Familial resemblance and shared latent familial variance in recurrent fall risk in older women

Abstract

Background: A possible familial component to fracture risk may be mediated through a genetic liability to fall recurrently. Methods: Our analysis sample included 186 female sibling-ships (n = 401) of mean age 71.9 yr (SD = 5.0). Using variance component models, we estimated residual upper-limit heritabilities in fall-risk mobility phenotypes (e.g., chair-stand time, rapid step-ups, and usual-paced walking speed) and in recurrent falls. We also estimated familial and environmental (unmeasured) correlations between pairs of fall-risk mobility phenotypes. All models were adjusted for age, height, body mass index, and medical and environmental factors. Results: Residual upper-limit heritabilities were all moderate (P < 0.05), ranging from 0.27 for usual-paced walking speed to 0.58 for recurrent falls. A strong familial correlation between usual-paced walking speed and rapid step-ups of 0.65 (P < 0.01) was identified. Familial correlations between usual-paced walking speed and chair-stand time (−0.02) and between chair-stand time and rapid step-ups (−0.27) were both nonsignificant (P > 0.05). Environmental correlations ranged from 0.35 to 0.58 (absolute values), P < 0.05 for all. Conclusions: There exists moderate familial resemblance in fall-risk mobility phenotypes and recurrent falls among older female siblings, which we expect is primarily genetic given that adult siblings live separate lives. All fall-risk mobility phenotypes may be coinfluenced at least to a small degree by shared latent familial or environmental factors; however, up to approximately one-half of the covariation between usual-paced walking speed and rapid step-ups may be due to a common set of genes.

falls are associated with 80% and 90% of all nonvertebral and hip fractures among older adults, respectively (6), and treating fall-related injuries costs approximately $20 billion annually (2, 3). Both twin (1, 5) and family (4) studies have provided evidence for a familial component to fracture risk in older adults that is independent of bone density, potentially mediated through a genetic liability to fall recurrently. A recent study in older female twins enrolled in the Finnish Twin Study on Aging (FITSA), reported that familial genetic and nongenetic factors explain 40% of interindividual variation in recurrent falls (9).

A potential genetic regulation of recurrent fall and fracture risk is suggested in part through genetic influences on leg muscle performance in middle-aged and older women in the United Kingdom and Finland. FITSA and other twin cohort studies report that genetic factors influence individual variation in knee-extension strength (30, 31, 29, 32) and leg extensor power (1, 31, 29, 32), but not in ankle plantar flexion strength (30). A relatively small genetic effect on maximal walking speed (17, 20, 29) has also been reported. Furthermore, it has been suggested that a common set of genes may be coregulating knee-extension strength and leg extensor power (12) and maximal walking speed (11).

Muscle performance is only a small aspect of fall risk due to the capacity of multiple physiological body systems to compensate for muscle deficits and for individuals with muscle performance deficits to make cognitive decisions to avoid riskier activities (such as walking at maximal speeds). There are essential mobility tasks required for independent living that are fundamentally related to the risk of recurrent falls: ability to walk, ability to stand up from a chair, and ability to step rapidly to avoid an obstacle or as a protective response to common balance perturbations. Furthermore, impaired mobility is associated with a fivefold increased probability for recurrent falls (10). However, data on the potential genetic regulation of these fall-risk mobility phenotypes are very limited. In older male twins enrolled in the U.S. National Heart Lung and Blood Institute Study (NHLBI) Twin Cohort (6), genetic factors were reported to influence over 50% of the interindividual variation in usual-paced walking speed and timed chair stand performance. In adult women of various ages, enrolled in the Australian Twin Registry and Research Program (11), FITSA (21), and the Swedish Twin Registry (34), genetic factors were reported to regulate various balance measures that relied on self-reported data (34) and included assessments not closely related to fall risk in the community (11, 21).

There is a dearth of information on the potential genetic susceptibility to fall recurrently. In the current study, we examine for the first time familial resemblance in fall-risk mobility phenotypes and recurrent falls in a family study of community-dwelling older female siblings in the United States. Furthermore, we examine for the first time, the extent to which shared latent familial factors coregulate fall-risk mobility phenotypes.

METHODS

Subjects and setting.

