A. INTRODUCTION AND PURPOSE

With the aging of the population, osteoporosis is rapidly becoming one of the major health challenges for older people living in the United States. {27,51,104} Osteoporosis refers to a decreased density of bone mass blow the level that is needed to maintain skeletal function and provide adequate mechanical support. {51,71,76} Although risks and rates of osteoporosis may vary across groups, some bone loss appears to be part of the natural aging process in that bone mass declines for everyone after middle adulthood, and briefly escalates for women during the postmenopausal (50 to 59 years} period when estrogen levels drop. {27} The seriousness of the emerging osteoporosis problem is reflected in recent estimates which indicate that the proportion of the population made up of the elderly and postmenopausal women is expected to increase in 50 percent in the next two decades. {91, 104}

In addition to aging and estrogen deficiencies, a third cause of osteoporosis, which has received less attention, is disuse and disease. {77,89} This form of osteoporosis results from the cumulative breakdown of bone mass due to inactivity, misuse or immobility associated with paralysis or muscle weakness. {89} A significant increase in disability-related osteoporosis is also expected due to the marked increase in survivorship among people aging with early-life disabilities such as polio, cerebral palsy, and spinal cord injury. {54,89} Despite this expected rise in osteoporosis rates among persons aging with disability, empirical investigations in this population have been scant. What studies do exist of disability-related osteoporosis have focused primarily on spinal cord injury within the first few years after onset. {32,33} Particularly noteworthy is the absence of data on the long-term effects of disuse and misuse for the estimated 1.6 million polio survivors living in the United States today. {89}

The purpose of this investigation is to respond to this gap in our knowledge by examining the effects of disability, aging, and menopause-related estrogen deficiency on bone mineral loss for a sample of female polio survivors and age-matched non-disabled controls. The primary objectives of the study are:

1. To compare percentages of bone loss in several areas of the body for female polio survivors and age-matched controls to determine whether living an average of 48 years with impaired mobilization results in significantly less bone mineral mass.

2. To discover whether the cumulative effects of aging on percentage of bone loss differ between females from disabled and non-disabled samples.

3. To discover whether the accelerated bone loss of the postmenopausal period can be mitigated by estrogen replacement therapy (ERT}, and whether the effects of this intervention are different for women in these two groups.

Results from this last objective may have important clinical implications. If female polio survivors with long-term use of ERT have significantly lower levels of bone loss compared to their counterparts with irregular or no ERT, then, initiating treatment early in menopause may constitute a promising intervention strategy for reducing the risks of falls and excess disability which further complicate the later-life effects of polio.

B. BACKGROUND AND HYPOTHESES

Aging and Bone Loss. The production of bone mass normally increases until approximately age 30. {27,44,73,76} following a short period of stabilization, age-related bone loss begins as changes occur in the remodeling cycle of special bone cells called osteoblasts and osteoclasts. {76} The normal remodeling cycle begins with the osteoclast cells removing or absorbing bone. The cycle is completed when this cavity, or bone loss, is gradually refilled through the work of osteoblast cells. {76} Bone loss occurs either when an increased number of osteoclasts create excessive cavities, or when a drop occurs in the production of osteoblasts leading to an incomplete refilling of the cavities. {76} Investigators explain the strong empirical association between osteoporosis and age as a function of a slow, but steady, decrease in osteoblastic activity that beings at approximately age 30 for trabecular bone [hard area of bone located on the periphery or outside of the bone surface.] and age 40 for cortical bone. [hard area of bone located on the periphery or outside of the bone surface.] {63,75,76,77}

Postmenopausal Bone Loss and ERT. The acceleration of bone loss that occurs during the immediate postmenopausal period has been explained as the result of an increase in osteoclast cells, which has more potential for structural damage than does a decrease in age-related osteoblastic activity. {34,63,75} Heightened osteoclastic activity leads to deeper cavities during the bone remodeling process and the potential for a complete perforation of the trabeculae in cancellous bone, which results in the elimination of the bone formation phase of the remodeling cycle. In contrast, the slow decline in osteoblastic activity associated with long-term aging leads only to a thinning of the trabeculae. {76}

How much acceleration occurs in the rate of bone loss during the post-menopausal period? Most researchers agree that cortical bone loss increases from less than 0.5 percent per year before menopause to 2-3 percent annually for a period 8 to 10 years following menopause. The rate of loss then returns to baseline levels following this postmenopausal period until later in life when loss of cortical bone again increase due to age-related processes. {28,34,53,75,89} For trabecular bone, estimates for an accelerated postmenopausal bone loss vary widely. Different cross-sectional studies have placed this estimate for women to be anywhere from a decrease of 0.6 per cent per year, {77} which indicates no acceleration, to a decrease of 4-5 per cent per year for six years. {63} Recently, Riggs and Melton (1992) made a "tentative" conclusion that trabecular bone in the postmenopausal period undergoes a large but short-term acceleration of loss. {76} They also explained the wide range in estimates of this acceleration phase due to the type of termination of menopause and the age at which it occurs. Natural termination of menopause, which typically occurs around age 50 to 52, results in less severe accelerated trabecular bone loss than oophorectomy (or hysterectomy) "because the onset of sex-steroid deficiency is more variable and more gradual" (p. 1677) {75} In contrast, women undergoing hysterectomies are typically much younger than women who experience menopause naturally, and consequently have higher rates of bone regeneration, which ironically leads to more bone loss. {77}

