| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Department of Biobehavioral Health, The Pennsylvania State University, University Park
b Center for Developmental and Health Genetics, The Pennsylvania State University, University Park
c Institute of Gerontology, University College of Health Sciences, Jönköping, Sweden
d Department of Medical Epidemiology, The Karolinska Institute, Stockholm, Sweden
e Department of Psychology, University of Southern California, Los Angeles
Carol H. Gold, Department of Biobehavioral Health, 315 East Health and Human Development Bldg., The Pennsylvania State University, University Park, PA 16802 E-mail: gum{at}psu.edu.
Decision Editor: Fredric D. Wolinsky, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. All living pairs of unlike-sex twins born between 1906 and 1925 were identified through the Swedish Twin Registry and sent surveys assessing health and other factors. This population-based sample consisted of 605 twin pairs. Paired sample t tests were used to analyze gender differences in health-related measures, including a three-level measure of health problems based on physicians' ratings.
Results. Women had more total health conditions, not life-threatening health conditions, somewhat life-threatening cardiovascular conditions, and physical and psychological symptoms. Men had more very life-threatening health conditions and cardiovascular conditions. No gender differences were found in somewhat life-threatening health conditions, total cardiovascular conditions, or self-rated health.
Discussion. Important gender differences and similarities in health were found using an unlike-sex twin design that reduced variability due to background characteristics. This design also minimized problems caused by gender differences in survival. Research on gender and health in older persons requires more detailed approaches to address the complexity of this topic.
WOMEN have a longevity advantage over men in all industrialized countries and nearly all developing countries (Smith 1993
). The basis of this differential is clearly multifaceted, and theories invoking biological, behavioral, and sociocultural mechanisms have been proposed. Despite their greater longevity, women report poorer health and more symptoms than do men. The purpose of this study was to characterize gender differences in health using unlike-sex twin pairs. Because twins are matched for age and share the same familial influences, such a design allowed us to characterize health differences that are not contingent on differential survival or background characteristics.
Biological theories on women's longer survival focus on the role of sex hormones in modulating lipid levels (Hazzard 1986, Hazzard 1989, Hazzard 1990a, Hazzard 1990b), citing evidence that estrogen increases high density lipoprotein (HDL) levels and decreases low density lipoprotein (LDL) levels and that androgens physiologically suppress HDL levels. HDL values correlate negatively and LDL values correlate positively with the incidence of ischemic heart disease, leading Hazzard and others (e.g., Smith 1993) to postulate that a major explanation for the sex differential in longevity is the significant protection that estrogen provides women against heart disease. Differential hormonal modulation of immunity is another theory proposed as an explanation for some of the gender differences in longevity and health. Estrogen appears to increase immune responses (Waldron 1983). Hazzard 1984 proposed that women's immune advantage may also contribute to lower rates of atherogenesis, assuming that there is an immune component in that disease process. Mascart-Lemone, Delespesse, Servais, and Kunstler 1982 found there was a slower decline with age in cell numbers of T-lymphocytes in women as compared with men. Genetic explanations include those that highlight genetic evidence of a more robust immune function in women (Buckley 1986; Waldman 1986). Other biological theories concentrate on the timing of weight gain and fat distribution, both of which are implicated in heart disease (Hazzard 1986, Hazzard 1990b).
Behavioral theories have focused mainly on lifestyle, including men's higher rates of tobacco and alcohol use. Behavioral theories have also included differential use of health care and differential health and illness behavior (Mechanic 1976; Verbrugge 1990). It has been observed that women are more sensitive (perceptually) and more attentive (behaviorally) to their own health needs (Verbrugge 1985); this greater sensitivity can result in more symptom perception and reporting.
Sociocultural and environmental theories concentrate on such issues as the increased occupational hazards experienced by men (Bell et al. 1990) and the societal and cultural pressures on men to engage in risk-taking and self-destructive behavior (Smith 1993). Likewise, women have acquired risks due to stresses, unhappiness, and multiple and conflicting role pressures (Frankenhaeuser 1991; Verbrugge 1990).