All women enrolled in the Study of Osteoporotic Fractures (SOF) between 1986 and 1988 were eligible (n = 9,704). SOF participants were aged 65 yr and older, Caucasian (99.6%), had no bilateral hip replacements, and were able to walk without assistance. Two hundred and six sibling-ships, defined as having at least two full-sibling sisters were identified (n = 433), with a range of two to five sisters per family. Siblings were identified during enrollment as women referred their sisters or reported having a sister already enrolled. The sibling sample in SOF was similar to the total nonsibling sample in age (71.9 vs. 71.7 yr) and reported a similar prevalence of fair or poor self-reported general health (19 vs. 17%), but had fewer recurrent fallers 13 vs. 21%. The analysis sample for all fall-risk traits consisted of a subset including 186 sibling-ships (n = 401) who had data available for all potential covariates. The study protocol was approved by Institutional Review Boards represented from each SOF clinic, and all participants provided written informed consent.

Recurrent fall-risk phenotypes: mobility and recurrent falls.

Time to stand up from a chair five times without the use of arms from a 45-cm-high seat was recorded. The number of step-ups completed (including stepping up and back down) in 10 s on a 23-cm-high step while grasping a handrail was obtained. Usual-paced walking speed (m/s) was assessed over a 6-m course. Postcards were mailed every 4 mo and each postcard queried participants as to whether or not they had fallen and, if so, how many times. Recurrent falls were defined as two or more falls over 2 yr. A threshold of two or more falls was used since more frequent falls are associated with increased fracture risk (25) and isolated falls are commonly associated with environmental causes of falls such as slippery surfaces (5).

Covariates.

Age was determined and standing body height (m) and weight (kg) were measured (22) and body mass index (BMI) was calculated (kg/m²). Self-reported physician diagnoses of diabetes, stroke, and any arthritis regardless of site were documented. Other measurements included: cognitive impairment (<23/26) on the modified Mini Mental State Examination (14, 43), visual impairment defined as being in the first quartile of visual acuity (4), and orthostatic hypotension (drop in systolic blood pressure ≥20 mmHg on standing).

Self-reported measures of educational level (completed high school, Yes/No), smoking, alcohol consumption (≥1 beverage in past 30 days), average daily dietary protein intake in grams (6), self-reported fair/poor general health (Yes/No), and sedentary lifestyle (>20 h/day spent lying down or sitting) were documented. Use of central nervous system (CNS) active medications (including sleep aids, anti-anxiety, benzodiazepines, barbiturates, nonbenzodiazepine sedative/hypnotics, antidepressants, and muscle relaxants) were identified from medications brought into the clinic.

Lifetime physical activity was assessed using the Paffenbarger questionnaire (18, 19). Participants were asked to report the number of times they were physically active in the past year for various ages: their current age, age 50, age 30, and as a teenager. Number of times active in each of 33 possible physical activities listed was weighted according to its intensity level defined as low, medium, and high. Weighting of the number of times physically active according to intensity level defined as low, medium, and high, was calculated as follows: 5*(number of times they were active for light activities) + 7.5*(number of times they were active for moderate activities) + 10*(number of times they were active for heavy activities). The weighted number of times physically active was then averaged over each of the four ages at which it was assessed. Examples of low-level activities included walking, bicycling, skating, badminton, gardening, and golfing with cart. Medium-intensity activities included activities such as hiking, swimming, dance exercise, aerobic dance, square dance, other dance, and golfing with walking. Examples of high-intensity activities included jogging, running, skiing, racquetball, and squash. Due to the skewed distribution of lifetime physical activity, we used poor lifetime physical activity as a covariate defined as low quartile of lifetime physical activity (Yes/No).

Statistics.

The potential role of familial influences on fall-risk mobility phenotypes and recurrent falls was assessed by calculating residual upper-limit heritability. Residual upper-limit heritability is an estimate of the proportion of phenotypic variation due to additive effects of genes and shared nongenetic factors after adjusting for measured covariates. Higher values of residual upper-limit heritabilities imply greater familial contributions to individual differences (13). Residual upper-limit heritabilities were estimated using maximum likelihood methods under the variance component framework in SOLAR (Sequential Oligogenic Linkage Analysis Routines) (1). The household option in SOLAR was not used as adult siblings live separate lives. The total phenotypic variation is partitioned into a “familial” component including shared genetic and nongenetic factors, an individual-specific measured “covariate” component (e.g., age, BMI, height, etc.), and a residual “unmeasured environment” component.

