One of the most frequently noted observations among healthcare professionals is the considerable variation in the vitality of our rapidly growing population of elders. Why is it that some individuals survive to their Los, 80s or 90s, relatively vigorous and free from chronic illness, and maintain their cognitive abilities, while others experience considerable physical and cognitive impairment with a reduction in activities of daily living? Is it because these individuals are genetically different or because they have accumulated different experiences throughout life? In the terms of quantitative genetics, we want to know how important, relatively speaking, genes and environments are for the individual differences we see among individuals. Clearly, both genes and environments are important for aging. What is not as well delineated is whether the influence of genetic effects is the same throughout the last half of the lifespan, or whether the importance of these effects is different for different characteristics like health, memory, and functional abilities.
Twins represent an opportune natural "experiment" for studying the relative importance of genetic and environmental effects. Identical twins share all of their genes in common. Hence, any differences between members of a pair must be due to environmental differences. Fraternal twins, like other siblings, share half of their genes in common. Thus, we compare the similarity of identical and fraternal twin pairs in order to estimate how important genetic effects might be. If genetic effects are of importance, then identical twins should be twice as similar as fraternal twins. The extent to which identical twins are different provides an estimate of the importance of what is known as "nonshared environments;' that is, those individual-specific environmental factors that cause differences among family members.
Although the use of twin studies for the study of human health and behavior has a tradition dating from the late 1800s, there have been relatively few twin studies of aging. The first twin study of aging was that by Kallmann, Jarvik, and colleagues, known as the New York State Psychiatric Institute Study of Aging. After this pioneering effort, which focused mostly on cognitive abilities, mental health, and longevity, most of the recent studies have been broad multidisciplinary efforts using population-based samples of twins (Pedersen, 1996). There are currently seven twin studies of aging, four of which are based on twins in Sweden and Denmark. These studies includethe following: the Swedish Adoption/Twin Study of Aging (sATsA), which includes a sample of twins separated at an early age and reared apart; ocro Twin, comprising pairs who are 80 years of age or older (focusing on "successful agers"); the GENDER Study of unlike-sexed twins (which helps us get a handle on sex linked differences in aging); the Longitudinal Study of Aging in Danish Twins (LSADT ); the Minnesota Study of Adult Development; the Black Elderly Twin Study (BETS ); and the National Heart Lung and Blood Institute (NHLBI) study of male twin veterans. Each of these studies provides a unique contribution to the understanding of how genetic and environmental factors affect human aging. In the following, I will provide a review of some of the results, starting with cross-sectional findings and moving to longitudinal findings when these are available.
Perhaps one of the most pertinent characteristics for the gerontologist to study is the subject's self rated health, as this measure reflects the individual's own opinion of his or her status and is a predictor of survival. Self rated health can be assessed with the simple question, "On the whole, how do you rate your health?" or augmented to elicit additional information, as in "How would you rate your health compared to 5 years ago?" and "Does your health prevent you from doing things you would like to do?" Countless epidemiological studies have demonstrated that older individuals report poorer health than younger individuals do. What is perhaps more relevant to questions of genetics and aging, however, is the finding that there is a considerable increase in variation in self rated health in older age groups. Both the range and degree of individual difference are greater in the elderly Does this increase in variation reflect genetic differences among the elderly or an accumulation of experiences and exposures that influence health? Lifespan developmental theories of aging predict that the latter is the case.
Early findings from SATSA. by Harris and colleagues (1992) indicate that the cross-sectional increase in individual differences with increasing age indeed reflects an increase in environmental sources of variation. The notable increase in variation after the age of 70 was exclusively due to an increase in environmental effects. The next question is whether this finding is an age-related finding or whether it is a function of the birth cohort. SATSA twins who were 70 or older at the time of measurement were born prior to World War I. Could it be that the SATSA results merely reflect the very special circumstances for individuals surviving the calamities of the beginning of the twentieth century?
We have recently been able to replicate the SATSA findings in a sample of Swedish twins interviewed fifteen years later (Svedberg, Lichtenstein, and Pedersen, 1999). The same conclusion emerges-that the increase in variation is a result of increased environmental effects, especially after 70 years of age. Thus, it is unlikely that the cross-sectional findings are exclusively the result of a historical effect. However, an important question still remains: What happens if we follow the same sample of subjects across time? Will the same group of individuals become more variable in their responses across time? Or do we introduce extra variability by evaluating samples cross-sectionally?
