Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 60
Biosocial Surveys 3 The Taiwan Biomarker Project Ming-Cheng Chang, Dana A. Glei, Noreen Goldman, and Maxine Weinstein It takes courage, optimism, and a remarkable degree of cooperation to field a study that engages colleagues from opposite ends of the earth. The Taiwan biomarker project—SEBAS (Social Environment and Biomarkers of Aging Study) for short—owes its success to dedicated teams in both Taiwan and the United States. Our discussion in this chapter begins with an overview of the Taiwan biomarker project. We then review project logistics, summarize our substantive findings to date (and comment on the work we see in the near term), and offer some thoughts about how our experience might inform future research. Along the way, we engage in a candid discussion of the problems we have encountered. OVERVIEW OF THE PROJECT SEBAS builds on a longitudinal study of the elderly and near-elderly population of Taiwan. The Study of Health and Living Status of the Elderly in Taiwan was initiated by the Taiwan Provincial Institute of Family Planning (now the Bureau of Health Promotion, Department of Health) as a collaborative effort with Albert Hermalin at the University of Michigan. The first survey was conducted in 1989; 4,049 persons age 60 and older were interviewed (a response rate of about 92 percent) (Chang and Hermalin, 1989). The survey included eight modules on (1) marital history and other demographic characteristics; (2) household roster, social and economic networks, and exchanges; (3) health, health care utilization,
OCR for page 61
Biosocial Surveys and health-related behaviors; (4) occupational and employment histories; (5) activities and attitudes; (6) residential history; (7) economic and financial well-being; and (8) emotional and instrumental support (Weinstein and Willis, 2001). Since 1989, follow-up interviews have been conducted in 1993, 1996, 1999, and 2003. In both 1996 and 2003, the study drew a refresher sample to provide a sample of persons age 50 and older. The initial impetus for the biomarker arm of the Taiwan study of the elderly grew out of a seminar on the cumulative effects of stress on health that was presented by Burton Singer at the Office of Population Research at Princeton University in 1995. The focus of his presentation was the MacArthur Study of Successful Aging—a study of predominantly high-functioning individuals drawn from community-based cohorts that were part of the Established Populations for Epidemiological Studies of the Elderly. The longitudinal study of the elderly in Taiwan seemed to offer an opportunity to do a population-representative study—albeit of persons middle-aged and older—that incorporated biomarkers. The study had been going on for some years, there was a strong base of sociodemographic data, the institute in Taiwan had a competent staff and substantial experience fielding surveys, we had a long and productive history of cooperative work with each other, and we knew that the study sample was cooperative and responsive. We have already presented a (simplified) diagram of our basic theoretical model in the predecessor to this volume, Cells and Surveys: Should Biological Measures Be Included in Social Science Research? (Weinstein and Willis, 2001, p. 259). Underlying the study was our interest in exploring the (often) reciprocal relationships linking the social environment with stressful experience and with health outcomes, and in elaborating the physiological responses that lie between those links and between stressful experience and health outcomes. There are huge—and growing—literatures linking the social environment with exposure to challenge, linking the social environment with health outcomes, and some linking exposure to challenge with health outcomes. What we hoped to add to the discussion (primarily) were better data on the physiological pathways that lie between the environment and health outcomes and the physiological effects of exposure to challenge. Our original approach to incorporating physiological dysregulation was based on the concept of allostatic load. The idea behind allostatic load is that stressful experience causes a chain of physiological changes that interrupt normal processes; repeated or prolonged exposure to such stressors can result in physiological dysregulation (McEwen, 2002; McEwen and Stellar, 1993). Proponents of the framework would argue that allostatic load can be viewed as an index of the relative degree of failure at a physiological level—a marker of the cumulative physiological costs of efforts to cope with life’s challenges.