The distributions of fall-risk mobility phenotypes were examined for nonnormality and chair-stand time was transformed using log and number of step-ups using square root. Distributions were further assessed for outliers (±4 SD) that were removed due to probable data entry errors. There were five women excluded with scores of zero on the step-up test and one woman excluded with a time of 46.2 s on the timed chair-stand test. Intercorrelations for each of the three pairs of fall-risk mobility phenotypes were determined from between effects linear regression models. Residual upper-limit heritabilities were then estimated in base- and full-multivariate models. Base-multivariate models were adjusted for age, BMI, and height. Full-multivariate models were adjusted for age, BMI, and height in addition to other selected covariates. Covariate selection was performed in SOLAR using variance component frameworks to account for correlations among family members. Specifically, we used the screening option in SOLAR in which all covariates were initially screened in the same order one at a time and only a subset of those covariates demonstrating a P value <0.15 were included in the full-multivariate model. A conservative criteria (P value <0.15) was used to remain in the multivariate model, so that potentially important covariates would not be unnecessarily eliminated. We also performed principal component factor analysis on the three fall-risk mobility phenotypes to generate a mobility function factor (11). Upper limit heritability of this mobility function factor was determined using the same methods as the three separate fall-risk mobility phenotypes.

Using bivariate maximum likelihood analyses in SOLAR, we estimated the degree to which shared latent familial factors and common (unmeasured) environmental factors coinfluence performance on rapid step-ups, chair-stand time, and usual-paced walking speed. Bivariate methods partition the total phenotypic correlation (ρ_P) between two quantitative traits into genetic correlations and environmental correlations. Because calculations of genetic correlations in our study of sibling pairs potentially include common environmental effects, they are referred to as familial correlations. Familial correlations (ρ_F) estimate the degree to which the same genetic factors, including the same gene or a common set of genes and the same unmeasured (unadjusted) familial environmental factors coinfluence two phenotypes within an individual. Environmental correlations (ρ_E) specifically estimate the degree to which the same unmeasured (or unadjusted) environmental factors coinfluence two phenotypes within an individual. All familial and environmental correlations were calculated using the combined set of covariates selected for inclusion in the full-multivariate models of residual upper-limit heritability for any two pairs of traits.

The ρ_F and ρ_E were obtained for pairs of selected fall-risk mobility traits from the phenotypic variance-covariance matrix. Statistical significance was determined using likelihood ratio statistics. We subsequently calculated a residual phenotypic correlation (ρ_P), or the sum of both residual familial and unmeasured environmental components (7a), between two traits: ρ_P = [sqrt(h2r1)*sqrt(h2r2) *ρ_F] + [sqrt(1−h2r1)*sqrt(1−h2r2)*ρ_E]. The h2r1 and h2r2 were residual upper-limit heritabilities for the two traits that are recalculated in SOLAR using the combined set of covariates that were selected for inclusion in the full-multivariate models of residual upper-limit heritability for any two pairs of traits. The ρ_F and ρ_E represent familial and common (unmeasured) environmental correlations, respectively, between any two given pairs of traits after adjusting for covariates.

Covariates with genetic and/or environmental components and as well as known independent associations with fall risk among older adults were included in all multivariate methods. We necessarily controlled for these covariates because of their potential influence on estimates of upper-limit heritability and familial and environmental correlations.

RESULTS

Table 1 provides descriptive characteristics of the sample. The mean age of women was 71.9 yr (SD=5.0), with 37% of women aged 65–69, 37% aged 70–74, 17% aged 75–79, 7% aged 80–84, and 2% aged 85–88.

Table 1.