These questions may, at first glance, seem confusing. After all, we are born with all of our genes-doesn't it make sense that any increase in variability is due to the environment? The key is to remember that although we are born with a full complement of genes, not all of them are active at any one moment, and certainly not in all cells. Genes may turn on and off, and the timing of functioning certainly differs among individuals. Similarly, not all individuals have the same experiences or culturally determined rites of passage at the same rate. Typical examples of genes turning on and off are puberty and menopause. There are certainly many more, such as age-related activation of genes for certain diseases. Environmental effects may be accumulated in a continuous fashion, but in some circumstances, single events may be sufficient to result in lifelong changes.
The best way to address these questions necessitates longitudinal assessments of a relatively large sample of individuals over several years. Although longitudinal data concerning self rated health are available from saTSA, we have not yet been able to perform the most appropriate types of analyses to those data. Instead, I will digress with an example from the realm of cognitive abilities.
The first cross-sectional (meaning at the same time) assessment of SATSA, including twins from 50 to 85 years of age, indicated that the relative importance of genetic effects-the heritability-for general cognitive abilities was substantial. Eighty percent of the variation in cognitive abilities was attributable to genetic differences among individuals (Pedersen et al., 1992). This finding was remarkable at the time. Despite earlier findings, predominantly from young people under the age of 40, few imagined that genetic influences would account for such a great proportion of individual differences in cognitive abilities in the second half of the lifespan. The design of SATSA, with twins separated at an early age and reared apart as well as conventionally reared twins, provided one of the few opportunities to disentangle the genetic and shared familial environmental factors as sources of familial similarity. These first results suggested that genetic effects cast a long shadow and that the experiences early in life are less important as a source of individual differences later in life.
Other cross-sectional analyses of SATSA demonstrated that heritability for general cognitive abilities was somewhat lower in the oldest group of sATSA twins (those 70 and older) (Finkel et al., 1995). This finding was subsequently confirmed in octo Twin, for which the estimate was 60 for octo- and nonagenarian twins (McClearn et al., i997), suggesting that there may be a decrease in the importance of genetic effects or an increase in the importance of environmental effects late in life. What we may be witnessing is the effect of history, in that there may be differences in the importance of genetic effects for individuals born 191 or earlier, much like the findings reported above for self rated health. This hypothesis is not unreasonable, considering the selection effect that the influenza epidemics and outbreak of WWI may have had on Europeans.
Remember that these results are cross-sectional-from one point in time. When the second wave of SATSA results came in three years later, at first glance the cross-sectional results appeared to be confirmed. Across a three-year period, the heritability estimates remained stable (Plomin et al., 1994). Indeed, when yet another wave of data was analyzed, the estimates again appeared to be stable. However, these early analyses have serious limitations: In order to be included in the analyses, both members of the pairs had to participate at all time points. It might be that those who continue to participate are very different from those who drop out of our studies. Examination of average levels of performance indicated that for those pairs in which both members participated at all three waves, performance was relatively stable (only slight declines). In contrast, there was an initial lower level followed by a slightly sharper decline in performance between the first two occasions for pairs in which one or both ceased to participate after Time 2, and those pairs participating at Time r only performed the least well at that occasion. Most of the individuals who terminated participation were too sick to participate or died before the next testing occasion.
This pattern is a feature of a phenomenon known as "terminal decline" (Berg,1996), which is characterized as a period just before death (a to 5 years), during which there are often declines in cognitive and functional abilities that predict subsequent mortality better than knowledge of specific diseases. Thus, those individuals who could not or did not participate in subsequent waves of saTSA may have entered into this period of "terminal decline." By requiring that a twin participate in all measurement occasions, serious limitations on the generalizability of the results were imposed; the net result was that we evaluated only stable individuals, and not those experiencing decline.
What is the consequence of these limitations for subsequent genetic analyses? Examination of the patterns of intraclass correlations for identical and fraternal pairs (the sine qua non for estimates of heritability) indicated that similarity differed as a function of participation patterns. Thus, by requiring that both members participate in all occasions of measurement, a pattern mimicking greater genetic stability was imposed, and we had limited the likelihood of finding change in the relative importance of genetic and environmental influences. In fact, the pattern of correlations for pairs dropping out suggested that environmental influences might be of considerable importance for the timing of entry into "terminal decline."
Up to this point, the sATSA data indicated that there may be some decrease in heritability in the oldest old, but it was not clear whether the results reflect cohort differences or true aging effects. Recent developments in various statistical methods allow for missing values for some participants and some variables. "Cohort-sequential" analyses combining cross-sectional, from one test time, and longitudinal information confirm a general decline in mean levels of cognitive performance but no change in total variance for general cognitive abilities. Thus, earlier predictions that individual differences should increase with advancing age were once again not supported.
