Depression, Falls, and Risk of Fracture in Older Women

Background  Previous studies have suggested that depression is associated with falls and with low bone density, but it is not known whether depression leads to an increased risk of fracture.

Subjects and Methods  We conducted a prospective cohort study in elderly white women who were recruited from population-based listings in the United States. At a second visit (1988-1990), 7414 participants completed the 15-item Geriatric Depression Scale and were considered depressed if they reported 6 or more symptoms of depression. We measured bone mineral density (BMD) in the spine and hip using dual energy x-ray absorptiometry at the second visit, and asked participants about incident falls (yes/no) at 4 follow-up visits. Nonvertebral fractures were ascertained for an average of 6 years following the depression measure, and verified radiologically. We determined incident vertebral fractures by comparing lateral spine films obtained at the first visit (1986-1988) with repeat films obtained an average of 3.7 years later (1991-1992).

Results  The prevalence of depression (Geriatric Depression Scale score ≥ 6) was 6.3% (467/7414). We found no difference in mean BMD of the hip and lumbar spine in women with depression compared with those without depression. Women with depression were more likely to experience subsequent falls than women without depression (70% vs 59%; age-adjusted odds ratio [OR], 1.6; 95% confidence interval [CI], 1.3-1.9; P<.001), an association that persisted after adjusting for potential confounding variables (OR, 1.4; 95% CI, 1.1-1.8; P=.004). Women with depression had a 40% (age-adjusted hazard ratio [HR], 1.4; 95% CI, 1.2-1.7; P<.001) increased rate of nonvertebral fracture (124 fractures in 3805 woman-years of follow-up) compared with women without depression (1367 fractures in 59,503 woman-years of follow-up). This association remained strong after adjusting for potential confounding variables, including medication use and neuromuscular function (HR, 1.3; 95% CI, 1.1-1.6; P=.008). Further adjustment for subsequent falls appeared to explain part of this association (HR, 1.2; 95% CI, 1.0-1.5; P = .06). Women with depression were also more likely to suffer vertebral fractures than women without depression, adjusting for history of vertebral fracture, history of falling, arthritis, diabetes, steroid use, estrogen use, supplemental calcium use, cognitive function, and hip BMD (OR, 2.1; 95% CI, 1.4-3.2; P<.001).

Conclusions  Depression is a significant risk factor for fracture in older women. The greater frequency of falls among individuals with depression partially explains this finding. Other mechanisms responsible for the association between depression and fracture remain to be determined.

FIFTEEN PERCENT of elderly individuals report clinically relevant symptoms of depression.¹ Depression is associated with increased disability,² poor physical function,^3-6 falls,^7-12 and low bone density,^13-15 all of which increase susceptibility to osteoporotic fractures.^16-18 Older individuals who are depressed may also be at increased risk for falls and fracture due to the effects of antidepressants or sedatives.^12,19-22 Patients with depression have poorer recoveries following fractures.^23-25 However, it is not known whether depression is associated with an increased risk of fracture. To determine the risk of falls and fracture in women with depression, we measured depression, physical function, and bone density in a cohort of older women, and followed them for incident falls and fractures.

A total of 9704 white ambulatory women who were at least 65 years of age were recruited from population-based listings in Baltimore, Md, Minneapolis, Minn, Portland, Ore, and the Monongahela Valley, Pa, between September 1986 and October 1988 for the prospective Study of Osteoporotic Fractures.¹⁸ At a second visit (1988-1990), 7414 of these women (78% of survivors) completed at least 10 of 15 items on the Geriatric Depression Scale (GDS) and underwent measurement of bone density. These 7414 women are the subjects of this analysis. This study was approved by the appropriate institutional review boards, and all subjects provided written informed consent.

The 15-item GDS is a validated and reliable checklist of depressive symptoms designed to detect current depression in the elderly.²⁶ This self-report scale consists of 15 yes/no questions regarding symptoms of depression (eg, have you dropped many of your activities and interests?), including 5 reverse scored items (eg, do you feel happy most of the time?). We used the standard cutoff point of 6 or more symptoms to define depression, with 6 to 10 symptoms indicating mild to moderate depression, and 11 or more symptoms indicating severe depression.²⁷ This cutoff point of 6 or more symptoms has a sensitivity of 88% and a specificity of 62% compared with a structured clinical interview for depression.²⁸ For the 3% of participants who completed between 10 and 14 GDS items, we estimated the total score by dividing the reported number of symptoms by the proportion of items completed.