The postmenopausal period is estimated to result in anywhere from a third to half of the bone loss a woman experiences in her entire lifetime, which has been placed at an average of 50% of trabecular bone and 30% of cortical bone. {76} This contrasts with average lifetime estimates of bone loss for men of only 30% and 20% respectively. {76}

A variety of treatments have been proposed for the prevention of osteoporosis, with one of the most effective being Estrogen Replacement Therapy (ERT). {28,45,50,73,76,95} ERT involves using a group of antiresorptive drugs to prevent bone loss by decreasing the rate of bone resorption during the bone-remodeling phase. {7} In one study, long term regular use of ERT was found to increase a woman's trabecular bone mass by 50 percent and decrease her chances of osteoporosis-related fractures by 50 percent compared to non-medicated controls. {28} Although the acceleration period of postmenopausal bone loss is estimated to last only five years for trabecular bone and 10 years for cortical bone, continued use of ERT throughout the rest of a woman's lifetime appears to be necessary in order to effectively maintain the lower rates of bone loss. {104} (For a review of the research on the effects of ERT see *Riggs & Melton, 1992). {76} Empirical data focusing on how women who already are experiencing bone loss related to a disability traverse the vulnerability of the postmenopausal period have yet to be published. {89}

* Riggs LR and Melton LJ. The Prevention and Treatment of Osteoporosis. New England Journal of Medicine.1992; 327(9): 620-627. No abstract available. "A concise review of the modalities of osteoporosis treatment, distinguishing antiresorptive from bone-forming regimens. Also makes specific recommendations concerning preventative measures."

Chronic Disability and Bone Loss. Disuse and disease osteoporosis describe the cumulative breakdown of bone mass due to limited mobility and inactivity.{89} To our knowledge, no data have been published to date on the risks of osteoporosis among polio survivors. The closest related study is one by Werner, Warning and Maynard,{99} who detected a possible link between chronic polio and osteoarthritis (OA). They found that 68% of their post-polio sample [cross-sectional study of 61 post-poliomyelitis survivors] suffered from mild, moderate, or severe OA in the hand and wrist, which compares to an expected rate of 30% in previous studies of nondisabled individuals. {99} They also found that OA in the hand and wrist was present in a significantly higher percentage of the sample with severe locomotor limitations than it was for those with mild and moderate locomotion limitations. Maynard and his colleagues interpret their findings as supporting on "overuse" hypotheses. {99} In a recent study of stroke survivors, rather than polio survivors, the authors found that the bone mineral densities of the non-paralyzed side were significantly higher than those of the paralyzed side. {43} Several studies of individuals with spinal cord injury have also found increased bone density loss and increased risk of osteoporosis in paralyzed limbs. {33} A major limitation of the SCI literature on osteoporosis and disability is that most of the studies have been restricted to males and to an outcome period of less than 10 years.

Hypotheses. The following three hypotheses guided this investigation. The first focuses on the main effects of age and sample membership, while the second and third stipulate significant interactional effects.

H1: Due to the long-term effects of disuse and misuse, female polio survivors will have significantly higher average levels of bone mineral loss for Greater Trochanter and Ward's Triangle areas [See Annotated X-RAY Graphic at Left] of the hip compared to age-matched non-disabled controls.

H2: Age of participants will be positively related to percentage bone loss for both samples, however, the magnitude of the relationship will be greater for polio survivors than it is for controls.

This hypothesis stipulates a two-way interaction between age and sample membership. Specifically, mean differences in percentage bone loss between polio and control females are expected to increase as chronological age increases. Within the context of cross-sectional data, this means that as age of participants increases from 50 to 80 and over, bone loss will increase faster for polio survivors than it will for controls. We refer to this prediction as the "accelerated aging" effect because it suggests that female polio survivors are "aging" more quickly than controls in terms of bone mineral loss.

The third hypothesis stipulates both a main effect of estrogen replacement therapy and a three-way interaction of disability, age and ERT.

H3: Regardless of sample membership, women who received continuous ERT since the beginning of menopause will have significantly less bone loss compared to those who received either irregular or no ERT, although the beneficial effects of ERT are expected to vary by area of the body due to differences in the percentage of trabecular bone, which is more sensitive to estrogen levels. In addition to significant main effects, the effects of ERT are also expected to differ by the interaction of sample membership and chronological age. Lack of ERT is expected t pose the greatest risk of bone loss for those women who are experiencing the combined effects of postmenopausal estrogen deficiencies, long-term disability, and "aging;" that is, female polio survivors with no history of ERT who are 60 years of age or older at T.O.M. [time-of-measurement]

C. METHODS

Sample. The sample utilized in this investigation is different from that described above for the entire LLE [Later Life Effects] Study (see page 17 & 18 of main study). The osteoporosis sample consisted of a subset of 43 female polio survivors selected from the LLE Study and an augmented sample of 42 female controls selected from both the non-disabled control subsample of the LLE Study (N= 23) and patient files from the Bone Diagnostic and Treatment Centre in Long Beach, CA (N= 19). We were unable to include all of the female participants from the polio (N= 79) and control subsamples (N = 35) because of both technical and logistic problems which prevented some participants from getting access to the bone densitometer equipment necessary to perform the bone scans. The additional non-disabled subjects were selected at random from the Bone Diagnostic Centre based on the absence of any major risk factors for osteoporosis (e.g., falls, extensive periods of disuse, etc) and a one-to-one age match with polio subjects. Each augmented control subject was matched to a polio survivors within one chronological year of birth.