A seeming contradiction in the study of gender differences in longevity is that, despite the fact that women live longer than men, women seem to have more health conditions and problems than men do. The issue is complex. For example, MacIntyre, Hunt, and Sweeting 1996 found in an examination of two British data sets that the gender in excess for a health problem varied by the condition in question and the time in the life span. The authors presented results of other research that confirmed their findings of complex relationships between gender and health (e.g., Haavio-Manila 1986; Kandrack, Grant, and Segall 1991; Wingard, Cohn, Kaplan, Cirillo, and Cohen 1989). Verbrugge 1985 reported that women have higher morbidity from acute conditions and nonfatal chronic diseases. Men, however, have proportionately fewer health conditions, but these were mainly identified as life-threatening chronic diseases, for example, coronary heart disease, atherosclerosis, and emphysema (Johnson, Mullooly, and Greenlick 1990; Verbrugge 1985; Verbrugge and Patrick 1995). However, a limitation in many of the studies examining gender differences in health was the use of only a dichotomous classification of diseases (i.e., fatal and nonfatal). A clearer depiction of gender differences in health may require a more complex paradigm.
A majority of studies have found that women report more symptoms than do men (Gijsbers van Wijk, van Vliet, Kolk & Everaerd, 1991; Grimby and Wiklund 1994; Hale, Perkins, May, Marks, and Stewart 1986; Kroenke and Spitzer 1998; Tibblin, Bengtsson, Furunes and Lapidus 1990). Verbrugge 1985 explained this phenomenon in part as a result of the kinds of conditions women have in excess—the nonfatal chronic conditions and acute conditions that have bothersome symptoms (e.g., arthritis, urinary tract infections). The reporting of symptoms is affected by a number of variables (van Wijk and Kolk 1997), including ability to recognize symptoms and frequency of occurrence. However, there is less gender differential in reporting physical symptoms than in reporting psychological symptoms, where there is more ambiguity in connection to illness and which it may be more culturally acceptable for women to report (Mechanic 1976).
Self-rated health is a commonly used indicator variable, particularly in gerontological research. Despite its subjectivity, it has shown considerable strength as a predictor of mortality (Berg 1996; Idler and Benyamini 1997; Rakowski, Fleishman, Mor, and Bryant 1993; Sundquist and Johansson 1997). Numerous studies have found no gender differences in self-rated health among older persons (Arber and Cooper 1999; Jylha, Guralnik, Ferrucci, Jokela, and Heikkinen 1998; Leinonen, Heikkinen, and Jylha 1998; Musil 1998; VanderZee, Buunk, and Sanderman 1995). Johnson and colleagues 1990 found that in 65–79-year-olds, there was little gender difference in self-reported health, but among those 80 and older, women had better self-ratings of health. Others have found that among older persons, on average, men report worse health than women (Mutran and Ferraro 1988; Ries 1983; Verbrugge and Wingard 1987). In contrast to the above findings, results from a population study of Swedish 70–79-year-olds found that women reported feeling less healthy than men, although there was no sex difference in number of definable diseases (Osterlind, Lofgren, Sandman, Steen, and Winblad 1986). In a study of older persons in 11 countries across Europe, men reported higher self-ratings of general health than did women, and more men than women rated their health as better than that of their peers (Schroll, Ferry, Lund-Larsen, and Enzi 1991). Thus, there are conflicting findings regarding gender differences in self-reported health.
A major methodological problem in the study of gender differences in health is the issue of survival differences between the two genders. Gender studies of aging, in particular, are affected by mortality differences in men and women. At a given chronological age, men represent a more selective "surviving elite" sample than women. An ideal study design to reduce the problems of gender differences in survival would be to match pairs of men and women who are as similar as possible. One promising approach is to examine intact pairs of unlike-sex twins. The genetic and environmental similarities inherent in such a sample may ameliorate to a considerable extent the selection biases that confound comparisons of older men and women. By using unlike-sex twins, the women are selected to the same extent as the men with regard to genetic (apart from sex-linked genes) and early environmental influence. We contend that this approach provides a better estimate of gender differences than conventional population studies that cannot control as well for these background characteristics.
