases as a whole have also declined with modernization and industrialization. The commonly held concept that degenerative diseases increased until the mid-20th century is based on trends in cause of death that are uncorrected for misclassiﬁcation of cause of death. Whenever methods of correcting for misclassiﬁcation of cause of death are employed, degenerative diseases as a group appear to decline. They begin to decline in the uncorrected data after the 1940s or 1950s. It is likely that cardiovascular mortality has declined more than any of the other major infectious disease categories, such as respiratory tuberculosis (responsible for the largest proportion of the infec- 115 tious disease decline in England and Wales and the US) or inﬂuenza, pneumonia, and bronchitis (responsible for the largest proportion of the decline in infectious disease mortality internationally). The exception here is neoplasms, which appear to increase as a cause of death until recently in most modernized/industrialized countries. This could be due in part to cigarette smoking, a behavior correlated with modernization. Exactly what aspects of modern environments and modern lifestyles are responsible for the decline in degenerative disease mortality prior to the 1950s are largely unstudied. 4. Mortality is not the only possible deﬁnition of health; morbidity can also be considered. However, due to the relationship of morbidity and mortality, morbidity may increase while mortality declines, and vice versa. Historical information on morbidity is even more incomplete than on mortality. However, it appears that the prevalence of morbidity declined during the early part of the 20th century but began to increase in the latter part of the 20th century. This increase could be due to better survivorship of morbid individuals due to recent improvements in medical treatment. 5. The ﬁeld of evolutionary medicine, among others, has argued that the \novel" environmental conditions experienced by modern/industrialized humans should be detrimental to health, since we are \adapted" to a hunter-gatherer lifestyle. This is not an evolutionarily necessary conclusion. Given that modern environments are at least partly \built" by humans, the effects might be considered a \genotype by environment correlation," which are often positive. ACKNOWLEDGMENTS I thank Sara Stinson, Paul Leslie, Tom Brutsaert, Janet Padiak, and two anonymous reviewers for comments on the manuscript, and Lynne Shultis for help in preparing the ﬁgures. LITERATURE CITED Acsadi G, Nemeskeri J. 1970. History of human life span and mortality. Budapest: Akademiai Kiado. Antia R, Regoes RR, Koella JC, Bergstrom CT. 2003. The role of evolution in the emergence of infectious diseases. Nature 426: 658–661. Azambuja MIR, Duncan BB. 2002. Similarities in mortality patterns from inﬂuenza in the ﬁrst half of the 20th century and the rise and fall of ischemic heart disease in the United States: a new hypothesis concerning the coronary heart disease epidemic. Cad Saude Publica 18:557–577. Barker DJP. 1989. Rise and fall of Western diseases. Nature 338:371–372. Barker DJP. 1999. The fetal origins of coronary heart disease and stroke: evolutionary implications. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 246–250. Barrett R, Kuzawa CW, McDade T, Armelagos GJ. 1998. Emerging and re-emerging infectious diseases: the third epidemiological transition. Annu Rev Anthropol 27:247–271. Becker N. 1998. Cancer mortality and prevention in the European Union. Eur J Surg Oncol 24:370–374. Bull JJ. 1994. Perspective: virulence. Evolution 48:1423–1437. Burke G, Sprafka J, Folsom A, Luepker R, Norsted S, Blackburn H. 1989. Trends in CHD mortality, morbidity, and risk factor levels from 1960 to 1986: the Minnesota Heart Survey. Int J Epidemiol [Suppl] 18:73–81. 