OCR for page 62
Biosocial Surveys The realization of allostatic load used in the few studies that had already been done by the time of the first biomarker collection in Taiwan was a simple index of 10 parameters (and our choice of biomarkers was driven by them) based on data from various domains of physiological regulation. Early in the MacArthur studies (and in some of our work that compares the Taiwan data with the MacArthur results), allostatic load was measured by summing the number of parameters for which individuals fell into the highest risk quartile over each of the components. In recognition of its limitations (e.g., equal weight given to each parameter, the definition of high risk as only a single tail of a given distribution, exclusion of parameters pertaining to immune function and, more generally, the limited number of markers), this formulation is evolving (Seplaki, 2004, 2005, 2006a) and undoubtedly will continue to evolve, as our understanding of physiological pathways increases. We have found, for example, that the simple score can be improved by incorporating additional biomarkers and allowing for two tails of risk (when relevant). At the same time, a simple index has the potential to mask important information disclosed by analysis of individual biomarkers. As discussed by Wachter (2001, p. 335), the measure is an example of “dimensionality reduction” that could be accomplished in a number of ways. Findings to date suggest that indicators over multiple systems provide greater power than single markers in understanding the costs and consequences of stressful experience (Cohen, 2000; Karlamangla, Singer, McEwen, Rowe, and Seeman, 2002; Seeman, McEwen, Rowe, and Singer, 2001). Still, the pre-2000 surveys provided data that allowed us to take a longitudinal look at the effects of the social environment on the physiological markers and the effects of the social environment on health. Our findings are discussed following a description of the project logistics. PROJECT LOGISTICS We did a pilot test of the biomarker collection in December 1997-January 1998. The idea was to assess feasibility: we obtained information regarding the logistics of collecting specimens and transporting them to the lab, training interviewers and staff, pretesting the instruments, and assessing participation rates (see Weinstein, Goldman, Hedley, Lin, and Seeman, 2003). SEBAS I—the first biomarker collection—was fielded between July and December 2000. We drew a random subsample of 1,713 persons ages 54 and older from those interviewed in 1999. Older persons (those age 71 and older in 2000) and persons in urban areas were over-sampled (for details see Goldman, Lin, Weinstein, and Lin, 2003 or Goldman, Glei, Turra, Glei, Lin, and Weinstein, 2006a). The study consisted of two parts: a face-to-face interview and a hospital-based examination. Our goal was to complete hospital examinations for 1,000 partici-
OCR for page 63
Biosocial Surveys pants. The choice of a hospital setting for the tests was driven in large part by three factors. The first was the requirement in Taiwan that phlebotomy be performed by (or under the close supervision of) a physician. Second, we wished to provide some incentives for participation. The hospital setting allowed us to offer an abdominal ultrasound and provided the setting in which a physician could perform an examination similar to what the respondents would have received under National Health Insurance coverage. Third, our concern for the safety of the participants was the paramount consideration. We wanted to ensure that any problems could be addressed swiftly and effectively. The face-to-face interviews were conducted by a local public health nurse who was well known and highly respected in the local area. A total of 1,497 persons responded to the home interview (93 percent of the elderly survivors and 91 percent of the near-elderly survivors). The in-home interview updated information on each respondent’s living situation, employment, and marital status. It included questions on health (e.g., measures of depressive symptoms, cognitive function, activities of daily living, and instrumental activities of daily living), use of health care services, participation in social activities, stressful experiences and assessment of sources of stress and anxiety, and how respondents were affected by the 1999 earthquake (7.3 on the Richter scale and centered in Chi-Chi). The respondents also provided a subjective assessment of their position on a “social hierarchy ladder” separately for Taiwan as a whole and for their community. At the end of the interview, the interviewer evaluated whether the respondent’s health would allow participation in the hospital protocol. Exclusion criteria included living in an institution, being seriously ill, needing a catheter or diaper, being on kidney dialysis, or having another health condition that would preclude blood drawing. About 7 percent of the respondents were ineligible for the examination. Eligible respondents were asked to participate in a health examination at a nearby hospital. The interviewer explained the protocol, scheduled the examination (for several weeks later), and arranged transportation if needed. Hospitals were chosen by the Bureau of Health Promotion (BHP) based on reputation, accessibility, whether they had the interest—and capacity—to participate in the study, and whether they were within or close to a selected primary sampling unit (PSU). In the end, 24 hospitals were recruited; a few served more than one PSU. The night before the hospital appointment, a BHP staff member together with a public health nurse delivered a urine collection container (for an overnight 12-hour urine specimen) to the respondent’s home, explained the proper procedures, provided written instructions for urine collection, and answered questions. They obtained informed consent for the health examination and reminded the participants not to eat anything from midnight until
OCR for page 64
Biosocial Surveys after the examination was completed. The next morning—the morning of the hospital examination—a member of the BHP met the participant at home, picked up the 12-hour urine specimen, and accompanied the participant to the hospital.1 During the hospital visit, a team of seven staff members coordinated the respondent’s visit, processed the blood (collected at the hospital) and urine specimens, and processed the forms. Participants were asked about their health history, family disease history, health-related behaviors, and current long-term medications. A member of the team also confirmed that the participant had followed the urine collection procedure, had fasted since midnight, and did not have any contraindications to a blood draw. In addition, the team member reinterviewed the participant in order to obtain responses to questions that remained incomplete from the initial home interview and to resolve any inconsistencies in the home interview that were found when the staff reviewed and edited the questionnaire the previous evening. At the hospital, the participant provided a spot urine specimen, a phlebotomist drew a blood specimen, and a nurse measured the participant’s height, weight, waist and hip circumference, and blood pressure. Two blood pressure readings (about one minute apart with the respondent in a seated position) using a mercury sphygmomanometer on the right arm were taken at least 20 minutes after the participant arrived at the hospital. A physician performed a medical examination that included a third blood pressure reading, an abdominal ultrasound, and health counseling. Among participants in the medical exam, compliance with the clinical protocol was high: all but 10 individuals collected the 12-hour urine sample, provided a sufficient volume of blood for analysis, and completed the medical exam. After the examination, participants were provided with breakfast, were given a small gift of nutritional supplements, and were accompanied back to their home. By noon on each day of the hospital visit, a staff member of Union Clinical Laboratories (UCL) based in Taipei collected the blood and urine specimens from the hospitals. UCL was responsible for transporting the specimens to Taipei. They followed standard lab procedures for the assays (details regarding the assays are provided in Seeman et al., 2004, and Goldman, Glei, Seplaki, Liu, and Weinstein, 2005) and returned the results to the BHP within two weeks.2 Several weeks after the 1 Some respondents chose to come to the hospital on their own or by taxi. In these cases, the respondents brought the urine collection container with them. 2 In addition to the routine standardization and calibration tests performed by UCL, nine individuals (outside the target sample) contributed triplicate sets of specimens during the early stages of the fieldwork. In each case, two sets were submitted to UCL and a third was sent to Quest Diagnostics in the United States.