As a subject pool, unlike-sex twins comprise a challenging yet compelling focus for multidisciplinary inquiry into gender differences in health and aging. Despite their promising utility for such studies, however, unlike-sex twins have been the least analyzed study group of twins. We have found no published studies, to date, that have used the unique characteristics of an unlike-sex twin sample to examine gender differences in aging and health.
The primary goal of this study was to assess gender differences in various measures of health conditions, health symptoms, and self-rated health among older persons by comparing brothers and sisters in a sample of unlike-sex twins. On the basis of the theoretical explanations for gender differences in longevity and health described above, and this review of the literature on gender differences in health conditions, symptoms, and self-reported health, we propose the following hypotheses:
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
) because the analytic models used to study genetic influences included only same-sex twins. The population-based Swedish Twin Registry is a scientific resource maintained at the Karolinska Institute in Stockholm. Although the unlike-sex twins' birth records were not entered into the registry, they were stored for possible future research. Approval was requested and granted by the Swedish Data Inspection Office and the Ethics Committee of The Karolinska Institute in Stockholm to carry out the study. The birth records of the unlike-sex twins were entered into the Swedish Twin Registry and linked to the National Population Registry for updated demographic information and addresses. A total of 1,699 pairs were identified as possible unlike-sex twin pairs who were still living, all of whom were sent the survey instrument. Linkage to this population registry also provided mortality data for validation and outcomes studies.
A modified Dillman 1978 technique was used to increase the response rate, which included two reminder letters and a duplicate survey packet sent to those who had not yet returned their surveys. The total number of returned questionnaires was 1,843 (54% individual response rate). These consisted of 605 cases in which both twins participated (n = 1,210); 5 cases in which both siblings answered the questionnaire, but there was doubt about whether they were twins (n = 10); 301 sisters who returned questionnaires but whose brothers did not; and 322 brothers who returned the surveys but whose sisters did not. Therefore, the male response rate was 55%, and the female participation rate was 54%. A total of 349 siblings returned blank surveys or expressed that their sister or brother was not interested in participating. Thus, 1,206 (36%) siblings did not respond in any way. It is not known how many of these nonrespondents were deceased or had an unknown address, but in light of the recent match with the updated address registry (within 6 weeks of the survey mailing), it is likely those numbers are small. The age distribution of the final sample of 605 confirmed twin pairs was as follows: 43% were 69–72 years old, 31% were 73–76, 15% were 77–80, and 10% were 81–88.
Survey Instrument
The survey instrument consisted of questions pertaining to health status and health-related behaviors and attitudes, as well as demographics and psychosocial factors such as social support. Whereas some questions were specific to this questionnaire, most were questions that have been used in other major Swedish population studies, including the Swedish Adoption/Twin Study on Aging (SATSA; see Pedersen et al. 1991) and the Origins of Variance in the Old-Old: Octogenarian Twins (OCTO-TWIN) study (see McClearn et al. 1997), both longitudinal studies based on subsets of the Swedish Twin Registry.
Measures and Analysis
Self-reported health conditions.
This was a list of common illnesses and conditions that had been used in a survey mailed to members of the Swedish Twin Registry in 1963, as well as additional items added by the SATSA study in 1984 (Harris, Pedersen, McClearn, Plomin, and Nesselroade 1992). Participants were asked to respond "yes" or "no" to the question "Do you have or have you had ______ [then came the list of 48 health problems and conditions]?" A summary variable was computed. A weakness of a summary measure of these health problems is that it gives equal weight to conditions that vary in their seriousness. To obtain a measure that reflects seriousness of health problems, a five-member expert panel of physicians evaluated the conditions in terms of their degree of being life threatening. The physicians (one male MD; two female MDs; one male MD/PhD; and one female MD/PhD) were asked to rate each of these health conditions according to one of the following three categories: very life threatening, somewhat life threatening, and not at all life threatening. If there was no unanimity on the seriousness of a particular condition, the majority rating was used. The rater reliability (calculated as an intraclass correlation) was .78 (Winer 1971). Summary scores for each of these rating categories were computed as the sum of conditions reported in the category for each respondent. Table 1 lists the conditions in each of these three categories.
|
). A variable for the somewhat life-threatening conditions was computed that excluded prostate disorders and uterus or ovary operation. An additional summary variable was computed for the total health conditions score, which also excluded the sex-specific conditions. Verbrugge 1985 encouraged this procedure when analyzing gender differences in general measures of health so that the health problems are those for which both sexes are biologically eligible. A summary score of cardiovascular conditions was also computed, as well as one for very life-threatening cardiovascular conditions (the first six conditions in column 1 of Table 1 ) and one for somewhat life-threatening cardiovascular conditions (the first three conditions in column 2 of Table 1 ), based on the physicians' ratings. There were no cardiovascular conditions that had been rated as not at all life threatening.