Procedures. A complete description of the equipment and methodology used to perform bone density scans and derive measure of total 'bone mineral density' (BMD) for different areas of the body is presented above in the general "Methods" section of the final report under Procedures' (see pages 19, 20 & 31 of the full report [off page link]). Total BMD is defined as the total mass in grams per squared centimeter (cm) of area for different regions of the spine and hip. To obtain measures of the risk of osteoporosis each BMD estimate was converted into a percentage of bone mineral loss by comparing the total BMD score for each region of the hip and spine to normative data for "healthy" young females of the same ethnic group. {35,45,62}

Measures. The percentage of bone mineral loss in the Greater Trochanter and Ward's Triangle areas of the hip were selected as dependent measures for two reasons. First, measures of bone loss in the lumbar spine proved to be unreliable due to the high incidence of an osteoarthritic artifact which inflates estimates of bone density. Second, using both the Greater Trochanter and Ward's Triangle allowed for a comparison of mineral loss in two different types of bone. The Greater Trochanter area of the hip is composed of 60 percent cortical bone and 40 percent trabecular bone, while the proportions of bone are reversed for Ward's Triangle area of the hip (i.e., 40% cortical and 60% trabecular). Previous research by Riggs and Melton indicates that mineral loss in trabecular bone appears to be more sensitive to intervention with ERT than is the case for cortical bone. {76,77}

The independent variables used as predictors of bone loss were obtained by self-report and included: Chronological Age at time-of-measurement (T.O.M.); Age at Time of Last Menstrual Period; ERT Status since time of last period, measured on an ordinal scale of 'No ERT', 'Non-continuous ERT', and 'Continuous ERT'; and type of Termination of Menses, measured as 'Natural' versus 'Surgical'.

Analytic Strategies. The data for this investigation were analyzed using the same two-level strategy described in the main part of the report under Methods (see pages 32 and 33 of the full report)

D. RESULTS

1. Descriptive Results

   Sample Characteristics. The data in Table 22 show the success of our matching procedure. With mean ages of 59.6 and 61.2, respectively, there were no significant differences in chronological age at time-of-measurement between polio and control subjects.

   FOCUSED INVESTIGATION -- RISKS OF OSTEOPOROSIS

   Table 22. Polio vs. Control Sample characteristics for comparative analysis of risks of osteoporosis. Group N Mean Age STD LLE Polio Sub-Sample of Females 43 59.6 6.98 Non-Disabled Female Controls: 42 61.2 6.46 LLS Sub-Sample 17 62.2 6.98 Bone Diagnostic Centre 25 60.8 6.14 Table 23. Polio vs. control Menopause characteristics for osteoporosis analysis.   POLIO NON-DISABLED     Variable N Mean/% STD N Mean/% STD Termination of Menses *     Natural vs. 18 42.9% ---- 26 63.4% ----     Surgical 24 57.1% ---- 15 36.6% ---- Age at Last Period     Natural vs. 16 50.5** 7.97 25 50.4** 4.04     Surgical 22 41.3 5.87 14 41.2 6.35 Time on ERT **     Natural vs. 16 3.4** 4.34 21 2.2 3.84     Surgical 17 9.1 9.34 14 4.8 6.85 * p = £ .05         ** p = < -.01

   In addition, as Table 23 reveals, there were also no between sample differences in 'Age of Last Period'. On average, female polio survivors experienced their last menstrual period at 45.2 years of age compared to 47.1 for female controls.

   There were, however, significant between sample differences in both termination of menses and the length of 'Time on ERT'. Somewhat surprisingly, the percentage of women with surgically versus naturally terminated menses was greater for polio survivors (57.1% vs. 42.9% than it was for non-disabled controls (36.6% surgical vs. 63.4% natural). Equally surprising and counter to expectations, polio survivors reported an average length of time on ERT that was more than twice as long as controls, with means of 6.3 years vs. 3.1 years, respectively (p= £ .01). Consistent with the mean differences in duration of ERT, the data in figure 32 indicate that female controls were much more likely to have no history of ERT than were polio survivors, with 51% versus 24% reporting 'None' on a self-administered questionnaire (see figures 32 & 33). Conversely, the percentage of women with a 'Continuous' history of ERT use was significantly larger for polio survivors than it was for non-disabled controls (i.e., 33% vs. 15%, respectively).

   Although there were no between-sample differences in the timing of onset of menopause, differences in termination of menses did account for significant within-sample differences in both age of last period and time on ERT (see Table 23). Not surprisingly, women in both samples whose menses were terminated surgically experienced the onset of menopause at a significantly younger age compared to women whose menses terminated naturally. Interestingly, average age at last period by type of termination of menses was almost identical for both samples, with means of 50.5 and 50.4 for polio and controls with naturally terminated menses compared to 41.3 and 41.2 for polio can controls with surgical termination. Termination of menses also accounted for significant differences within the polio sample in length of time on ERT. Female polio survivors with surgical termination of menses had been on estrogen replacement therapy almost three times longer than their counterparts whose menses were terminated naturally (9.1 vs. 3.4 years of ERT use). In contrast for controls, differences in termination of menses had no significant effect on the duration of ERT. Together, these findings suggest that polio survivors may have been more likely to be on 'estrogen replacement therapy', in large part, because they were more likely to have their menses terminated surgically.

   Summary of Descriptive Results. Results of descriptive analysis raise two important and interrelated questions. First, why should the incidence of hysterectomies, or surgically terminated menses, be greater for polio survivors than it is for non-disabled controls; and second, why should differences in termination of menses result in longer duration of ERT for polio survivors only? Although an empirical response to these questions goes beyond the scope of available data, three possibilities are offered to stimulate further research.