Characteristics of 401 women among 186 sibling-ships

Measures	Percent, Mean (SD), or Median (IQR)
Fall-Risk Traits
Recurrent falls, %	13.3%
Chair-stand time (s to complete 5 repetitions)	11.8 (9.8–14.3)
Rapid step-ups (number complete in 10 s)	9.2 (8.0–11.0)
Usual-pace walking speed, m/s	1.00 ± 0.20
Physical, Medical, and Environmental factors
Age, yr	71.9 (5.0)
BMI, kg/m2	26.2 (4.9)
Height, cm	158.3 (5.9)
High school education, %	68.3%
Orthostatic hypotension, %	13.7%
Fair-to-poor self-rated heath, %	18.4%
Stroke, %	2.2%
Arthritis, %	61.8%
Protein intake, g/day	348.0 (135.3)
Visual acuity corrected for near distance (no. correct)	51 (45–54)
Diabetes, %	7.7
Cognitive impairment, %	12.2
Uses CNS-active medications, %	30.2%
Sitting or lying ≥4 h/d, %	9.5%
Lifetime physical activity (intensity-weighted no./yr.)	3,050 (1,300–5,480)
Current smoker, %	11.0
Consumes alcohol, %	50.1

IQR, interquartile range; CNS, central nervous system.

Twenty-one percent of all women fell at least once during the first year of study and 12% fell at least twice over the first two years (recurrent falls). Eight percent of women had diabetes, 62% had arthritis, 14% had orthostatic hypotension, 2% had a stroke, and 30% used a central nervous system-active medication. Intercorrelations of usual walking speed and number of step-ups, number of step-ups completed, and time to stand up from a chair, and usual walking speed and time to stand up from a chair, were 0.59, −0.53, and −0.42, respectively.

Table 2 shows the proportion of the total phenotypic variation due to familial factors (residual upper-limit heritabilities), covariates, and environmental (unmeasured) factors in the base- and full-multivariate models for all fall-risk mobility phenotypes. Results were similar in the base- and full-multivariate models. Residual upper-limit heritabilities were moderate to strong for all fall-risk phenotypes (P < 0.05, for all), ranging between 27% for walking speed to 58% for recurrent falls after controlling for age, body size, and other covariates.

Table 2.

Proportion of total phenotypic variation attributable to covariates, familial factors, and environment

	Base-Multivariate Modela			Full-Multivariate Modelb
	Covariates	Familial Factors	Environment (unmeasured)	Covariates	Familial Factors	Environment (unmeasured)
1. Recurrent falls	0.10a	0.63	0.27	0.18e	0.58	0.24
		(0.29)c			(0.31)c
2. Chair-stand time	0.14a	0.33	0.47	0.21f	0.43	0.36
		(0.13)d			(0.13)d
3. Number of rapid step-ups	0.18a	0.51	0.31	0.29g	0.44	0.40
		(0.13)d			(0.13)d
4. Usual-paced walking speed	0.13a	0.24	0.63	0.22h	0.27	0.51
		(0.13)c			(0.13)c

n = 401.

^aAdjusted for age, height, and BMI;

^badjusted for age, height, BMI (all models) and selected covariates as indicated: arthritis (AR), diabetes (DI), stroke (ST), visual impairment (VI), poor general health (HEA), cognitive impairment (CI), and orthostatic hypotension (OH) education (ED), smoking (SM), alcohol (AL), protein intake (PR), CNS-active medications (CNS), sedentary lifestyle (SED), and poor lifetime physical activity (PA);

^cresidual upper-limit heritability (mean ± SE), all P < 0.05;

^dresidual upper-limit heritability (mean ± SE), all P < 0.01;

^eadditionally adjusted for VI, CNS;

^fadditionally adjusted for AR, ST, VI, HEA, CI, ED;

^gadditionally adjusted for AR, ST, HEA, CI, AL, PR, CNS, PA;

^hadditionally adjusted for DI, ST, VI, HEA, CI, PA, SM.

The proportions of the total phenotypic variation in the mobility function factor due to covariates, familial (residual upper limit heritabilities), and unmeasured environmental factors were 21%, 33% (SE=13), and 46%, respectively, in base-multivariate models adjusted for age, height, and BMI and 33%, 34% (SE=13), and 33%, respectively, in full-multivariate models adjusted for age, height, BMI, arthritis, stroke, visual impairment, poor general health, cognitive impairment, orthostatic hypotension, alcohol, and sedentary lifestyle. Residual upper limit heritability estimates were significant in both base- and full-multivariate models (P = 0.01 for both).