OCR for page 65
Biosocial Surveys fieldwork, participants were sent the results of the standard tests based on the blood and spot urine specimens and the findings from the physical examination. Participants whose results were outside normal ranges were encouraged to see a physician for further examination and were informed about health counseling services available at the hospital. Of the 1,497 respondents to the in-home interview, 1,023 participated in the hospital protocol (75 percent of the near-elderly and 61 percent of the elderly). Among the approximately 24 percent who declined to participate (as noted above, 7 percent were ineligible), the primary reasons were that the respondent felt that she or he was healthy and did not need an exam, that the exam was too much trouble, that she or he had just had a health exam, and that the respondent had no free time or was out of town during the several-day period that the exams were offered. Disproportionately high nonparticipation rates were found among the healthiest respondents and the least healthy. Overall, persons who received the medical exam reported the same average health status (on a five-point scale) as those who did not. Although respondents over age 70 were less likely than younger persons to participate, sex and measures of socioeconomic status were not significantly related to participation (Goldman et al., 2003). A full list of the biomarkers and clinical measures that we obtained is presented in Box 3-1. Briefly, in addition to a hematology panel routinely collected as part of a health examination, we tested for total and high-density lipoprotein (HDL) cholesterol, glycosylated hemoglobin, dehydroepiandrosterone sulfate (DHEAS), IGF-1, and IL-6 from the blood specimens. We used the 12-hour urine specimens for assays of cortisol, epinephrine, norepinephrine, dopamine, and creatinine. At the suggestion of the study section that initially reviewed the proposal, we also added determination of apolipoprotein E (ApoE) genotype. We stored three sets of aliquots of the specimens: one set at the BHP, one at UCL, and one at Georgetown University. Our experience with shipment to the United States was scary: one set of the reliability/validation specimens—fortunately not specimens from our sample—was held up in customs on the West Coast. This happened despite the fact that we had obtained all the necessary clearances. We also found that we had to ship the (sample) specimens from Taiwan to the United States in two separate shipments; evidently the amount of dry ice required for a single shipment exceeded the post-9/11 guidelines. The storage freezers are kept at –80 degrees C. All the storage facilities have emergency CO2 back-up systems; the Georgetown specimens are housed at the Medical Center in a building that also has an emergency electrical generator. We have not yet received any requests for the use of the stored specimens (other than our own plans
OCR for page 66
Biosocial Surveys BOX 3-1 Biomarkers and Clinical Measures Collected During the SEBAS 2000 Measure Physical Examination Anthropometry: height, weight, waist and hip circumference Systolic and diastolic blood pressure (3 readings) Examination of chest, heart rate, breathing, breasts, abdomen, arms, legs, lymph and thyroid glands for abnormalities (similar to National Health Insurance Exam) Abdominal ultrasound (liver, pancreas, gallbladder, kidneys) Fasting Blood Sample Total and HDL cholesterol Glycosylated hemoglobin DHEA-S ApoE genotype Immune function and growth factor: IL-6, IGF-1, Other routine blood tests (e.g., blood cell counts, hemoglobin, glucose, triglycerides) 12-hour Urine Sample Cortisol Norepinephrine Epinephrine Dopamine Creatinine SOURCE: Glei et al. (2006). for additional assays); however, for the second round, we have made a provision for an advisory board to review such requests. Moving forward to the second round of biomarker collection we have realized that one issue that we did not anticipate adequately is change in assay techniques. We recently discovered that standards for some assays have changed since 2000; indeed, they have changed to the extent that the reagents we used in 2000 are no longer available. UCL will be performing duplicate assays on the stored pilot study specimens with the new assays to see how well the results replicate the previous assays; however, at least for the assays that require fresh specimens, we will not be able to distinguish between the effects of sample deterioration and change in assay method. An important question in adding biomarkers to general surveys
OCR for page 67
Biosocial Surveys relates to whether the additional burden imposed by the biomarker collection compromises participation in the parent study. We have no evidence that this has happened in Taiwan. The most recent round of household interviews, carried out by BHP as part of the Study of Health and Living Status, was done in 2003, based on a questionnaire similar to those in earlier waves of the survey. The response rate of 92 percent is consistent with rates from previous waves of the survey, suggesting that previous participation in the biomarker study did not affect participation rates in the main longitudinal study.3 Our pretest for the second round of biomarker collection was conducted in September-October 2005. Of the 66 pretest participants in the household survey who were eligible for the hospital protocol, three reported that they were either tired of participating in the study as a whole or had a bad experience with the earlier (2000) biomarker study. It remains to be seen what will happen as we go into the field with the second round of the biomarker work. Issues relating to human subjects and the process of informed consent are ongoing concerns for us. In particular, we are working with an elderly population, many of whom—particularly women—cannot read (31 percent overall, 50 percent among women). For the second round, we have consent procedures for respondents who are able to provide consent for themselves and separate procedures for those who require the assistance of a proxy. The low literacy rates affect what we can include in the protocol as well. A complex protocol, with regard to the urine collection for example, is difficult and time-consuming to explain. In the second round, we had hoped to collect three saliva specimens (at bedtime, next morning before getting out of bed, and 30 minutes after waking up), but our pretest suggested that participants, particularly illiterate ones, simply could not handle both sets of instructions (for urine and saliva). The data from SEBAS I have been released through the Inter-university Consortium for Political and Social Research. The files include almost all the information that was collected at the time of the National Institutes of Health–funded 2000 study, along with some basic demographic data that were collected during the previous waves of the study (funded by the Taiwanese government). In order to protect the identity of the participants, some data were not included in the public release; we replaced identifiers for respondent, town of residence, PSU, hospital, examining physician, and town where respondent was during the earthquake with randomly generated identifiers, and we excluded other identifiers (i.e., for interviewer, type of organization in which the respondent works) and all of the 3 Among the SEBAS participants, the response rate in 2003 was 97 percent. This very high response rate may be attributable, however, to the fact that it is based only on those who did not refuse the examination in 2000 and who were not lost to follow-up by 2000.