   One possible explanation for these between sample differences suggests that female polio survivors have a higher incidence of hysterectomies because they received more frequent medical attention than non-disabled controls and, therefore, the probability of detecting gynecology problems was greater for this group of women than it was for controls. A second and related explanation posits that because of the long-term effects of disability female polio survivors have a higher incidence of gynecological problems compared to non-disabled controls and, therefore, a higher incidence of surgically terminated menses. A third possible explanation for differences in frequency of hysterectomies points to the effects of discrimination or "disabilityism" within the medical profession, which undermines the quality of health care disabled women receive. It may be that female polio survivors, like women with spinal cord injury, are less likely to receive regular gynecological exams and therefore more at risk for serious problems which could be treated non-surgically if detected earlier. {90,100}

   Regardless of the underlying cause of differences in termination of menses, the higher rate of ERT and the longer average time on ERT reported by the polio sample raise two additional questions. First, did our sample of female polio survivors actually experience less of the accelerated trabecular bone loss associated with estrogen deficiency in the immediate post-menopausal period compared to non-disabled controls due to higher rates of both surgical termination and continuous ERT usage? And, second, can greater use of ERT in the polio group compensate for predicted sample differences in both trabecular and cortical bone that are due to the long-term effects of disuse and misuse as opposed to the short term effects of estrogen deficiency? Previous research suggests that the answer to the second question will be negative. To date, ERT has not been found to be effective treatment for preventing cortical bone loss, which is more sensitive to the long-term effects of aging and disability than it is to estrogen levels. {76}

2. Predictive Results

   Main Effects of Age and Sample. Figure 30 provides support for H1 has it applies to the main effects of age and sample membership on bone loss in the Greater Trochanter area of the hip. Adjusting for differences in the dependent variable due to age, female polio survivors had significantly higher percentage of cortical bone loss in the Greater Trochanter area of the hip compared to age-matched, non-disabled controls (22.7% vs. 14.8%, p= .05). Conversely, controlling for the effects of sample membership, increases in the chronological age of participants were positively related to significant increases in the percentage of bone mineral loss. Women in the youngest age group at T.O.M., those 50 to 59 years of age, had significantly less bone loss in the Greater Trochanter area compared to their older counterparts, age 70 and above (15.4% vs. 29.2%, p=£ .05) However, when the dependent variable changes to the Ward's Triangle area of the hip, which is composed primarily of trabecular bone, we observe no differences in bone loss between polio survivors and controls (no figure represented). This absence of between-sample differences in bone loss for Ward's triangle may have to do with the previous finding indicating that polio survivors were on ERT significantly longer than controls, and, therefore had less exposure to the effects of postmenopausal estrogen deficiency on trabecular bone.
   

   Figure 30. Polio vs. Control: Adjusted main effects of age and sample on percentage bone loss for Greater Trochanter area of hip.

   Accelerated Aging Effect. Figure 31 provides partial support for the accelerated aging effect contained in H2. While the overall interaction between sample membership and chronological age is significant (p= £ .01), mean differences in percentage bone loss between samples only achieve significance for those in the 60 to 64 age category. Female polio survivors in this age group have twice the amount of Trochanter bone loss as controls (28.7% vs. 13.2%, p = £ .001), and it approximates the estimated lifetime average of cortical bone loss for all women of 30%. As differences in subgroup elevations indicate, the general trend is for bone loss to increase faster with an increase in chronological age of polio survivors than it does for controls, although this "accelerated aging" effect occurs only for those between the ages of 50 and 64. After this age, the gap between polio survivors and controls diminishes until it actually reverses for those in the oldest age group (70 to 88 years), with non-significant percentage differences of 27.8& bone loss for polio versus 30.8% for controls.
   

   Figure 31. Polio vs. Control: Differential effects of age on percentage bone loss in hip (Greater Trochanter) by sample (p£ .01).

   A similar but non-significant accelerated aging trend was also observed for the Ward's Triangle area of the hip (no figure presented). Again, bone loss increased faster with age for the polio survivors than it did for controls, but the between-sample differences observed for the 60 to 64 year old cohort only approached significance. Polio survivors in this age group had an average (Trabecular) bone loss of 46% compared to 34% for controls (p = £ .10). While not statistically significant, this result suggests that by age 60 to 64 female polio survivors had already experienced a level of bone loss in the trabecular bone-dominated area of the hip (Ward's Triangle) that is almost equivalent to the estimated lifetime average for all women of 50%.

   One explanation for the reversal in the "accelerated aging" effect observed for the Greater Trochanter area can be traced to previous findings reported above in conjunction with an application of the life course perspective to polio data (see part 2 of the general Results section, pages 51-64). From these analyses and those of other researchers, {40,42} we learned that both age and historical period of acute onset were inversely related to severity of initial impairment, and that severity of initial impairment was inversely related to chronological age at T.O.M. This means that participants who were older at T.O.M., on average, experienced less severe effects of polio at onset and during their lifetime compared to their younger counterparts. These differences in initial and residual impairment may have resulted in less disability-related bone loss due to disease and disuse compared to younger survivors who, on average, contracted polio at an older age and were more severely impaired. Female polio survivors in the oldest age cohort may resemble their age-matched, non-disabled counterparts more closely at T.O.M. because they were less different from them to begin with. This interpretation is reinforced by the finding that mean levels of bone loss observed for the oldest cohort of polio survivors for both the Greater Trochanter and Ward's Triangle areas of the hip are consistent with population estimates of the average lifetime bone loss for all women for each type of bone -- i.e., 30% for cortical and 50% for trabecular, respectively. {76}

   Main and Interactional Effects of ERT. The data presented in Figures 32 and 33 indicate that, contrary to the first part of H3, ERT use had no significant main effect on bone loss for either the Greater Trochanter or Ward's Triangle areas of the hip. Averaged across sample membership, women with 'Noncontinuous' and 'Continuous' use of ERT did not have lower levels of bone loss compared to women with 'No History' of ERT. Moreover, although the general trend was for polio survivors to have more cortical bone loss than non-disabled controls across all three levels of ERT, the predicted two-way interaction of 'sample membership by ERT' was non-significant for both dependent variables. Partial support for H3 was observed, however, in the pattern of planned contrasts between sample groups within levels of ERT status. These within-level contrasts are referred to as "nested effects".
   