Familial, environmental, and residual phenotypic correlations between pairs of fall-risk mobility phenotypes are shown in Table 3. The three pair-wise sets of chair-stand time, usual-paced walking speed, and number of rapid step-ups all demonstrated moderate residual phenotypic correlations ( ρ_P : 0.33–0.45). A strong familial correlation was evidenced between usual-paced walking speed and number of rapid step-ups (0.65 ± 0.24), P < 0.01. Familial correlations between usual-paced walking speed and chair-stand time (−0.02) and between chair-stand time and number of rapid step-ups (−0.27) were both nonsignificant. Moderate environmental correlations were evidenced for all three pairs of fall-risk mobility phenotypes ( ρ_P : 0.35–0.58), P < 0.05 for all. The familial correlation for usual-paced walking speed and number of rapid step-ups was stronger than the corresponding environmental correlation (0.65 ± 0.24 vs. 0.35 ± 0.12).

Table 3.

Pair-wise correlations among fall-risk mobility phenotypes

Phenotype	ρF ± SE	ρE ±SE	ρP	Covariates*
Chair-stand time & No. of rapid step-ups	−0.27 ± 0.22	−0.58 ± 0.12†	−0.45	AR, ST, VI, HEA, CI, ED, AL, PR, CNS, PA
Usual walking speed & No. of rapid step-ups	0.65 ± 0.24*	0.0.35 ± 0.12*	0.43	AR, DI, ST, VI, HEA, CI, SM, AL, PR, CNS, PA
Usual walking speed & Chair-stand time	−0.02 ± 0.32	−0.50 ± 0.12†	−0.33	ST, VI, HEA, CI, ED, SM, PA

*All correlations are adjusted for age, height, and BMI and a subset of the following covariates as specified: AR, DI, ST, VI, HEA, CI, OH, ED, SM, AL, PR, CNS, SED, and PA.

^*P < .01,

^†P < .001. Note: ρ_F is the familial correlation, ρE is the environmental correlation, and ρ_P is the residual phenotypic correlation.

Finally, we assessed the robustness of our results and reran all analyses without the exclusion of the six women with extreme outlier data. Instead, we substituted the lowest and highest acceptable values (e.g., those within ±4 SD of the mean) for these women and essentially obtained the same results (data not shown).

DISCUSSION

We identified moderate residual upper-limit heritabilities in fall-risk mobility phenotypes and recurrent falls in older female sisters after controlling for extensive physical, medical, and environmental factors. While all three pairs of fall-risk mobility phenotypes demonstrated a moderate degree of familial resemblance, only usual-paced walking speed and number of rapid step-ups were suggested to be coinfluenced to a relatively large degree by shared familial factors, with over one-half of the covariation between these two traits potentially due to a common set of genes. All three pairs of fall-risk mobility phenotypes were suggested to be coinfluenced to a relatively small degree by shared (unmeasured) environmental factors.

Our residual upper-limit heritability findings for all recurrent fall-risk phenotypes suggest that recurrent fall-risk runs in families of older sisters, with as much as 27% to 58% of the interindividual variation in these traits potentially attributable to genetic factors. Since our older sibling sample is estimated to have spent a minimum of 40 yr living apart since living together with their parents as children, we believe our residual upper-limit heritabilities closely approximate the proportion of interindividual variation due to genetic factors after accounting for important genetic and environmental confounders. Because recurrent falls was the only phenotype with a residual upper-limit heritability >50%, this suggests a potential stronger role of genetic influences on the interindividual variation in recurrent falls than of unmeasured environmental factors. Conversely, a stronger role of unmeasured environmental influences on interindividual variation in usual-paced walking speed, chair-stand time, number of rapid step-ups, and the mobility function factor than of genetic factors is suggested.

Our estimate of the residual upper-limit heritability for recurrent falls was consistent although higher (0.58 vs. 0.40) than familial effects for recurrent falls identified among female twins in Finland (22). One reason for the increased familial resemblance in our older sample is our inclusion of women aged 77 to 88 yr old (range: 65–88 yr), whereas the Finnish study included younger women aged 63–76 yr. Risk factors for falls are multifactorial and environmental sources of risk are suggested independent of physiological deficits (7, 9, 16), particularly for younger adults (15, 24, 33). Thus it is possible that environmental sources of recurrent falls were more prevalent in the younger twin cohort (63–76 yr) than in our older family cohort (65–88 yr), which would be consistent with lower estimates of familial effects in the younger twin cohort (22). In fact, recurrent fallers demonstrated similar maximal walking speeds and maximal isokinetic knee extensor strength at baseline as nonfallers, which would not be expected if falls were primarily due to physiological sources (22).