116 T.B. GAGE Coale AJ, Demeny P, Vaughan B. 1983. Regional model life tables and stable populations. New York: Academic Press. Cohen MN. 1989. Health and the rise of civilzation. New Haven: Yale University Press. Cohen MN, Armelagos GJ, editors. 1984. Paleo-pathology at the origins of agriculture. Orlando: Academic Press. Costa DL. 2000. Understanding the 20th century decline in chronic conditions among older men. Demography 37:53–72. Crimmins EM. 2004. Trends in the health of the elderly. Annu Rev Public Health 25:79–98. Crimmins EM, Saito Y. 2000. Changes in the prevalence of diseases among older Americans: 1984–1994. Demogr Res Vol. 3– 9. http:// www.demographic-research.org. Cutler DM, Miller G. 2005. The role of public health improvements in health advances. Demography 42:1–22. Cutler DM, Richardson E. 1997. Measuring the health of the U.S. population. In: Baily M, Reiss P, Winston C, editors. Brookings papers on economic activity: microecomonics. Washington, DC: Brookings Institute. p. 217–271. Davenport F. 1976. Inﬂuenza viruses. In: Evans AS, editor. Viral infections of humans: epidemiology and control. New York: Plenum Press. p 373–396. Demirovic J, Blackburn H, McGovern P, Sprafka J, Doliszny K. 1993. Sex differences in coronary heart disease mortality trends: the Minnesota Heart Survey, 1970–1988. Epidemiology 4:79–82. Dobson M. 1997. Contours of death and disease in early modern England. Cambridge: Cambridge University Press. Doll R, Peto R. 2001. The causes of cancer. Rev Epidemiol Sante Publique 49:193–200. Eaton SB, Eaton SB III. 1999. The evolutionary context of chronic degenerative diseases. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 251–259. Eaton SB, Konner M, Shostak M. 1988. Stone Agers in the fast lane: chronic degenerative diseases in evolutionary perspective. Q Rev Biol 69:353–367. Ebert D. 1999. The evolution and expression of parasite virulence. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 161–171. Ewald PW. 2000. Plague time. New York: Anchor Books. Fenner F. 1970. The effects of changing social organization on the infectious diseases of man. In: Boyden SV, editor The impact of civilization on the biology of man. Canberra: Australian National University Press. p 48–76. Floud R, Wachter K, Gregory A. 1990. Height, health and history. Cambridge: Cambridge University Press. Fogel RW. 2005. Changes in the disparities in chronic diseases during the course of the 20th century. Perspect Biol Med 48: 150–165. Fogel RW, Engerman SL, Trussell J. 1982. Exploring the uses of data on height: The analysis of long-term trends in nutrition, labor welfare and labor productivity. Soc Sci Hist 6: 401. Gage TB. 1988. Mathematical hazard models of mortality: an alternative to model life tables. Am J Phys Antrhopol 76:429–441. Gage TB. 1991. Causes of death and the components of mortality: testing the biological interpretations of a competing hazards model. Am J Hum Biol 3:289–300. Gage TB. 1993. The decline of mortality in England and Wales 1861 to 1964: an analysis using competing hazards models. Popul Stud 47:47–66. Gage TB. 1994. Population variation in cause of death: level, gender and period effects. Demography 31:271–296. Gage TB. 2000. Demography. In: Stinson S, Bogin B, Huss-Ashmore R, O’Rourke D, editors. Human Biology: an evolutionary and biocultural approach. New York: John Wiley and Sons. p 507–552. Gage TB, McCullough JM, Weitz CA, Dutt JS, Abelson A. 1989. Demographic studies in human population biology. In: Little M, Haas J, editors. Human population biology: a transdisciplinary science. Oxford: Oxford University Press. p 45–68. Gersten O, Wimoth JR. 2002. The cancer transition in Japan since 1951. Demogr Res 7:272–306. Godfrey KM, Barker DJP. 2000. Fetal nutrition and adult disease. Am J Clin Nutr 71:1344–1352. Grayston JT. 2000. Background and current knowledge of Chlamydia pneumoniae and atherosclerosis. J Infect Dis 181:402– 410. Havlik RJ, Feinleib M, editors. 