   Figure 32. Polio vs. Control: Interaction effects of ERT by Sample on percentage bone loss in hip (Greater Trochanter), adjusting for age at T.O.M.

   
   

   Figure 33. Interaction effects of ERT and Sample on percentage bone loss for Ward's Triangle area of hip, adjusting for age at T.O.M. (polio: 'No ERT' vs. 'Continuous ERT', p£ .05)

   For the Greater Trochanter area of the hip, polio survivors in the 'None' and 'Noncontinuous' categories of ERT had almost twice the rate of bone loss compared to age-matched, non-disabled controls with the same level of ERT usage (see Figure 32). Mean percentages for these nested effects are 25.7% versus 15.7% and 22.3% versus 13.0%, respectively (p= £ .05). Whereas, among women with 'Continuous' post-menopausal use of ERT there were no significant differences in bone loss by sample membership (20.2% vs. 16.8%).

   Consistent with H3, these results suggest two important findings: first, female polio survivors aging with the combined effects of long-term disability and post-menopausal estrogen deficiency appear to be more at-risk for osteoporosis compared to their non-disabled counterparts who are also aging without the benefit of estrogen replacement therapy, but who lack the added risk factors of disuse and misuse which also contribute to bone loss; and, second, that continuous use of ERT may mediate the impact of these risk factors by compensating for the long-term effects of disability on trabecular bone loss. The lack of significant main effects of ERT usage for either sample may have to do with the fact that cortical bone, which makes up 60% of the Greater Trochanter area, is less sensitive to estrogen levels than is trabecular bone.

   As the data in Figure 33 indicate, the pattern of effects for Ward's Triangle area o of the hip is considerably different. Unlike what was observed above for the differences in level of ERT usage are associated with significant differences in the percentage bone loss, but only for the polio sample. Female polio survivors with no history of ERT had significantly higher levels of trabecular-dominated bone loss compared to their counterparts who share the same long-term disability but who have had the advantage of continuous use of estrogen replacement therapy since the onset of post-menopausal period (42.9% vs. 31.5%, p = £ .05). Interestingly, among age-match non-disabled sample, differences in ERT status had no effects on percentages of bone loss. Using the Ward's Triangle area of the hip as the dependent variable, there were also no significant two-way interactions between sample membership and level of ERT usage.

   Findings from Figures 32 and 33 raise the question of why there was no main effect of ERT status on bone loss for the Greater Trochanter area of the hip and only partial evidence of a main effect of ERT for Ward's Triangle? Previous research in this area suggests that women on ERT will have significantly lower levels of bone loss compared to women with no on noncontinuous use of ERT. {28,75,76,95} One possible explanation for our failure to observe a significant ERT-main effect may have to do with a potential confounding between sample membership, length of time on ERT, and continuity of ERT use. If polio survivors were on estrogen replacement therapy longer than controls for both continuous and noncontinuous levels of ERT, then the hypothesized differences between these groups in percentage of bone loss could have been suppressed by differences in length of treatment. Evidence of such a confounding could explain why subgroup differences by ERT level were observed for the Greater Trochanter area of the hip, but not for Ward's Triangle. Unfortunately, follow up analyses not presented here provides no support for this explanation. At this stage of the investigation, further explanations for the lack of an ERT main effect go beyond the scope of our data.

   Three-Way Interactions with ERT. Figure 34 shows the results of the combined effects of age, disability and ERT status on bone loss for the Greater Trochanter area of the hip, adjusting for the effects of age at last period on the dependent variable. In this model, age at last period was used as a proxy to control for the potential effects of sample differences in the average age of menopause on percentage bone loss. Age at last period was chosen as a covariate because it is significantly related to type of termination of menses. For both samples, those having surgically terminated menses were considerably younger at age of last period compared to those whose menses terminated naturally. It was necessary to substitute age of last period for the dichotomous measure of termination of menses because the small sample size prevented the inclusion of a fourth classification factor, which would have expanded the design from eight to 24 cells and resulted in more missing data or empty cells. Similarly, to enhance statistical power, it was also necessary to replace the original ordinal version of ERT with dichotomous measure based on 'ERT usage' versus 'No ERT use'.
   

   Figure 34. Polio vs. Control: Combined effects of age, disability, and ERT on percentage bone loss in hip (Greater Trochanter), adjusting for age at last menstrual period (** p£ .01).