Bone mineral density (BMD) of the lumbar spine and hip (including femoral neck, intertrochanteric region, trochanter, and the Ward triangle) was measured using dual energy x-ray absorptiometry (QDR 1000; Hologic, Waltham, Mass). The interscanner coefficient of variation was 1.5% for the spine, 1.2% for the femoral neck, and 0.9% for an anthropometric hip phantom.²⁹

We ascertained self-reported incident falls (yes/no) at follow-up visits (1989-1996). Women returned postcards, or were contacted by telephone, for incident nonvertebral fractures every 4 months for an average (±SD) of 6 ± 2 years (range, up to 8 years) following the depression measure. The cumulative rate of postcard returns was 99%. All reported fractures were confirmed by reviewing radiographic reports, blinded to results from the GDS. Unless they also suffered a nontrauma fracture, participants who suffered fractures resulting from severe trauma (eg, motor vehicle crash, struck by a car, hit by rapidly moving projectile, or assault) were excluded from the analysis.

We determined incident vertebral fractures by comparing lateral spine films obtained at the first visit with follow-up films obtained at a third visit (1991-1992, an average of 3.7 years later) in 6281 (85%) of 7414 participants. Lateral spine radiographs were assessed for fractures using quantitative morphometry.³⁰ A woman was classified as having an incident vertebral fracture if any vertebral height decreased by 20% or more and 4 mm or more between the baseline (1986-1988) and follow-up radiographs (1991-1992). Thus, some "incident" vertebral fractures may have occurred before the depression measure was administered in 1988-1990.

Self-reported age, marital status, education, medical history, smoking, alcohol use, current perceived health status (excellent/good vs fair/poor/very poor), caffeine intake, supplemental calcium, estrogen, thiazide, nonthiazide diuretic, and steroid use were determined from a questionnaire administered at the first visit and reviewed by an interviewer.³¹ Lifetime physical activity was estimated using a modified Paffenbarger scale that assesses the type and duration of weight-bearing activities.^32,33 Participants were asked about the frequency with which they performed each of 40 activities at 4 times during their lives (teenager, ages 30 and 50 years, and previous year). The number of times per year each activity was performed was multiplied by 5 kcal/min for low-intensity (eg, walking or gardening), 7.5 kcal/min for medium-intensity (eg, dancing or tennis), or 10 kcal/min for high-intensity activities (eg, jogging or skiing). Scores for these 4 times were added to estimate lifetime physical activity (weighted number of times per year), which has been associated with increased bone density in this cohort.³⁴ We measured current physical activity by asking subjects the question, "Did you participate in any physical activity, recreation, or sport in the past week?"

We determined history of vertebral fractures by radiograph in all women at the first visit. We ascertained self-reported history of falls (yes/no) at the second visit (1988-1990). Also at the second visit, participants were asked: "In the past 12 months, have you taken any medications for anxiety or nerves or to relax muscles? If yes, write down the name of the medication you use most often for anxiety, nerves or relaxation." Participants who listed taking a tricyclic antidepressant or trazodone at least twice per week were categorized as users of antidepressants. No participants listed monoamine oxidase inhibitors, selective serotonin reuptake inhibitors, or other antidepressants. Those who listed taking a benzodiazepine, barbiturate, or other sedative/hypnotic at least twice per week were categorized as users of sedatives/hypnotics.

Body mass index (BMI), blood pressure, cognitive function, neuromuscular function, and functional status were measured at the second visit. Weight was measured (in light indoor clothing with shoes removed) using a balance beam scale, and height was measured using a stadiometer.³⁵ Body mass index was calculated as weight in kilograms divided by the square of the height in meters. Blood pressure was measured with the participant supine and after she had been standing for 1 minute³⁶; orthostatic hypotension was defined as a drop in systolic blood pressure of 20 mm Hg or more on standing. Cognitive function was measured by a trained examiner using Digit Symbol, a subtest of the Wechsler Adult Intelligence Scale.³⁷ Scores equal the number correct within the timed trial, lower scores indicating poorer performance.