Self-reported symptoms.
Participants were asked to indicate on a checklist of 30 symptoms if they had had them in the past 3 months (yes or no). This checklist was based on a self-administered instrument that had been developed for use by participants before they took part in a clinical study (Tibblin 1986). A total score was computed as well as separate summary scores for the psychological symptoms (n = 10) and for the physical symptoms (n = 20) that comprised the list.
Self-rated health.
Four self-rated health questions were analyzed separately. These are questions that had also been included in the SATSA mail surveys in Sweden (Harris et al. 1992) and are based on the Duke Older American Resources Survey. Three of them—"How would you rate your health?" "How would you rate your health compared to others your age?" and "Does your health prevent you from doing what you would like to do?"—were highly correlated; a summary measure of perceived health status was computed as the total of the scores on the three questions. For all questions, a higher score indicates a more positive self-rating. Because the metrics for each item were not the same, the variables were standardized to a mean of 0 and a standard deviation of 1 before summing the three parts. The Cronbach's alpha for this summary measure was .79 for the women and .78 for the men. The additional self-rated health question ("How would you compare your health to that of 3 years ago?") was examined only as a separate item because of the different nature of this question (e.g., a person who was better now because 3 years ago he had had a heart attack could have a higher score than a person who was very healthy both at the time of the survey and 3 years ago). Paired-sample t tests were used to analyze gender differences in the variables described above. The degrees of freedom used to test significance are the number of pairs minus 1. All significance levels are based on two-tailed tests.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
|
On the basis of previous studies, we had expected that women would have more not life-threatening conditions and fewer very life-threatening conditions than their brothers. Results confirm both of these expectations. However, among those conditions regarded as somewhat life-threatening, there were no gender differences, regardless of whether the sex-specific conditions were included.
Validation procedures of our classification system were carried out using 6-year mortality data on our sample. Each member of the sample was assigned to one of three categories based on his or her most severe level of condition(s) reported in the survey data: (1) not at all life threatening, (2) somewhat life threatening, and (3) very life threatening. Analyses using independent sample t tests found that those in Category 2 were more likely to have died than those in Category 1, t(345) = 2.101, p < .05, and those in Category 3 were more likely to have died than those in Category 2, t(747) = 3.604, p < .001.
With regard to cardiovascular conditions (see Table 3 ), the expectation from epidemiologic studies was that women would have fewer total conditions. The results do not support this expectation: There appears to be no gender difference in this summary variable. However, men had significantly more of the very serious cardiovascular conditions and women had more of the somewhat life-threatening cardiovascular conditions. It is interesting to note that the t-test values are quite similar for both severity levels of cardiovascular conditions; it is only the sign that varies.
Health Symptoms
As expected from previous research on this topic, sisters had significantly more total symptoms than their brothers (see Table 4 ). This was true for both the physical and the psychological symptoms.
Self-Rated Health
As predicted, we found no gender differences in this sample on any of the measures of self-rated health (see Table 5 ).