OCR for page 68
Biosocial Surveys information regarding the arrangements for the hospital visit and reasons for nonparticipation. We did not anticipate the complex issues involved in responding to requests for cross-walks between the biomarker data and previous waves of the study. Apart from the 2000 wave of the study, which has been made publicly available as described above, the longitudinal data are the intellectual property of the BHP. Researchers who want to use the data can apply to the BHP for their use; upon approval, such researchers have been provided copies of (some) data. The problem arises because the biomarker study does not have any contractual arrangements with these researchers regarding release or dissemination of the longitudinal data. If those data have already been shared with other researchers, and if we were to provide a cross-walk with the biomarker data, the possibility of deductive identification of the participants becomes very real indeed. A second problem—and we are working on this issue now—concerns the release of a second round of data. Again, the greatest concern is the need to protect our participants from deductive identification. We are working with Jim McNally, director of the National Archive of Computerized Data on Aging, and his staff in preparation for this release. There is also the need for ongoing support and funding by the research group—support that may not be available after the completion of the grants that fund the Taiwan project. This funding concern is also relevant to the storage and review of the use of the biological specimens. We have not yet identified a solution; however, the Behavioral and Social Research Branch of the National Institute on Aging is currently investigating these issues; we hope for some guidelines—and some support—as we go forward. The need for a curator (for the archived specimens and for the data) to ensure that these resources continue beyond the life of the grant (or the original researchers, for that matter) is very real indeed. Such an effort cannot be sustained indefinitely out of pocket of the research grant or the investigators. This section on logistics would not be complete without acknowledging the work of our colleagues in Taiwan. We cannot overstate the contributions of the dedicated staff at the BHP. The three team leaders—Yu-Hsuan Lin, Shu-Hui Lin, and I-Wen Liu—and their supervisor, Yi-Li Chuang (now director of the Population and Health Research Center of the BHP) were remarkable. The interviewers, field workers, drivers, supervisors—absolutely everyone—pitched in to ensure the success of the project. SUBSTANTIVE FINDINGS AND NONFINDINGS Our study design integrated both biological and survey data from a population-representative sample; here we focus on results based on both
OCR for page 69
Biosocial Surveys sources, although our discussion includes findings from all aspects of the design. In Taiwan—as elsewhere—the social environment matters for an individual’s health, but, unlike some previous work, we found that not all aspects of social ties or social support seem to matter. Specifically, contact with friends and participation in social or religious activities are associated with lower rates of functional disability, better survival, better self-rated health, fewer depressive symptoms, better cognitive function, and lower levels of allostatic load (Beckett, Goldman, Weinstein, Lin, and Chuang, 2002; Cornman, Goldman, Weinstein, and Chang, 2003; Glei et al., 2005; Goldman, Glei, and Chang, 2004a; Seeman et al., 2004; Yeager et al., 2006). Contrary to expectation—especially in light of the traditional importance of the extended family system and filial piety in Taiwan—coresidence patterns, the number of children, and contact with these children reveal little, if any, association with these health outcomes. In addition to highlighting the importance of measuring different components of the social environment, the results of these studies demonstrate the need to include a broad range of health outcomes. For example, whereas perceptions about social support matter little for measures of physical well-being in this population, these perceptions appear to be protective against depressive symptoms (Cornman et al., 2003). Not only are adverse aspects of the social environment associated with health decline and mortality, but also a socially rich environment is significantly related to maintaining health (Beckett et al., 2002). Our research has also demonstrated that standard socioeconomic measures, most notably education and income, are related to health. A recent analysis, however, shows that the biomarkers measured in SEBAS do not account for the relationship between socioeconomic status (SES) and the two health outcomes considered—overall self-rated health and mobility difficulties (Dowd and Goldman, 2006). In particular, there is no evidence that sustained activation of neuroendocrine markers, including cortisol, is an important mediator in the relationship between SES and health. These results place an increased burden of proof on researchers who argue that psychosocial stress is an important link in the pathway linking low SES to poor health. In an effort to gain new insights into social disparities in health, the Taiwan survey incorporated a recently developed instrument of subjective social position. This measure asks respondents to use the visual aid of a ladder to position themselves relative to other people in their community and society. An evaluation of this instrument reveals that the ladder captures diverse aspects of respondents’ lives that extend beyond the conventional indicators of education, occupation, income, and wealth (Goldman, Cornman, and Chang, 2006). Cross-sectional analyses of the social determinants of health suggest that the ladder has a stronger asso-
OCR for page 70