More interesting were the analyses of the genetic and environmental components of variance. Inspection of the longitudinal trends for the separate cohorts clearly demonstrated that there is a longitudinal decrease in genetic variance for the oldest cohorts (Finkel et al.,1998). Heritability is relatively stable longitudinally, with approximately 80 percent of differences attributable to genetic factors in the younger cohorts. On the other hand, heritability decreases from approximately 80 percent at Time r to 60 percent at Time 3 in the older cohorts. (Incidentally, this is the same heritability as that reported for twins 80-plus years old in octo Twin.) We may be observing the effect of terminal decline in cognitive abilities. It may be that twin similarity for cognitive abilities decreases as members of a twin pair begin to decline at slightly different times or the rate of decline differs for the members of the pair. Thus, it appears as though environmental factors are important for the timing of entry into or the trajectory of terminal decline.
INFLUENCES ON DECLINE AND RATE OF CHANGE
Thus far we can see that the increase in individual differences across age groups reflects an increase in environmental influences. Further, genetic influences for health and cognition appear to be relatively stable in mid life but decrease in importance in the oldest-old. SATSA analyses of cognitive abilities suggest that environmental influences might be important for decline, but we have not yet been able to evaluate what influences the rate of decline. Preliminary analyses of various indicators of decline by Reynolds and colleagues (1997) suggest that environmental influences are most important for the rate of change for all cognitive abilities with the exception of a measure of perceptual speed. The picture is complicated by methodological difficulties but should become clearer once we are able to apply growth-curve models to the data.
The findings regarding rate of decline are consistent with findings for measures of functional ability (such as grip strength and manual dexterity) by Grant (1997). Variation in decline in these measures was exclusively due to environmental influences. These results have considerable practical implications, as these traits are related to those activities that may lead to disability and the need for assistance. Thus, appropriate environmental interventions are likely to be effective in reducing the rates of decline in ability to perform daily activities.
These are but a few examples of how genetic and environmental influences are important for stability and change late in life. There are no general rules for how important genetic factors are; their importance is a function of the trait, age, and sex. Cross-sectional differences are not always mirrored in longitudinal changes. Influences on stability are not necessarily the same as those that influence rate of change. And, perhaps most important, we should remember that evidence for a significant genetic influence does not mean that a trait is immutable. Finding the genes for disease will help us understand interactions with the environment and ways to cure, prevent, or delay onset of disease and reduce disability.
Berg, B. D. 1996. "Aging, Behavior, and Terminal
Decline, - In J. E. Birren and K. W Schaie, eds., Handbook of the Pychology of Aging, 4th ed. San Diego, Calif.:
Finkel, 4-th ed. San Diego, et al. i995. "Heritability of Cognitive AbilAcademic Press.
Finkel, D et al 1995 "Heritability of Cognitive Abilities in Adult Twins: Comparison of Minnesota and Cross-SecSwedish Data." Behavior Genetics 25;421-31
Finkel, D., et al. 1998 "Longitudinal and Cross-Sectional Twin Data on Cognitive Abilities in Adulthood: The Swedish Adoption/Twin Study of Aging." Developmental Psychology 34;1400-13
Harris J R., et al 1992. "Age Differences in Genetic and Environmental Influences for Health from the
Swedish Adoption/Twin Study of Aging."Journal of Gerontology: Psychological Sciences 47:rai3-ao.
Grant, J. D. c997. An Assessment of the Genetic and Envronmental Contributions to Stability and Change in Selected Indica rs of FunuimialAbility. Unpublished doctoral thesis, College Park: Pennsylvania State University.
McClearn, G. E., et al. i997. "Substantial Genetic Influence on Cognitive Abilities in Twins 80 or More Years Old." Science 276: i56o-3.
Pedersen, N. L. 1996. "Gerontological Behavior Genetics." In J. E. Bitten and K. W Schaie, eds., Handbook of the Psychology of Aging, 4th el. San Diego, Calif:: Academic Press.
Pedersen, N. L., et al. 1992. A Quantitative Genetic Analysis of Cognitive Abilities During the Second Half of the Life Span." Psychological Science 3:;46-53.
Plomin, R, et al. 1994. "Variability and Stability in Cognitive Abilities Are Largely Genetic Later in Life." Behavior Genetics 24:207-If .
Reynolds, C. A., Gatz, M., and Pedersen, N. L.1997. Sources of lndividual V ariation fr Cognitive Decline: Quantitative Methods for Describing Patterns of Change. Psychiatric Genetics Conference, Santa Fe, N.Mex.
Svedberg, P, Lichtenstein, P, and Pedersen, N. L. i999. "Age and Sex Differences in Genetic and Environmental Factors for Self Rated Health: A Twin Study.' Unpublished manuscript.
Nancy L. Pedersen, Ph.D., is professor of psychology a genetic epidemiology, Department of Medical Epidemiology, Karolinska Institute, Stockholm, Sweden.…