Brother–Sister Correlations
Examination of the interclass correlations (Pearson product-moment correlations) between brothers and sisters revealed statistically significant values for the majority of the variables; the only exception was the summary measure of very life-threatening cardiovascular conditions. However, these correlations were for the most part relatively small. The variables with the largest correlations (.23 in each case) are the two summary variables (with and without sex-specific conditions) for total number of health conditions.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Also of note are the results of no gender differences for self-reported health, total cardiovascular conditions, and a category of health conditions described as somewhat life threatening. This latter category included the largest number of health conditions according to the physicians' trichotomous ratings, and the results held true even when sex-specific conditions were excluded from the analyses. The significance of this finding is twofold. First, it is important that this category of illness be assessed separately when examining the health status of older people. Simply dichotomizing health conditions into those that are fatal and nonfatal obscures the significance of a broad range of intermediate health conditions that contribute to one's health profile. For example, the finding of no gender differences in the summary score of cardiovascular conditions seems to be contrary to what one might expect from some of the epidemiologic literature. Applying the categories of somewhat life threatening and very life threatening, a more defined picture emerges. Men had significantly more of the latter conditions, and women had significantly more of the former. These findings are supported by previous studies that have found that men have more serious heart conditions and women have higher rates of hypertension (Johnson et al. 1990; Verbrugge and Patrick 1995; Verbrugge and Wingard 1987).
Second, this category of somewhat life-threatening conditions consisted of diseases and conditions that, although they may not be the direct cause of mortality, can contribute to mortality. The issue of comorbidity and its impact on mortality was investigated in a recent study (Fillenbaum, Pieper, Cohen, Cornoni-Huntley, and Guralnik 2000). Two of the five chronic health conditions examined in that study were in this intermediate group of conditions; the other three were in the most serious group. Clearly, our health profile was more than a sum of fatal and nonfatal conditions.
This research could not directly test the biological theories regarding women's longevity advantage. However, the finding that, compared with their sisters, the male twins had more very life-threatening conditions, the majority of which were serious cardiovascular problems, may provide indirect support for some of the biological theories. These include Hazzard's theories that women live longer than men because of the significant protection that estrogen provides women against serious heart disease (Hazzard 1990b); that women have an advantaged immune function against atherogenesis (Hazzard 1984); that men exhibit their greatest weight gain about 10 years before women, which correlates with the 10-year lag in symptoms of heart disease between the two genders (Hazzard 1990b); and that men more often have android patterns of fat distribution, which lie in the abdominal area (Hazzard 1986) and are associated with increased risk of cardiovacular disease. One of the other diseases on this list of most serious conditions is cancer, the risk of which is affected by the strength of one's immune function. Our finding that men had more of the very life-threatening conditions also indirectly supports Waldron 1983 theory that estrogen increases immune responses and the genetic theories of a more robust immune function in women (Buckley 1986; Waldman 1986).
The greater number of total health conditions that the sisters had provides some support for Verbrugge 1985 sociocultural theory that women have acquired health risks due to role pressures that result in more health problems, as well as her behavioral theory regarding increased use of health care (where conditions would be diagnosed) and reporting of health conditions by women. Verbrugge's theories predict that, compared with men, women will have more health problems that are likely to be not life threatening, but bothersome (Verbrugge 1985). Our results showing that the sisters had more of the not-at-all life-threatening conditions offer support for these theories, as well.
Our findings on symptom reporting support behavioral theories regarding gender differences in illness behavior (Mechanic 1976; Verbrugge 1990). Greater sensitivity can result in more symptom perception and reporting. Also, with more chronic health conditions (Johnson and Wolinsky 1994) and more of the "bothersome" conditions that Verbrugge 1985 described (e.g., rheumatoid arthritis), it is not surprising that women report more symptoms. Of particular interest is that for psychological symptoms, the difference between the women's and men's means was nearly twice that for the physical symptoms. This finding is supported by literature on symptom reporting, which indicates that men are more likely to report clearly defined symptoms than ambiguous symptoms (Mechanic 1976). It is also not surprising in the context of cultural pressures on men not to admit psychological distress and cultural messages to women that support acknowledgement of psychological symptoms (Mechanic 1976).
Previous research on gender differences in self-rated health has indicated mixed findings, depending on the age and country of the sample. As predicted, our analyses of age- and background-matched twins found no gender differences in any of the measures of self-rated health. Although these results do not appear consistent with the gender differences in health conditions that we found, research has shown that self-rated health entails more than just number of health problems (e.g., Idler and Benyamini 1997). It may be that the fewer but more serious conditions that men are more likely to have are counterbalanced by the more numerous but not life-threatening conditions that women are more likely to have. The results may also reflect the importance of the intermediate category in how the twins viewed their health, as no gender differences were found in that category. Future research should examine the relationship between self-rated health and the severity-level profile of health conditions. These results are also not consistent with the gender differences found in number of symptoms reported. Again, self-rated health appears to be a complex assessment that measures something more than number of symptoms and health problems.