Biosocial Surveys ciation with health outcomes than the conventional measures of SES, but we await confirmation of these findings from longitudinal data (Hu, Adler, Goldman, Weinstein, and Seeman, 2005). A question of interest—particularly to those of us doing comparative work—is whether biological profiles are relatively constant across broad populations or whether they reveal large variations, potentially attributable to environmental factors. A detailed comparison among SEBAS, the MacArthur study, and the Wisconsin Longitudinal Survey shows that with regard to risk factors for cardiovascular disease, men are at a clear disadvantage vis-à-vis women in the United States but not in Taiwan. The disparate findings for these two populations lead us to speculate that environmental factors, such as cultural practices or social roles (e.g., higher levels of stratification by sex in Taiwanese society), as well as inherent sex differences, affect these biomarkers (Goldman et al., 2004). In contrast, our analysis of the ApoE gene demonstrates that not all intercountry variation can be attributed to social factors. SEBAS provides the first estimates of the frequency of ApoE alleles for a national sample of the Taiwanese population. The results support earlier evidence based on select samples (Gerdes, Klausen, Sihm, and Foergeman, 1992) that Chinese populations have substantially lower frequencies of the ApoE-4 allele of this gene (a risk factor for Alzheimer disease and ischemic heart disease) than most other national and ethnic groups and underscore the need to obtain better information on the prevalence of dementia in Chinese populations. When we look at the links between the social environment and physiology and between challenge and physiology, we find generally smaller effects than those found in the MacArthur study. Seeman et al. (2004), for example, used the longitudinal data in SEBAS on levels of social integration and extent of social support to predict allostatic load in 2000. Only a small set of significant links emerged from this analysis. The analysis by Goldman et al. (2005) of perceived stress and its relation to physiological dysregulation identified several biomarkers for which high or low values were significantly associated with perceptions of stress. At the same time, however, they found that, in general, the associations between indexes of perceived stress (at baseline and at earlier waves) and physiological dysregulation were small. We are still struggling with a generally unwieldy set of analyses exploring associations between stressful experiences (such as the death of family members, relocation, and financial difficulties) and physiological dysregulation, as well as the potential moderating role of “vulnerability,” defined in terms of social position, social networks, and coping mechanisms. Still, our analyses identified the potential importance of traumatic experiences among the older population: questions that assess losses and reports of distress resulting from the 1999 earthquake show that damage to the home is significantly associated with
OCR for page 71
Biosocial Surveys an increase in depressive symptoms, most notably among middle-aged women (Seplaki, Goldman, and Weinstein, 2006b). The generally weak findings regarding links between the biomarkers and the social environment (including environmental stressors) convinced us of the need to include a broader set of questions related to stressful experiences and perceptions of stress in the second round of the biomarker collection. We have added questions related to traumatic experience, perceptions of stress, caregiving, major life events, and daily hassles. Certainly, in terms of evaluating the framework of allostatic load, these are critical pieces. Our feeling is that much of the earlier work discussing allostatic load—which found associations between SES and allostatic load—assumed that lower SES was associated with greater exposure to challenge (or possibly that lower SES was associated with a poorer response to challenge). It seems like a reasonable assumption, but it remains one that needs to be tested. The data in SEBAS II, along with the work of the National Survey of Midlife Development in the United States (MIDUS) II study, which includes a broad range of psychosocial measures along with similar biomarkers, may provide some insights into these questions. Of course, deficient measures of stressful experience may not be the only reason we are not finding strong associations. It may be that they simply aren’t there, or—as we discuss below—we are not collecting the “right” biomarkers. It may also be unreasonable to expect to find the same relationships across societies. Taiwan, for example, may be a more socially integrated and less stratified society than the United States. Finally, as noted above, we have recently been able to explore links between our measures of physiological dysregulation and downstream health outcomes. Based on mortality data for SEBAS I respondents between 2000 and 2003, Turra et al. (2005) demonstrated that biomarkers are predictive of survival even in the presence of extensive controls for sociodemographic factors and self-reported measures of physical and mental health. These biomarkers include both clinical markers—the cardiovascular and metabolic system measures customarily collected during physical examinations, which have well-defined clinical thresholds for normal function—and nonclinical markers—measures of neuroendocrine and immune dysfunction. The findings of this and an earlier analysis suggest that individuals may underestimate their probability of dying because they have no information about risk factors that act silently on the body and are not detected by clinical exams (Goldman, Glei, and Chang, 2004a). Additional results demonstrate that the nonclinical markers appear to be better predictors of mortality than the clinical markers (although the small number of deaths in the three-year period makes us somewhat wary of pressing
OCR for page 72
