NIH considers the Observational Study (OS) an important component of the Women's Health Initiative (WHI), but it has not given it nearly the amount of attention given the Clinical Trial (CT). The OS will draw from women who have participated in at least one CT screening visit but who are either not eligible or not willing to participate in the CT, and who agree to participate in the OS. It is anticipated that 45 clinical centers will enroll approximately 100,000 women in the OS, following them for an average of nine years. Data collected at baseline are to be related to subsequent clinical events in order to examine the associations of both putative risk factors and protective factors with subsequent disease occurrence. An observational study of this scope would require a very large investment if it were initiated independently. Designed as a WHI component, however, it can take advantage of the baseline information that will have been collected in the course of the screening to form the CT cohort. The marginal cost of the OS—that amount that would be required in addition to what is already required by CT protocol—is therefore relatively reasonable for the information to be acquired.
The OS thus has the potential to complement the CT by providing another (larger) cohort, although one in which hypothesized preventive measures are not assigned at random. Not only could investigators use this cohort to pursue hypotheses related to those of the CT, they would have the opportunity to describe myriad relationships among putative risk factors, participant characteristics, physiologic changes, and morbidity and mortality. Diseases such as Alzheimer's disease, osteoarthritis, and stroke are common, yet much remains to be learned about risk factors. The sequence of events, behaviors, and diagnoses is more readily apparent in a prospective cohort study such as the OS than in case-control studies, diminishing at least some of the difficulties in interpreting results. This longitudinal study presents a good opportunity to be able to better differentiate among antecedents, intervening variables, and effects, thus providing information on predictors of these diseases and others of public health importance.
The committee was provided with a proposed list of variables to be measured at baseline and at the three year follow-up using questionnaires. The OS Subcommittee–made up of selected NIH and Vanguard Clinical Center investigators—suggested many variables,
most of which were then included in the CT questionnaires. * The variables remaining in the supplemental questionnaire, which will be given only to OS participants, reflect more speculative issues than those addressed by the CT. However, although the WHI Manual of Operations and Procedures notes that “(t)he content of the supplemental questionnaire is directed to the epidemiology of primary outcomes that are relatively common, serious, and capable of being readily ascertained by means of the planned OS follow-up mechanisms” (1993), the committee was provided with little information about the possible specific etiologic associations and the additional hypotheses being studied in the OS.
DESIGN AND METHODS
The OS is designed as an observational prospective cohort study. Women who have been contacted and expressed interest in the CT, but who are ineligible or unwilling to participate, will be asked to participate in the OS. Being drawn from the same area and age group, the women in the OS and CT will be in some ways comparable and, as volunteers, are likely to be different from the population as a whole. The women in the OS are likely to differ from CT participants in other respects, including their health behavior, history, and status. The incidence rates of disease and the extent of exposure to risk factors may be different among the CT, OS, and the general populations.
Despite, the OS participants having been drawn from a “convenience” sample, not a population-based sample, the OS population will be useful for examining associations between possible risk factors and disease; there is no a priori reason to believe that associations between possible risk factors and diseases will be different in this subgroup.
In addition to the information obtained by questionnaires, a fasting blood draw (and in some cases a urine collection) will provide specimens to be frozen and stored. The May 1991 Repository Feasibility Study from the National Cancer Institute estimates that between 10 and 20 million samples will be stored for the entire CT and OS populations. Serum storage is potentially valuable because it is likely that over the next several years hypotheses concerned with biological markers of exposures or of early disease will be formulated outside of the WHI that can be examined in this study population. A 1 percent subsample stratified on age, racial/ethnic group, and socioeconomic status will have repeat baseline information after the first screening visit so that the reliability of the data can be determined and error correction methods applied. This reliability testing strengthens the study.
Questionnaire categories include: Personal Information, Medical History, Reproductive History, Family History, Personal Habits, Psychosocial, Food Frequency, and OS Supplement.
All 100,000 OS participants will receive follow-up materials by mail at year one, and will be scheduled to return to the clinic for a follow-up visit at year three. Between 5 and 6 percent of the OS group will return to the clinic every three years for physical examination and laboratory measures; all 100,000 OS participants will be contacted annually by mail for follow-up. Bone densitometry studies and urine samples will be collected from the OS participants at the three osteoporosis substudy centers every three years. The Clinical Coordinating Center estimates the cost of the first year follow-up by mail and the year three follow-up visit to be $8.5 million.
The OS will have more power than the CT for testing associations in minority populations, since the OS anticipates enrolling 20,000 minority women. Investigators will attempt to recruit minority women to reflect their proportion in the general population of women aged 50-79 years in each clinical center region (approximately 17 percent of the population in the United States is nonwhite). Overall, anticipated OS enrollment by racial/ethnic subgroup is: African American: 10,000; Hispanic: 6,000; Native American: 2,000; Asian and Pacific Islanders: 2,000; and white: 80,000.
The committee feels that the elements of the analytic plan, as described by NIH, seem appropriate. Regression methods will be used to estimate relative risks. Nested case-control sampling can be used to reduce the number of blood specimens that will need to be analyzed for hypotheses concerned with serum. In instances in which measurements are stable over time, case-cohort sampling can be used.
Including the OS as a WHI component is inexpensive relative to a free-standing observational trial of this size, because the infrastructure to support the study would already be in place. For example, the same number of women need to be contacted initially for the CT regardless of whether they are retained for the OS. NIH is unable to distinguish the overall costs of the OS from the costs of the CT because both components were generated simultaneously. The only available estimate of OS costs comes from the Clinical Coordinating Center. It estimates the marginal costs of the OS (i.e., those costs over and above the cost of the CT, which includes the vast identification, recruitment, and screening process) to be approximately $10,570,000, that is, $105 per person over the planned 12-year duration of the OS recruitment and follow-up period. If accurate, that is a good cost per person; if it is a low estimate, the actual cost could still be reasonable.
Although data collection is included in the WHI cost estimate, many costs of analysis and publication are not considered. When additional funds are sought for analysis, both the cost and the scientific yield of the OS will increase.
The committee agrees with NIH that the OS is potentially a rich source of data within the WHI at a reasonable cost. However, the committee is concerned that it was not provided with historical material explaining the additional hypotheses being tested in the OS (beyond the DM, HRT, and CaD supplementation hypotheses). It is unclear whether the lack of such material reflects the prudent concentration of NIH staff on the CT when collecting material for the committee or a true lack of recent attention to the additional hypotheses being tested.
The committee recommends that NIH treat the OS as a precious investment even though its cost relative to the entire WHI is low. The progress of the OS should be monitored carefully to ensure that the quality of the data is as high as possible, losses to follow-up are minimized, and carefully framed investigations are pursued.
The committee also strongly recommends that NIH make the OS data available to qualified investigators outside the WHI network early enough to maximize the use of the data. Outside investigators should be able to add ancillary studies to the OS. Such data availability will increase the scientific and public health yield of this public investment, increasing the justification of the expenditure. Clear policies about the timing and conditions under which the data and cohort will be made available to outside investigators need to be established from the outset.
The committee's recommendation for the CT Objective Prescheduled Reassessment, outlined in Chapter 2, has implications for the OS. For example, if the CT should end after eight years, the coordination and funding of the OS would be significantly affected. The committee encourages NIH and the WHI OS Subcommittee to consider these possibilities and devise contingency plans for the OS.