Interclass correlations between brothers and sisters were statistically significant for the majority of variables analyzed, indicating rank-order stability. Although these correlations are relatively small, they suggest that familial effects are important for these measures despite the gender differences in group means for many of the variables.
There are a number of limitations regarding this study. Perhaps the most important is that the data are based on self-reports. Although the validity of self-reported assessments of health status has been questioned, some researchers have addressed this issue and found that self-reported health status measures can be substantiated with medical records and examinations (LaRue, Bank, Jarvik, and Hetland 1979; Liang 1986; Turner et al. 1997). Subjective assessments of health through surveys have been found to be predictors of mortality for both men and women (Ostlin 1990). Self-report has been shown to be reliable among older respondents (Sherbourne and Meredith 1992). These questions encompass dimensions of health status and perceptions about health in relation to aging (Levkoff, Cleary, and Wetle 1987) and thus are useful in aging research. Modeling analyses have confirmed that self-rated health is predicted by both physical health and functioning (Liang 1986). The validity of self-reported health status has been empirically verified as a single measure of overall health status, encompassing both disease and subjective assessment components (Segovia, Bartlett, and Edward 1989).
Another important limitation is that the physician-rated classification system of health conditions has not previously been used and validated. However, we believe that the results of our 6-year mortality data analyses of our sample for the three categories based on degree of threat of mortality suggest that our classification procedure has appreciable validity. Further validation efforts are planned using mortality data with cause of death.
The symptom checklist included in our questionnaire had not previously been used in a mail survey, and therefore its reliability has not been proven in mail-survey research with this population of participants. However, the format of this checklist was very simple and there were few omissions, indicating that responding to it was not a problem for the participants in our study. The checklist was developed by Swedish medical investigators for their native population, and it therefore reflects terminology of symptoms common to their culture. It was successfully self-administered by participants before their participation in a clinical study (Tibblin 1986).
A limitation inherent in any mail-survey study of older persons is that there is a risk that a family member or caregiver of the participant may have completed the survey for him or her. Our study twins were asked if they had had help with doing the survey, and if so, why. Of the 605 women, 14% said they needed help; the major reasons given were bad eyesight, difficulty writing, and "other." Of the 605 men, 13% said they needed help, with the same major reasons given. We believe that the help these participants received reduced missing data bias and that there is not likely to be any significant bias from receiving help to overcome physical handicaps. We do not know what the "other" reasons were, and this could be a source of potential bias.
The sample was limited to natives of Sweden, where the gene pool lacks the diversity of countries such as the United States. The conditions of economic security and universal health care provided by Sweden may have resulted in less variability in health status than elsewhere. Also, the question of generalizability of findings from a twin study must be addressed. However, the same factors that lead to these limitations also provide an opportunity to observe gender differences that may be confounded in a random sample of men and women from a more heterogeneous gene pool and from a population in which access to health care and economic security varies.
One of the strengths of a twin study is that it reduces the effects of uncontrolled and unmeasured factors, thus reducing the standard error and resulting in a more precise statistic. In this design of unlike-sex twins, it can be expected that the mean differences between the genders on most variables would be smaller than in a random group of men and women because twin brothers and sisters are more similar than are random men and women. At the same time, systematic variance would also be expected to be smaller, in light of the controls provided by this matched sample.