   Estimating the combined-effects model resulted in a significant three-way interaction among age, disability or sample membership, and level of ERT usage (p= £ .01). Observing a significant three-way interaction means that the effects of being in any one category of Age, Sample or ERT are not constant across levels of the other two variables. For example, as differences in the mean elevations of Figure 34 indicate, the effects on bone loss of being in the 60 and older age category are not the same for polio survivors versus controls, and within each sample, the effects of being this age are not the same by level of ERT. More specifically, it is only for polio survivors over the age of 60 that lack of estrogen replacement therapy is associated with a significant increase in percentage of bone loss. Polio survivors in the same age group but with a history of ERT do not differ significantly in percentage bone loss from age-matched controls or from their younger counterparts with or without ERT usage. Interestingly, and counter to expectations, among controls over the age of 60, there are no differences in bone loss between those with and without a history of ERT. Similarly, among participants under age 60, there were no significant differences in cortical-dominated bone loss between (Greater Trochanter) sample groups or within the same sample by level of ERT usage.

   We interpret this interaction as evidence of the "triple whammy" of aging, disability, and estrogen-deficiency on percentage bone loss, but only for the Greater Trochanter area of the hip. According to our data, only female polio survivors who meet all three conditions -- that is, are 60 years of age or older, have a life-long disability and no history of estrogen replacement therapy -- are at risk of loosing enough bone mineral mass to undermine skeletal function. Older polio survivors with continuous use of ERT, and younger survivors with ad without ERT, appear to be at no greater risk of osteoporosis than are non-disabled controls. Continuous ERT usage may also be helpful for women who have experienced the long term effects of polio in retaining their trabecular bone during the postmenopausal period, but this relationship needs to be pursued further in future research. The primary advantage of this combined effects model over previous models discussed is that it helps us to identify more precisely which sub-groups of polio survivors are most at risk of significant bone loss.

E. SUMMARY

Data from this exploratory study provide preliminary but consistent evidence of the greater risks of osteoporosis among female polio survivors compared to age-matched [non-disabled] controls. These greater risks are due to the combined effects of disability, post-menopausal aging and lack of ERT. This increased risk is illustrated by the fact that polio survivors loose bone mineral mass at a faster rate as age of participant increases from 50 to 88 compared to their non-disabled controls. By an average age of 65, female polio survivors without ERT have already lost more than 30% of cortical hip bone, which exceeds the average lifetime loss for the female population at large. We interpret these findings as providing initial support for the "accelerated aging" hypotheses. Our findings also suggest that early and continuous use of estrogen replacement therapy may constitute a promising intervention strategy to reduce the rate of bone loss and compensate for the long-term effects of disability, at least for the cortical-dominated bone in the Greater Trochanter area of the hip. The key question here is whether the benefits of ERT are associated only with pre-mature onset of menopause due to surgical termination.

Caution must be exercised in generalizing from these findings to other samples, however. Because of the small sample size, the importance of this analysis lies more in model building -- i.e., identifying significant predictors and subgroup differences -- than it does in providing tests of specific hypotheses. The next steps in this investigation of risks of osteoporosis among female polio survivors involve replicating these findings with a larger sample and systematically incorporating another dependent variable, such as Ward's Triangle, which is more sensitive to the effects of estrogen on bone mineral loss, across all analyses.

VI. CONCLUSIONS

Several important themes emerge from our investigation of the "the later-life effects" of aging with polio and stroke. They include: the importance of understanding the temporal structure of disability; the importance of measuring change (decline or improvement) over time in functional status; the importance of examining the role of the family -- both in terms of its supportive capacities and negative impacts; and, the importance of understanding the conceptual and methodological challenges of conducting cross-disability research.

The Utility of the Life Course Perspective

First finding from the LLE Study demonstrate the consequences of variations in the timing of onset of both primary and secondary disability. In both the polio and stroke samples, although less consistent for stroke, variations in the timing of acute onset were associated with significant differences in severity of initial and residual impairment, degree of functional limitation at time-of-measurement, depressive mood and acceptance of disability (measured at T.O.M.). For the polio sample, increases in age and historical period of acute onset were associated with greater impairment and functional limitations and more mental health problems. Whereas, for the stroke sample, the effects of age of onset varied significantly by laterality of CVA and gender, but contrary to expectations, the early-onset group was not consistently more vulnerable to psychosocial problems compared to those with later onset strokes, after age 60.

The timing of onset of stroke and the 'later-life effects' of polio (i.e., PPS [Post-Polio Syndrome]) also appear to influence how individuals interpret and respond to their changing life circumstances. Polio survivors who experience PPS at an early age, during their 50s, report significantly more negative consequences in terms of both depression and acceptance of disability compared to their counterparts who experience the new health problems and functional losses associated with PPS later, at a more normative age. Moreover, consistent with what we know about gender differences in social roles across the life course, female polio survivors appear to be more negatively affected by early onset of PPS then are their male counterparts. Interestingly, female stroke survivors also appear to be more adversely affected by the early, non-normative timing of their first CVA compared to both their older female counterparts and male stroke survivors; but, this relationship may be spurious do to the smaller number of female stroke participants.

The Importance of Measuring Change in Function

A second cross-sample theme highlights the importance of measuring change over time in functional mobility status rather than just relying on static measures of function at one or two time points. For both the polio and stroke samples, the amount of change (decline or improvement) they experienced in method of locomotion between residual or discharge from hospital and time-of-measurement was a significant predictor of their current physical and psychosocial well being. Not surprisingly, for the polio sample, both the amount and type of decline in mobility status were significantly related to their likelihood of developing post-polio syndrome. Those who declined the most between physical best and T.O.M., changing from unassisted ambulators to assisted ambulators and non-ambulators, were disproportionately represented among participants who met the criteria for PPS. They also had significantly lower scores on functional independence in basic and instrumental activities of daily living and on acceptance of disability compared to their counterparts who experienced no decline, regardless of their level of mobility. We interpret this result from the polio sample as consistent with Maenad and Roller's concept of "former passers." {54}. According to their definition, it is not surprising that this group should report the lowest scores on a acceptance of disability because they are the ones who were able to deny or "pass" for non-disabled most of their lives due to their mild residual impairment and the lack of visual evidence of disability.