Biosocial Surveys this claim), while the clinical markers generally have stronger relationships with (nonfatal) health outcomes. The findings also suggest that the physiological effects of the nonclinical measures, considered as primary mediators hypothesized to affect cardiovascular and metabolic outcomes in the allostatic load framework, are broader than those captured by the clinical markers in this analysis (Goldman et al., 2006a, 2006b). The recent availability of longitudinal information on health outcomes (from the 2003 survey and from registered deaths between 2000 and 2003) has enabled us to compare some of our findings based on analyses of the 2000 data with similar analyses that use health outcome data assessed three years after the collection of the biomarkers and include health controls at baseline. We find some potentially important discrepancies between cross-sectional and longitudinal results. Whereas analyses by Seplaki et al. (2004) based on the 2000 data indicate that the neuroendocrine and immune measures have significant associations with physical and mental function, the more recent analysis using reports of health in 2003 does not find these effects (Goldman et al., 2006a); one caveat is that the two studies used different types of dysregulation scores to capture extreme values of these nonclinical markers (grade-of-membership analysis versus more conventional cumulative dysregulation scores). In a separate set of analyses relating values of DHEAS to a range of health outcomes, estimates from statistical models based on 2000 data suggest stronger associations between high values of DHEAS and better health for women than for men (Glei et al., 2004). In contrast, statistical models that include measures of DHEAS and baseline health in 2000 to predict health outcomes in 2003 generally show that these associations exist for men, but not women. These two sets of comparisons underscore the limitations of cross-sectional analyses—in particular, the potential to overestimate associations between biomarkers and health because of possible reverse effects of poor health status on the biomarkers. The DHEAS results also raise intriguing questions as to why these types of biases may differ by sex. LIMITATIONS Of course, like all studies, our study has limitations. In terms of looking at the effects of stress on health, one might wonder whether an older population is the right target sample. Work by Crimmins and her colleagues on the data from the National Health and Nutrition Examination Survey (Crimmins, Johnston, Hayward, and Seeman, 2003) suggests that allostatic load flattens with age; more generally, some research suggests that SES gradients in health diminish with age (House et al., 1990, 1994). By using a sample age of 54 and older, have we missed the ages at which
OCR for page 73
Biosocial Surveys stressful experience and other social factors may have the greatest effect? We have some reassurance on this score: the probability of survival from ages 20 to 50 in 1970 was 0.90 for men and 0.94 for women (http://www.moi.gov.tw/stat/english/elife/1970.htm). Thus, while it seems reasonable to suppose that the effects of selective mortality are small, we cannot rule it out. What about the biomarkers themselves? Did we collect the right biomarkers? Did we collect enough biomarkers? Are the generally small effects of our measures of stress on the biomarkers a result of having chosen markers that are relatively insensitive? Some biomarkers, cortisol for example, have substantial diurnal variation; our use of the overnight urine for this measure was intended to capture basal levels of production, but perhaps we are missing important information about more transient levels of response or recovery time. At the outset of any project, it is difficult—perhaps impossible—to anticipate all the future uses of the data. Like all research, we began with a set of questions in mind and then discovered that we had additional questions. Even when we started, we knew that—ideally at least—some additional biomarkers would be desirable. The amount of blood we could collect was limited and assays are expensive; collection of cerebral spinal fluid is out of the question for a population-based study. Is there a better way of collecting a broad spectrum of markers? Dried blood spots are now being used successfully in a number of studies; it is a technique that allows collection in the home, and while we would have been unable to perform all the assays we wanted, our understanding is that additional assays based on blood spots are being developed. We are currently exploring the possibility of using tandem mass spectrometry on our stored serum specimens to identify peptides associated with the stress response; plasma would be an alternative compartment. Other approaches suggested by Singer and his colleagues (Singer, Ryff, and Seeman, 2004) include identifying time-related metabolic changes from, for example, urine specimens. Clearly, a major limitation affecting our existing analyses is that we have measured the biomarkers only once; the second round in 2006 will be of great importance, although we suspect that it will raise many new questions and concerns. An important improvement in our second round is that we will also be measuring some markers in the home (blood pressure, lung function, grip strength, timed walks, and chair stands). These additional measurements should help us assess more accurately the magnitude of nonresponse bias affecting estimates derived from the hospital-based biomarker collection in the first round of the survey. In the case of blood pressure, the use of home assessments along with measurements in the hospital is likely to provide us with information about the degree of
OCR for page 74