One of the limitations of a twin study of older persons is that both members of the pair must be living and must respond. The concern is that samples will be biased toward healthier individuals. In a comparison of twins from the OCTO-TWIN project with a population-based sample of nontwin octogenarians, investigators found significant effects for twin status in only 3 of 20 comparisons (Simmons et al. 1997). There were no significant effects for twin status in the domains of health care use and well-being; for all measures of vitality, except for left handgrip strength (twins were higher); for measures of physical functioning, except for instrumental activities of daily living scores (twins were higher); and for all measures of cognitive functioning, except for prose recall (singletons were higher). A recent study (Pedersen, Steffensson, Berg, Johansson, and McClearn 1999) evaluated potential survival and selection effects in the OCTO-Twin Study (of same-sex twin pairs). Not surprisingly, they found that members of those twin pairs who survived to participate had fewer illnesses at baseline (30 years before study) than did the members of the twin pairs who lost one or both members to mortality before the beginning of the study. More important, they found that there may be some selection or cohort effects for men, but not women, who survive to their 80s. Although only 14% of the twin pairs of our study were 80 and older, it is interesting that in this study we did find that men had more of the very life-threatening conditions than their sisters, despite a potential selection for healthy surviving men. It may be that the sisters in this unlike-sex twin design had selection effects similar to those of their brothers, that is, biased toward being healthier because of their shared genetic and early environmental factors with their brothers. This may suggest that our unlike-sex twin design was successful in helping to deal with the elite male survivor phenomenon that confounds research on gender differences in aging. If this is true, we can be more confident about the gender differences we found.
Another important issue regarding twin studies involves genetic influences on disease. Many disease conditions have been found to have a heritable component. These include cardiovascular conditions such as hypertension (Tambs et al. 1993), diabetes (Matsuda and Kuzuya 1994), asthma (Nieminen, Kaprio, and Koskenvuo 1991), lung cancer (Braun, Caporaso, Page, and Hoover 1994), and affective illness (Kendler, Pedersen, Johnson, Neale, and Mathe 1993). Thus, the interclass correlations for disease-related variables may be higher as compared with those that would be found between a random sample of men and women, due to these familial effects.
Whereas the data on which this study was based are informative regarding gender differences, they are cross-sectional in nature and limited to health-related assessments at one point in time. The question arises: Are these results sufficiently robust to multivariable analyses that incorporate additional variables regarding the current life circumstances of these brothers and sisters? In-person testing is being conducted on a subset of the twins described in this study. This data collection includes assessments of some of the same measures as in the mail survey, but also includes assessments of biomedical, physical, cognitive, and interpersonal functioning, as well as actual medicines being used and psychosocial measures of interest, including indicators of current life circumstances. Plans are being formulated for linkage to medical records and cause-of-death files. Thus, these longitudinal follow-ups and data linkages will provide extraordinary opportunities to find gender differences in the aging process through multivariate analyses of combinations of both objective and subjective measures of health and functioning. In addition, it will be possible to integrate these data with same-sex twin data to determine if sex-specific effects are operating in the genetic and environmental components of the variance in selected health-related phenotypes.
The major findings in this study are threefold. First, we have confirmed important gender differences in health indicators in older persons using a unique sample—a population-based sample of unlike-sex twins—that partially controlled for the confounding effects of elite male selective survival and genetic and environmental effects through this study design. Second, we have found that in this matched sample, there were no differences in self-rated health, despite the gender differences in reported health problems and symptoms. Third, we found that using a three-level severity-of-illness schema paints a much richer picture of gender and health than using the dichotomy of fatal and nonfatal conditions most commonly found in the literature. With this more complex paradigm, we have revealed an important category of illnesses and conditions in which no gender differences were found—the somewhat life-threatening health problems. This grouping included the largest number of problems and thus is a very critical category to examine when trying to disentangle the complexity of gender differences and similarities in health and aging. Perhaps it is also a reminder to focus not only on gender differences. Research on gender and health requires new approaches in measurement and study design. We have only begun to uncover the richness of this area of research.
|
Acknowledgments |
|---|
Received for publication September 26, 2000. Accepted for publication June 22, 2001.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  E. Guilley and C. J. Lalive d'Epinay Social Status and Mortality With Activity of Daily Living Disability in Later Life J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2008; 63(3): S192 - S196. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Braungart Fauth, S. H. Zarit, B. Malmberg, and B. Johansson Physical, Cognitive, and Psychosocial Variables From the Disablement Process Model Predict Patterns of Independence and the Transition Into Disability for the Oldest-Old Gerontologist, October 1, 2007; 47(5): 613 - 624. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. E. Alkema and D. E. Alley Gerontology's Future: An Integrative Model for Disciplinary Advancement. Gerontologist, October 1, 2006; 46(5): 574 - 582. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