Change over time in mobility was also significantly related to functional independence of stroke survivors at T.O.M., but the positive effects of improvement versus no improvement were restricted to independence in basic activities of daily living (ADLs) only. Change in mobility status had no direct effects on degree of independence in instrumental activities or on acceptance of disability.

The Importance of Family Involvement

The third cross-sample theme points to the importance of family involvement in mediating the impacts of disability-related stress and adversity on physical and psychological well being at T.O.M. Perhaps more important than the significant main effects of change in mobility status, are our findings that positive family function interacts with change in mobility status to moderate the overall negative relationship between decline, or no improvement, in mobility and functional independence or acceptance of disability. Moreover, support for the mediational effects of positive family function are stronger for the polio sample than they are for the stroke sample.

Among polio survivors who declined the most in mobility -- i.e., former "passers", those who report high levels of positive family functioning have significantly higher scores on acceptance of disability compared to their counterparts who experienced the same amount and type of functional decline but who report lower levels of positive family involvement. Within the polio sample, family function also interacts with severity of initial impairment and educational level to moderate the impacts of disability and socio-economic stress on mental health status. In both cases, the beneficial effects of 'high' positive family function are restricted to those who are experiencing the greatest functional limitations and/or adversity. We also observed significant gender differences in the mediational effects of family function. It is only for female polio survivors that low perceived levels of family functioning are associated with elevated depression scores. Among males, level of family function has no effect on depressive mood. This finding suggests that supportive ties to family members may not be equally important to men and women aging with the long-term effects of polio.

Positive communication and support from family also buffers the impact of no improvement in mobility status on the functional independence of stroke survivors. For this group, those who report high levels of positive family functioning are significantly more independent in basic activities of daily living compared to their counterparts with the same level of functional limitations but who report only moderate to low levels of family involvement. This finding is important because it demonstrates the capacity of "supportive" family relations to reinforce independence, rather than foster dependency, in persons aging with physical disability.

Within the stroke sample, we also have evidence of the potential negative effects of family involvement. Participants who reported that their stroke resulted in serious new family difficulties, including increased conflict and estrangement, were significantly less independent in instrumental activities of daily living at T.O.M. (i.e., shopping, traveling out of the home, and using the telephone) compared to those who reported no such change in the level of family stress. Moreover, lack of improvement in mobility status interacts with negative family impacts of stroke to further undermine the functional independence of sub-groups of stroke participants. Stroke survivors who experienced "the double whammy" of no improvement in mobility status and serious new family difficulties had significantly lower scores on functional independence in IADLs [Instrumental Activities of Daily Living] compared to all three of the other groups of stroke survivors.

Together, these findings illustrate both the "mixed blessing" of families and the "ripple effects" of disability on family relationships. Our data provide consistent evidence of the family's capacity to facilitate or undermine the well being of the individual and the success of rehabilitation interventions. Specifically, our findings highlight the need to involve families more in the research process in order to better understand the conditions under which families operate as sources of support or additional stress.

Challenges of Conducting Cross-Disability Research on Aging

When conducting cross-disability research, it is important to recognize that individuals with different disabling conditions, in this case polio and stroke, are coping with the long-term effects of disability at different stages of the life cycle, with different personal histories and health risk profiles, and with access to different levels of social support. Although emphasis on the temporal structure of disability is central to the life course perspective guiding this investigation, this model was not introduced until after the project was funded and data collection already underway. Because of this, we were unable to fully operationalize the life course framework in our selection and development of instruments and clinical assessment protocols. Specifically, we were unable to interrupt the study long enough to develop comparable and quantitative measures which could be used in between-sample analyses to test the utility of the life course perspective within a comparative framework. The net results was that within-sample analyses, particularly for the Polio sample, proved to be fruitful in terms of demonstrating the consequences of variations in the timing of disability than were the between sample comparisons.

Finally, findings from the LLE Study also point to the importance of education as both a personal and formal resource which contributes to positive adjustment to disability-related changes in physical and social functioning and increases the effectiveness of rehabilitation interventions (for Stroke sample only). Our findings regarding the "therapeutic" role of education highlight the need to look beyond the provision of traditional rehabilitation services to the use of continuing education programs as a promising intervention strategy.