1979. Proceedings of the conference on the decline in coronary heart disease mortality. NIH publication no. 79-1610. Washington, DC: US Department of Health, Education and Welfare. Hinde A. 2003. England’s population: a history since the Domesday Survey. New York: Oxford University Press. Holmes EC. 1999. Molecular phylogenies and the genetic structure of viral populations. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 173–182. Holmes EC. 2004. The phylogeography of human viruses. Mol Evol 13:745–756. Hoppa RD, Vaupel JW, editors. 2002. Paleodemograhy. Cambridge: Cambridge University Press. Howell N. 1982. Village composition implied by a paleodemographic lifetable: the Libben Site. Am J Phys Anthropol 59: 263–269. Human Mortality Database. 2004. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany). Available at www.mortality.org or www. humanmortality.de. Hunimk MG, Goldman L, Tosteson AN, Mittleman MA, Goldman PA. 1997. The recent decline in mortality from coronary heart disease, 1980–1990: the effect of secular trends in risk factors and assessment. JAMA 277:534–542. Lancaster HO. 1990. Expectation of life. New York: SpringerVerlag. Lee RD, Carter LR. 1992. Modeling the forecasting of U.S. mortality. J Am Stat Assoc 87:659–675. Livi-Bacci. 2000. The population of Europe. Oxford: Blackwell. Lovejoy CO, Meindl RS, Pryzbeck TR, Barton TS, Heiple KG, Knotting D. 1977. Paleodemography of the Libben Site, Ottawa County, Ohio. Science 198:329–352. Manton KG, Stallard E. 1984. Recent trends in mortality analysis. London: Academic Press. Manton KG, Stallard E, Corder L. 1995. Changes in morbidity and chronic disability in the U.S. elderly population: evidence from the 1982, 1984, and 1989 National Long Term Care Surveys. J Gerontol 50:194–204. Martelin T. 1987. Trends in elderly mortality in the Nordic countries. Compr Gerontol [C] 1:39–48. Maynard Smith J, Barker JP, Finch CE, Jardia SLR, Eaton SB, Kirkwood TBL, LeGrand E, Ness RM, Williams GC, Partidge L. 1999. The evolution of non-infectious and degenerative disease. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 267–272. McGovern PG, Folsom AR, Sprafka JM, Burke GL, Doliszny KM, et al. 1992. Trends in survival of hospitalized myocardial infarction patients between 1970 and 1985: the Minnesota Heart Survey. Circulation 85:172–179. McGovern PG, Pankow JS, Shahar E, Doliszny KM, Folsom AR, et al. 1996. Recent trends in acute coronary heart disease: mortality, morbidity, medical care, and risk factors. N Engl J Med 334:884–890. McKeown T. 1976. The modern rise of population. New York: Academic Press. Mesle F. 2004. Mortality in Central and Eastern Europe: longterm trends and recent upturns. Demogr Res 3:46–70. Moore JA, Swedlund AC, Armelagos GJ. 1975. The use of life tables in paleodemography. In: Swedlund AC, editor. Population studies in archaeology and biological anthropology: a symposium: memoirs of the Society for American Archaeology. Washington DC: Society for American Archaeology, p 57–70. Murray CJL, Lopez AD, editors. 1996. The global burden of disease: a comprehensive assessment of mortality and disability from diseases, injuries, and risk factors in 1990 and projected to 2020. Cambridge, MA: Harvard School of Public Health. National Cancer Institute. 1999. Cancer death rates among 50 countries. http://rex.nci.nih.gov/NIC_Pub_Interface/raterisk/rates34. html. ARE MODERN ENVIRONMENTS REALLY BAD FOR US? National Center for Health Statistics. 1978. Comparability of cause-of-death statistics, ﬁgures, tables, and technical notes describing trends in ischemic heart disease mortality. Hyattsville, MD: US Department of Health, Education, and Welfare. Nesse RM, Williams G. 1994. Why we get sick: the new science of Darwinian medicine. New York: Times Books. Nesse RM, Williams GC. 1999. Research designs that address evolutionary questions about medical disorders. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 16–22. Olshansky SJ, Ault AB. 1986. The fourth stage of the epidemiologic transition: the age of delayed degenerative diseases. Milbank Q 64:355–91. Omran AR. 1977. Epidemiological transition in the U.S. Popul Bull 32:1–42. Patel P, Mendal MA, Carrington D, Strachan DP, Leatham E, Molineaux N, Levy J, Blakeston C, Seymour CA, Camm AJ, Northﬁeld TC. 1995. Association of Helicobacter pylori and Chlamydia pneumoniae infections with coronary heart disease and cardiovascular risk factors. Br Med J [Clin Res] 311:711– 714. Potter CW. 2002. Inﬂuenza. New York: Elsevier. Preston SH. 1976. Mortality patterns in national populations. New York: Academic Press. Preston SH, Haines MR. 1991. Fatal years: child mortality in late nineteenth-century America. Princeton: Princeton University Press. Preston SH, Keyﬁtz N, Schoen R. 1972. Causes of death: life tables for national populations. New York: Seminar Press. Rogers RG, Hackenberg R. 1987. Extending epidemiologic transition theory: a new stage. Soc Biol 34:234–243. Salomon JA, Murray CJL. 2002. The epidemiologic transition revisited: compositional models for cause of death by age and sex. Popul Dev Rev 28:205–228. Schoﬁeld R, Reher D. 1991. The decline in mortality in Europe. In: Schoﬁeld R, Reher D, Bideau A, editors. The decline of mortality in Europe. Oxford: Clarendon Press. p 1–17. Siler W. 1979. A competing-risk model for animal mortality. Ecology 60:750–757. Stallones RA. 1980. The rise and fall of ischemic heart disease. Sci Am 243:53–59. Stearns SC. 1992. The evolution of life histories. Oxford: Oxford University Press. Stearns SC, editor. 1999. Evolution in health and disease. Oxford: Oxford University Press. 117 Strassmann BI, Dunbar RIM. 1999. Human evolution and disease: putting the Stone Age in perspective. In: Stearns SC, editor. Evolution in health and disease. Oxford: Oxford University Press. p 91–100. Szreter S. 1989. The importance of social intervention in Britain’s mortality decline c. 1850–1914: a re-interpretation of the role of public health. Soc Hist Med 1:1–37. Thompson W. 1929. Population. Am J Sociol 34:959–975. Vallin J. 1991. Mortality in Europe from 1720 to 1914: long-term trends and changes in patterns by age and sex. In: Schoﬁeld R, Reher D, Bideau A, editors. The decline of mortality in Europe. Oxford: Clarendon Press. p 38–67. Vallin J, Mesle F. 2004. Convergences and divergences in mortality. A new approach to health transition. Demogr Res 2:12–44. Van Blerkom LM. 2003. Role of viruses in human evolution. Yrbk Phys Anthropol 46:14–46. White KM. 1999. Cardiovascular and tuberculosis mortality: the contrasting effects of changes in two causes of death. Popul Dev Rev 25:289–302. Wilson C. 2001. On the scale of global demographic convergence 1950–2000. Popul Dev Rev 27:155–171. Wolleswinkel-Van-Den-Bosch JH, Looman CWN, Van-Poppel FWA, Mackenbach JP. 1997. Cause-speciﬁc mortality trends in the Netherlands, 1875–1992: a formal analysis of the epidemiologic transition. Int J Epidemiol 26:772–781. Wood JW, Milner GR, Harpending HC, Weiss KM. 1992. The osteological paradox. Curr Anthropol 33:343–370. Woods R. 1991. Public health and public hygiene: the urban environment in the late nineteenth and early twentieth centuries. In: Schoﬁeld R, Reher D, Bideau A, editors. The decline of mortality in Europe. Oxford: Clarendon Press. p 234–247. Woods R. 2000. The demography of Victorian England and Wales. Cambridge: Cambridge University Press. World Health Organization. 1955. International Classiﬁcation of Diseases, Revision 7 Vol 1. Geneva: World Health Organization. Wrigley EA, Schoﬁeld RS. 1981. The population history of England: 1541–1871. Cambridge, MA: Harvard University Press. Yanagishita M, Guralnick JM. 1988. Changing mortality patterns that led life expectancy in Japan to surpass Sweden’s: 1972–1982. Demography 25:611–624. Zatonski WA, McMichael AJ, Powles JW. 1998. Ecological study of reasons for sharp decline in mortality from ischaemic heart disease in Poland since 1991. Br Med J [Clin Res] 316:1047– 1051.