We assessed neuromuscular function and frailty using 5 measures: quadriceps strength, grip strength, gait speed, tandem stand, and time rising from a chair. Quadriceps strength was measured using a leg-extension chair (Bodymasters MD 110; Lafayette Instruments Co, Lafayette, Ind), grip strength of the dominant hand was assessed as the average of 2 attempts with a Preston dynamometer (Sammons-Preston Company, Burridge, Ill), and gait speed was determined as the time in seconds needed to walk 12 m. Poor performance on tandem stand was defined as holding a tandem stand position with eyes open for less than 10 seconds. Time rising from a chair was defined as seconds needed to complete 5 chair stands.

Functional status was measured on a 39-point scale with up to 3 points of difficulty (some difficulty, much difficulty, or unable to do) for each of 13 activities (eg, dressing, bathing, preparing meals, doing housework, shopping, or walking 2 or 3 blocks) based on a modified version of the Stanford Health Assessment Questionnaire.³⁸

Differences in baseline characteristics between subjects with and without depression were compared using χ² tests for dichotomous variables and t tests for continuous variables. We used analysis of covariance to compare adjusted mean BMD in subjects with and without depression. Because the exact dates of falls and of vertebral fractures were not known, we used logistic regression to analyze the association between depression and these outcomes.

We used proportional hazards models to analyze the association between depression and all nonvertebral fractures as well as the associations between depression and each of the 6 most common types of nonvertebral fractures. We verified the proportionality assumption of these models. To obtain adjusted risk estimates, we entered all variables listed in Table 1 into backwards elimination proportional hazards models that included depression. Variables that were associated with fracture (at P<.05) were retained in these models. To adjust for the variables related to neuromuscular function and frailty (tandem stand, grip strength, quadriceps strength, gait speed, and time rising from a chair), which were strongly intercorrelated, we selected the variable most strongly associated with the outcome. We then analyzed the association between depression and fracture using a proportional hazards model that included the variables identified by the backwards models. The survival curve for time to first nonvertebral fracture was estimated using an adjusted Cox proportional hazards model.³⁹ We used age-adjusted proportional hazards models to examine the rate of fracture associated with specific items on the GDS.

For all analyses, we report odds ratios (ORs) or relative hazards with 95% confidence intervals (CIs). Continuous data are presented as mean ± SD. Analyses were performed using the Statistical Analysis Software (SAS Institute Inc, Cary, NC).

The prevalence of depression (GDS ≥6) was 6.3% (467/7414). Compared with women without depression, women with depression were older and more likely to smoke, use antidepressants and sedatives, report fair or poor health and poor functional status, and perform poorly on measures of neuromuscular and cognitive function. Women with depression were more likely to have a higher BMI, to have gained weight during their adult years, and to have lost weight during the 2 years since entering the cohort. They were less likely to be married, well educated, or physically active (Table 1). Women (n = 211) who completed between 10 and 14 items on the GDS were more likely to be depressed than women (n = 7203) who completed all 15 items (12% vs 6%; P<.001).

Follow-up information regarding subsequent falls was available from 7317 women; 36 died and 61 were lost to follow-up. Follow-up for nonvertebral fractures was complete in 7069 women, 285 were excluded because they reported nonvertebral fractures that could not be confirmed radiographically, and 60 were excluded because they suffered fractures due to severe trauma. Follow-up for vertebral fractures was complete in 6281 women, 994 did not have follow-up lateral spine films, and 139 died or were lost to follow-up.

We found no difference in mean BMD between women with and without depression, even after adjusting for potential confounding variables (Table 2). In post hoc analyses, we found a significant interaction with depression, BMI, and BMD. Among the 2473 women in the highest tertile of BMI (>27.6 kg/m²), women with depression had a 4.6% lower mean spine BMD (0.88 ± 0.16 vs 0.93 ± 0.16 g/cm²; P = .007) and a 2.6% lower hip BMD (0.80 ± 0.12 vs 0.82 ± 0.11 g/cm²; P = .03) than women without depression adjusted for age, marital status, education, arthritis, diabetes, alcohol, current and lifetime physical activity, smoking, steroid use, estrogen, thiazide, nonthiazide diuretic, benzodiazepine, calcium supplement, perceived health, weight change, BMI, and quadriceps strength. Depression was not associated with BMD among the 4943 women in the lower 2 tertiles of BMI.