REFERENCES

• 7 Anonymous, 1992. The Choices for Treating Osteoporosis. In: Health after 50. [nothing available online] The Johns Hopkins Medical Letter, Vol. 4, Issue2, pp 6-7.
• 27 Ettinger, V., 1992. Osteoporosis: a silent disabler. [nothing available online] Rehabilitation Research and Training Center on Aging, Rancho Los Amigos Medical Center, Downey, CA, Geriatric Rehab. Prev, 4(1), 1&5.
• 28 Ettinger, V., Genant, H.K. and Cann, C.E., 1985. Long-term estrogen replacement therapy prevents bone loss and fractures. [abstract] Annals of Internal Med, 102, 319-324.
• 32 Garland, D.E., Maric, Z., Adkins, R.H. and Steward, C.A., 1993. Bone mineral density about the knee in spinal cord injured patients with pathological fractures. [nothing available online] Contemporary Orthopaedics, 26(4), 375-379
• 33 Garland, D.E., Steward, C.A., Adkins, R.H., Hu, S.S., Rosen, C., Liotta, F.J. and Weinstein, D.A., 1992. Osteoporosis following spinal cord injury. [abstract] Journal of Orthopaedic Research, 10, 371-378.
• 34 Georgious, E., Ntalles, K., Papageorgiou, A., Korkotsidis, A. and Proukakis, C., 1989. Bone mineral loss related to menstrual history. [abstract] Acta Orthop Scand, 60(2), 192-194.
• 35 Genant, H.K., Cann, C.E., Ettinger, B., and Gordan, G.S., 1982. Quantitative Computed Tomography of Vertebral Sprongiosa: A Sensitive Method of Detecting Early Bone Loss After Oophorectomy. [abstract] Annals of Internal Medicine, 97(5), 699-705.
• 40 Halstead, L. and Rossi, C., 1985. New problems in old polio patients: Results of a survey of 539 polio survivors. [abstract] Orthopedics, 8, 845-850.
• 42 Halstead, L.S., Wiechers, D.O. and Rossi, C.D., 1985. Results of a survey of 201 polio survivors. [abstract] Southern Medical Journal, 78, 1281-1287.
• 43 Hamdy, R.C., Krishnaswamy, G., Cancellaro, V., Whalen, K. and Harvill, L., 1993. Changes in bone mineral content and density after stroke. [abstract] Am J Phys Med Rehabil, 72, 188-191.
• 44 Hansson, T. and Roos, B., 1986. Age Changes in bone mineral of the lumbar spine in normal women. [abstract] Calcif Tissue Int, 38, 249-251.
• 45 Johnston, C.C. and Melton, L.J., 1990. Bone density measurement and the management of osteoporosis. [nothing available online] In: Primer on Metabolic Bone Diseases and Disorder of Mineral Metabolism, pp. 93-100.
• 50 Kiely, M., Ubin, R.A. and Kiely, J.L., 1989. Descriptive epidemiology of cerebral palsy. [nothing available online] Public Health Rev., 12, 79-101.
• 51 Kleerekoper, M. and Avioli, L.V., 1990. Evaluation and treatment of postmenopausal osteoporosis. [nothing available online] In: Primer on Metabolic Bone Diseases and Disorders of Mineral Metabolism, pp 151-154.
• 53 Lindquist, O., Bengtsson, C., Hansson, T. and Jonsson, R., 1983. Changes in bone mineral content of the axial skeleton in relation to aging and the menopause. [abstract] Scan J Clin Lab Invest, 43, 33-338.
• 54 Maynard, F.M., Julius, M., Kiesch, N., Lampman, R., Peterson, C., Tate, D., Waring, W. and Werner, R., 1991. The Late effects of polio: A model for the identification and assessment of secondary disabilities, (Final Report). [nothing available online] National Techinical Information Service, U.S. Dept. of Commerce, 5285 Port Royal Rd., Springfield, VA 22161.
• 62 Nilas, L., Borg, J., Gotfredsend, A. and Christiansen, C., 1985. Comparison of Single- and dual-photon absorptiometry in postmenopausal bone mineral loss. [abstract] J Nucl Med., 26(11), 1257-1262.
• 63 Nordin, B.E.C., Need, A.G., Bridges, A. and Horowitz, M., 1992. Relative contributions of years since menopause, age, and weight to vertebral density in postmenopausal women. [abstract] J Clin Endocrinology & Metabolism, 74, 20-23.
• 71 Polednak, A.P., 1989. Racial and ethnic differences in disease. New York: Oxford University Press.
• 73 Raisz, L.G. and Johannesson, A., 1984. Pathogenesis, prevention and therapy of osteoporosis. [abstract] Journal of Medicine, 15(4), 267-277.
• 75 Richelson, L.S., Wahner, H.W., Melton, L.J. and Riggs, B.L., 1984. Relative contributions of aging and estrogen deficiency to postmenopausal bone loss. [abstract] New England J Med, 311, 1273-1275.
• 76 Riggs, B.L. and Melton, L.J., 1992 The Prevention and Treatment of Osteoporosis. [No abstract available.] New England J Med, 327(9): 620-627.
• 77 Riggs, B.L. and Melton, L.J., 1986. Involutional Osteoporosis. [no abstract available] New England J Med, 314(26), 1676-1686.
• 89 Trojan, D.A., 1993. Osteoporosis. [nothing available online] In: Polio Network News, Vol. 9, Issue 3, pp. 4. International Polio Network, St. Louis, MO.
• 90 Trieschmann, R.B., 1987. Aging with a Disability. [nothing available online] New York: Demo Publications.
• 91 U.S. Senate Special Committee on Aging, 1991. Aging America: Trends and Projections. [nothing available online] Washington, D.C.: U.S. Government Printing Office.
• 95 Weinstein, L., 1957. Influence of age and sex on suseptibility and clinical manifestations in poliomyelitis. [nothing available online] New England J Med., 257, 47-52.
• 99 Werner, R., Waring, W. and Maynard, F., 1991 Osteoarthritis of the Hand and wrist. [nothing available online] In: Late Effects of Polio: A model for Identification and assessment of preventable secondary disabilities, (Final Report). National Technical Information Service, U.S. Dept. of Commerce, 5285 Port Royal Rd., Springfield, VA 22161.
• 100 Whiteneck, G. (Ed.), 1992. Aging with a Spinal Cord Injury. [nothing available online] NY.: Demos.
• 104 Worley, R.J., 1981. Age, estrogen and bone denisity. [abstract] Clin Obstetrics & Gynecology, 24(1), 203-218.