Our residual upper-limit heritabilities estimates for usual-paced walking speed (0.27 vs. 0.51) and timed chair-stands (0.43 vs. 0.56) were on average lower than those identified among male siblings in the United States (6), despite the overlapping 95% confidence intervals calculated as residual upper-limit heritability ±1.96 (SE) (i.e., 0.01–0.52 and 0.18–0.68, respectively). Given the clear component of muscle strength in these phenotypes, part of the discrepancy may be due to genes having a smaller influence on muscle strength in women than men (3) consistent with the lower heritability estimates among our female sample.

In contrast, our residual upper-limit heritability estimate for the rapid step-up test (0.51) was not consistent with one prior null finding of a single-stance rapid step test (17) among 83 female twin pairs aged 50 yr and older. A possible explanation for the different results is because a single-stance rapid step test as described by Haber and colleagues (11) involved dynamic weight shift up and down a 7.5-cm-high block as fast as possible using only one leg at a time. This is a very different task than the test used in the current study, which involved stepping up and down on a 23-cm-high step with both legs while using a handrail for support. Consequently, we expect that our two-legged rapid step test required significantly less ankle plantar flexion strength compared with the one-legged rapid step test performed by Haber and colleagues (11). Given a prior study reported no evidence of genetic effects on ankle plantar flexion strength in older females twins (30), the differential involvement of ankle plantar flexion strength between these two step tests may explain the different results consistent with higher familial effects among a step test with ankle plantar flexion strength involved to a lesser degree. There are no other published studies that have examined the familial influences on mobility function factors.

The direct phenotypic correlation between number of rapid step-ups and usual-paced walking speed was primarily attributable to familial genetic and nongenetic factors and, to a lesser degree, shared latent environmental factors. A familial correlation of 0.65 indicated that up to 42% of the covariation in rapid step-ups and usual-paced walking speed (or ρ_F squared) was attributable to shared familial influences, including a common set of genes and possibly some common environment. Due to the complex physiological systems involved in the regulation of self-selected walking speed and ability to coordinate rapid and accurate steps up and back down again, there are potentially many genes influencing these traits. However, these findings still suggest that older sisters may share a genetic predisposition toward speeds of usual-paced walking and rapid step-ups that may be influenced in part by a common set(s) of genes, in addition to a unique set(s) of genes. While genes influencing muscle strength and power have been identified (4) to the best of our knowledge, there are no data yet exploring the specific genes influencing fall-risk mobility phenotypes.

While our findings suggest moderate familial effects on all three fall-risk phenotypes, our findings suggest that the familial factors influencing the variance in usual-paced walking speed are largely unique from those influencing the variance in chair-stand time and that the familial factors influencing the variance in chair-stand time are largely distinct from those influencing the variance in rapid step-ups. While all three mobility tasks involve to varying degrees similar leg muscle groups, chair-stand time probably requires more muscle strength in relation to either usual-paced walking speed or number of rapid step-ups and, specifically, more involvement of eccentric muscle contractions. For instance, the timed chair-stand task involves taking on all of one's body weight on one's feet without the aid of arms while standing up and then subsequently lowering all one's body weight without the use of arms back to the seated position. In contrast, the usual-paced walking speed and rapid step-up tasks involve only a transfer of one's body weight from one position to another. Additionally, chair-stand time involves a relatively large eccentric muscle contraction component as participants lower their full body weight into a chair without the use of arms each time after standing. In the timed chair-stand task eccentric contractions involve a full 90-degree range of motion when sitting down; whereas walking includes very little eccentric contractions and stepping rapidly includes only a relatively small eccentric muscle contraction component with a 20- to 30-degree range of motion as the legs are bending to lower the body from the 23-cm-high step back down to the floor. Thus varying degrees of muscle strength and more specifically involvement of eccentric muscle contractions among the three fall-risk mobility traits, as well as potential differential degrees of genetic regulation on eccentric and concentric muscle contractions (8), may explain why relatively low familial correlations were identified between chair-stand times and usual-paced walking speed and between chair-stand time and number of rapid step-ups.