Biosocial Surveys white coat hypertension (involving people’s response to medical professionals), which we believe affected estimates in the first round. We know, too, that many of our biomarkers are affected by medication use; we have some of the relevant information that would allow us to disentangle these effects, but it is a complicated undertaking, made more complex by the difficulties in obtaining complete lists of medications, documenting the extensive array of traditional Chinese medicines, and identifying the effects of both traditional and Western medicines on the biomarkers. We remain concerned about the best way to measure physiological dysregulation; current scores—our own and those of other researchers—continue to have problems in terms of choice of cut points in the absence of clinical information, as well as potential biological interactions among these biomarkers. The problems with interactions among the biomarkers are exacerbated by limitations of conventional statistical models: they are not well designed to incorporate large amounts of information over successive waves, and we may not have the statistical power to get at some of the processes that interest us. SUMMARY Have we contributed to demographic knowledge? We would say yes. First, our analysis of SEBAS confirms earlier findings that the inclusion of biomarkers in household interview surveys improves the accuracy of the resulting health information (Goldman et al., 2003). Second, we demonstrate that the incorporation of biomarkers into statistical models of mortality substantially enhances the accuracy of the predictions, even in the presence of extensive control variables for self-reports of physical, mental, and cognitive health and sociodemographic information. Third, our exploratory analyses show that biological information can be used to identify biomarkers (and physiological systems) that do and do not account for demographic differentials in health and mortality (e.g., by sex), but that this may raise as many questions as it answers. Fourth, it is too early for us to answer the really big question: Does the inclusion of biomarkers in household surveys help us to understand SES differences in health, particularly with regard to the role of stressful experience? As described above, our research to date indicates that, at least among the older population in Taiwan, reports of stressful experience or perceptions of stress are only modestly correlated with the biological measures included in SEBAS and that these measures account for little of the association between SES and health outcomes. Overall we think that, despite the limitations and occasional mistakes, and despite the logistical and financial constraints, the Taiwan study has
OCR for page 75
Biosocial Surveys been remarkably successful in achieving many of its objectives. Would we argue that all social surveys should collect biomarkers? Certainly not. Our experience has shown us that it is a complex task, not to be undertaken by the faint of heart. Because no single study can be all things to all people, we would suggest that considerable care be given to the choice of biomarkers and especially to archiving specimens for future use when the inevitable new questions arise and better technologies become available for the analysis of those specimens. ACKNOWLEDGMENTS The work on this project has been financed by the Office of Behavioral and Social Science Research of the National Institute on Aging under grant numbers R01AG16661 and R01AG16790. We gratefully acknowledge their support. The project has required a team with a wide range of skills across multiple disciplines. We particularly wish to acknowledge the contributions of Teresa Seeman, who has been an important collaborator throughout the project. Her advice and cooperation regarding criteria for our choice of laboratory and issues regarding choice of assay were invaluable, particularly during the early stages of the project. She generously provided the MacArthur protocols for blood and urine collection; these protocols served as the basis for the ones we used in Taiwan. Our technical decisions have also benefited from the accumulated wisdom of Chris Coe and Paul Aisen, who provided guidance on assays. Chris Peterson helped us through the difficulties of the public release of the first round of data. I-Fen Lin provided help with questionnaire design and has been indispensable for translation issues. Burt Singer and Germán Rodríguez have provided outstanding statistical guidance. Jennifer Cornman, Christopher Seplaki, and Cassio Turra have been wonderful collaborators at different stages of the project. Kaare Christensen has provided significant assistance with the protocols for the home-based assessments of function in the second round of the study. We extend our thanks to all of them. REFERENCES Beckett, M., Goldman N., Weinstein, M., Lin, I.-F., and Chuang, Y.-L. (2002). Social environment, life challenge, and health among the elderly in Taiwan. Social Science and Medicine, 55(2), 191-209. Chang, M.-C., and Hermalin, A. (1989). The 1989 survey of health and living status of the elderly in Taiwan: Questionnaire and survey design. Comparative Study of the Elderly in Four Asian Countries. Research report 1. Ann Arbor: Population Studies Center, University of Michigan.
OCR for page 76
Biosocial Surveys Cohen, J.I. (2000). Stress and mental health: A biobehavioral perspective. Issues in Mental Health Nursing, 21, 185-202. Cornman, J.C., Goldman, N., Weinstein, M., and Chang, M.-C. (2003). Social ties and perceived support: Two dimensions of social relationships and health among the elderly in Taiwan. Journal of Aging and Health, 15, 616-644. Crimmins, E.M., Johnston, M., Hayward, M., and Seeman, T. (2003). Age differences in allostatic load: An index of physiological dysregulation. Experimental Gerontology, 38(7), 731-734. Dowd, J.B., and Goldman, N. (2006). Do biomarkers of stress mediate the relationship between socioeconomic status and health? Journal of Epidemiology and Community Health, 60, 633-639. Gerdes, L.U., Klausen, I.C., Sihm, I., and Færgeman, O. (1992). Apolipoprotein E polymorphism in a Danish population compared to findings in 45 other study populations around the world. Genetic Epidemiology, 9, 155-167. Glei, D.A., Chang, M.