Environmental correlations were identified for all three sets of fall-risk mobility traits independent of genetic and environmental covariates. The relatively strong environmental correlations identified between rapid step-ups and chair-stand times and between usual-paced walking speed and chair-stand times suggested that 34% and 25%, respectively, of the covariation in these two sets of traits may be influenced by a shared set of environmental factors. In addition, there was a moderate environmental correlation of 0.35 identified between number of rapid step-ups and usual-paced walking speed, suggesting that 12% of the covariation between these two traits may be influenced by a common set of unmeasured environmental factors. While adult siblings lead separate lives, there may be latent effects of growing up in the same household that extend well into old age that influence covariation in fall-risk mobility phenotypes. For example, shared attitude and values between siblings may lead to similar exercise and dietary practices, use of preventive health services or other community resources, and adherence to treatment regimens that all may improve health and function.

The current study has potential limitations. The same analyses performed in more extended family units would likely yield more precise estimates than those we derived from the female sibling-ships; however, even with our family study of sibling pairs, we were still able to detect significant findings. Our upper-limit heritability estimates may include effects from common environmental factors and common unmeasured confounders and, therefore, we may have overestimated a potential role of genetic factors in fall-risk phenotypes. Still, we believe that potential influences of common environment between pairs of sisters is minimal since older siblings live separate lives and we estimate have lived a minimum of 40 yr apart since living together with their parents.

Despite these potential limitations, our study is one of only two studies to have examined a possible role of genetic influences on recurrent falls in older adults. Furthermore, our study is the first family study to examine a potential role of genetic influences on recurrent falls in older women in the United States. Furthermore, unlike the narrow age range in the previous study of Finnish twins, our study included a wide age-range of older siblings and potentially frailer individuals who may fall for different reasons than younger healthier individuals. Therefore, our study provides new insight into the likely importance of genetic influence for a broader range of recurrent falls that occur among older adults. Our study is also the first study to examine a potential role of genetic factors on fall-risk mobility phenotypes in older women and, additionally, the first study (that we know of) to examine the degree to which shared family and shared environmental factors influence covariation between fall-risk mobility phenotypes in older adults of any sex.

In addition, the first study to examine the influence of genetics on recurrent falls among Finnish twins (22) did not account for genetic regulation on body height (26) and common geriatric conditions (2, 13, 14, 27, 28) that could potentially bias results. In fact, we adjusted all our analyses for age and the following carefully measured covariates with both genetic and environmental components: height (35), obesity (10, 19, 23), diabetes (21, 30), stroke (5), cognition (24, 42), orthostatic hypotension (18, 37), vision (37), habitual physical activity levels (23), and self-rated health (36). By adjusting for covariates with both genetic and environmental components, we removed the shared variance due to genetic and environmental influences on individual differences associated with that covariate. Thus the proportion of familial variance contributing to individual differences in fall risk in the current study cannot be explained by known genetic contributions from common acute and chronic geriatric conditions or habitual physical activity levels that are reported to have independent associations with fall risk in older adults (5, 9, 12, 16, 33).

In conclusion, our study found that fall-risk mobility phenotypes and recurrent falls aggregate in families of older Caucasian female siblings. These findings suggest a possible genetic liability in fall risk, particularly for recurrent falls. Furthermore, shared latent familial factors were suggested to coinfluence usual-paced walking speed and number of rapid step-ups, potentially due to a common set of genes. While shared unmeasured environmental influences may be an important source of covariation between fall-risk mobility phenotypes, uncommon sources are probably even more important. New studies are needed to confirm an existence of genetic liability in fall-risk among older women, particularly for recurrent falls, and to examine the extent to which common genetic and common environmental factors influence fall-risk mobility phenotypes. A better understanding of a possible role of genetic factors in fall risk may eventually lead to novel fall therapies and an increased potential for reducing fractures among older women.

GRANTS

The Study of Osteoporotic Fractures is supported by National Institutes of Health (NIH) funding under the following grant numbers: AG-05407, AR-35582, AG-05394, AR-35584, and AR-35583. Additional NIH support by AG-00181, P30-AG-024827, and UL1-RR-024153.