-C., Chuang, Y.-L., Lin, Y.-H., Lin, S.-H., Liu, I.W., Lin, H.-S., Goldman, N., and Weinstein, M. (2006, May). Results from the social environment and biomarkers of aging study (SEBAS) 2000: Survey Report (Chinese/English). Taiwan Aging Study Series, 9. Glei, D., Goldman, N., Weinstein, M., and Liu, I.-W. (2004). Dehydroepiandrosterone sulfate (DHEAS) and health: Does the relationship differ by sex? Experimental Gerontology, 39, 321-331. Glei, D., Landau, D.A., Goldman, N., Chuang, Y.-L., Rodríguez, G., and Weinstein, M. (2005). Participating in social activities helps preserve cognitive function: An analysis of a longitudinal, population-based study of the elderly. International Journal of Epidemiology, 34, 864-871. Goldman, N., Cornman, J., and Chang, M.-C. (2006). Measuring subjective social status: A case study of older Taiwanese. Journal of Cross-Cultural Gerontology, 21, 71-89. Goldman, N., Glei, D., and Chang, M.-C. (2004). The role of clinical risk factors in understanding self-rated health. Annals of Epidemiology, 14, 49-57. Goldman, N., Glei, D., Seplaki, C., Liu, I.-W., and Weinstein, M. (2005). Perceived stress and physiological dysregulation. Stress, 8, 95-105. Goldman, N., Lin, I.-F., Weinstein, M., and Lin, Y.-H. (2003). Evaluating the quality of self-reports of hypertension and diabetes. Journal of Clinical Epidemiology, 56(2), 148-154. Goldman, N., Turra, C.M., Glei, D.A., Lin, Y.-H., and Weinstein, M. (2006a). Physiological dysregulation and changes in health in an older population. Experimental Gerontology, 41, 862-870. Goldman, N., Turra, C., Glei, D., Seplaki, C., Lin, Y.-H., and Weinstein, M. (2006b). Predicting mortality from clinical and non-clinical biomarkers. Journal of Gerontology: Medical Sciences, 61(10), 1070-1074. Goldman, N., Weinstein, M., Cornman, J., Singer, B., Seeman, T., and Chang, M.-C. (2004). Sex differentials in biological risk factors for chronic disease: Estimates from population-based surveys. Journal of Women’s Health, 13, 393-403. House, J.S., Kessler, R.C., Herzog, A.R., Mero, R.P., Kinney, A.M., and Breslow, M.J. (1990). Age, socioeconomic status, and health. The Milbank Quarterly, 68(3), 383-311. House, J.S., Lepkowski, J.M., Kinney, A.M., Mero, R.P., Kessler, R.C., and Herzog, A.R. (1994). The social stratification of aging and health. Journal of Health and Social Behavior, 35(3), 213-234. Hu, P., Adler, N., Goldman, N., Weinstein, M., and Seeman, T. (2005). Relations between subjective social status and measures of health in older Taiwanese persons. Journal of the American Geriatrics Society, 53, 483-488.
OCR for page 77
Biosocial Surveys Karlamangla, A.S., Singer, B.H., McEwen, B.S., Rowe, J.W., and Seeman, T.E. (2002). Allostatic load as a predictor of functional decline. Journal of Clinical Epidemiology, 55(7), 696-710. McEwen, B.S. (2002). Sex, stress, and the hippocampus: Allostasis, allostatic load, and the aging process. Neurobiology of Aging, 23, 5, 921-939. McEwen, B.S., and Stellar, E. (1993). Stress and the individual: Mechanisms leading to disease. Archives of Internal Medicine, 153, 2093-2101. Seeman, T.E., Glei, D., Goldman, N., Weinstein, M., Singer, B., and Lin, Y.-H. (2004). Social relationships and allostatic load in Taiwanese elderly and near elderly. Social Science and Medicine, 59, 2245-2257. Seeman, T.E., McEwen, B.S., Rowe, J.W., and Singer, B.H. (2001). Allostatic load as a marker of cumulative biological risk: MacArthur studies of successful aging. Proceedings of the National Academy of Sciences, USA, 98(8), 4770-4775. Seplaki, C., Goldman, N., Glei, D., and Weinstein, M. (2005). A comparative analysis of measurement approaches for physiological dysregulation in an older population. Experimental Gerontology, 40, 438-449. Seplaki, C., Goldman, N., Weinstein, M., and Lin, Y.-H. (2004). How are biomarkers related to physical and mental well-being? The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 59, 201-217. Seplaki, C., Goldman, N., Weinstein, M., and Lin, Y.-H. (2006a). Measurement of cumulative physiological dysregulation in an older population. Demography, 43, 165-183. Seplaki, C., Goldman, N., Weinstein, M., and Lin, Y.-H.. (2006b). Before and after the 1999 Chi Chi earthquake: Traumatic events and depressive symptoms in an older population. Social Science and Medicine, 62, 3121-3132. Singer, B., Ryff, C.D., and Seeman, T. (2004). Operationalizing allostatic load. In: J. Schulkin (Ed.), Allostatis, homeostasis, and the costs of physiological adaptation (pp. 113-149). Cambridge, England: Cambridge University Press. Turra, C.M., Goldman, N., Seplaki, C.L., Weinstein, M., Glei, D.A., and Lin, Y.-H. (2005). Determinants of mortality at older ages: The role of biological markers of chronic disease. Population and Development Review, 31, 677-701. Wachter, K.W. (2001). Biosocial opportunities for surveys. In National Research Council, Cells and surveys: Should biological measures be included in social science research? (pp. 329-338). Committee on Population, C.E. Finch, J.W. Vaupel, and K. Kinsella, Eds. Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. Weinstein, M., Goldman, N., Hedley, A., Lin, Y.-H., and Seeman, T. (2003). Social linkages to biological markers of health among the elderly. Journal of Biosocial Science, 35, 433-453. Weinstein, M., and Willis, R. (2001). Stretching social surveys to include bioindicators: Possibilities for the Health and Retirement Study, experience from the Taiwan Study of the Elderly. In National Research Council, Cells and surveys: Should biological measures be included in social science research? (pp. 250-276). Committee on Population, C.E. Finch, J.W. Vaupel, and K. Kinsella, Eds. Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. Yeager, D.M., Glei, D.A., Au, M., Lin, H.-S., Sloan, R.P., and Weinstein, M. (2006). Religious involvement and health outcomes among older persons in Taiwan. Social Science and Medicine, 63, 2228-2241.