Women with depression were more likely to fall during follow-up than women without depression (70% vs 59%; age-adjusted OR, 1.6; 95% CI, 1.3-1.9; P<.001). This association persisted after adjusting for the variables listed in Table 1 (OR, 1.4; 95% CI, 1.1-1.8; P = .004).

Nonvertebral fractures occurred in 124 (28%) of 447 women with depression and 1367 (21%) of 6622 women without depression (P<.001). Participants who suffered subsequent fractures had a mean ± SD depression score of 1.9 ± 2.4 compared with 1.6 ± 2.1 in those who did not suffer fractures (P<.001). Greater numbers of depressive symptoms were associated with increased fracture risk (P for trend <.001; Figure 1).

Women with depression had a 40% (age-adjusted hazard ratio [HR], 1.4; 95% CI, 1.2-1.7; P<.001) increased rate of nonvertebral fracture (124 fractures in 3805 woman-years of follow-up) compared with women without depression (1367 fractures in 59,503 woman-years of follow-up). These results were similar when women with fractures due to severe trauma were included in the analysis (HR, 1.4; 95% CI, 1.2-1.7; P<.001) and when women who did not complete all 15 items on the GDS were excluded from the analysis (HR, 1.4; 95% CI, 1.2-1.7; P<.001). After adjustment for potential confounding variables, women with depression had a 30% increased rate of fracture compared with women without depression (Table 3 and Figure 2). Further adjustment for subsequent falls appeared to explain part of this association (HR, 1.2; 95% CI, 1.0-1.5; P = .06).

Women with depression had an increased rate of rib fracture, adjusted for history of vertebral fracture, history of falling, arthritis, diabetes, steroid use, estrogen use, supplemental calcium use, cognitive function, and hip BMD (Table 4). We also noted a trend for increased rate of hip, foot, and ankle fractures in women with depression, but women with depression did not have an increased rate of wrist, humerus, or other fractures. In post hoc analyses, 12 of 15 symptoms on the GDS were associated with an increased rate of fracture, including 8 that were statistically significant at the P<.05 level (Table 5).

Women with depression were more likely to suffer vertebral fractures than women without depression (11% vs 5%; age-adjusted OR, 2.3; 95% CI, 1.6-3.2; P<.001). Compared with women without depression, women with depression had a 2-fold increased odds of vertebral fracture, adjusted for history of vertebral fracture, history of falling, arthritis, diabetes, steroid use, estrogen use, supplemental calcium use, cognitive function, and hip BMD (OR, 2.1; 95% CI, 1.4-3.2; P<.001).

Depression is associated with a substantially increased risk of falls and fracture in elderly, white women. The adjusted risk of nonvertebral fracture associated with depression is similar to that associated with a 14% (SD, 0.6) lower mean calcaneal BMD.¹⁸ Although depression has been associated with low bone density in previous studies, low BMD did not explain the relationship between depression and fracture.

We attempted to explain the link between depression and fracture by exploring several other potential pathways in our analysis. Patients with depression have an increased prevalence of chronic medical illnesses. However, adjustment for history of stroke, hypertension, chronic obstructive pulmonary disease, myocardial infarction, arthritis, and diabetes did not affect the strength of the association between depression and fracture. Likewise, greater smoking or alcohol use and less physical activity among patients with depression did not explain this relationship. Use of antidepressants or sedative/hypnotic medications can lead to falls and fractures in older individuals,^12,20-22 but adjustment for these variables did not influence the association between depression and fracture. Depression has been associated with poor health and physical functioning in older individuals.^3-5 After adjusting for health, functional status, orthostatic hypotension, cognitive function, and neuromuscular function, however, the relative rate of fracture decreased from 1.4 to 1.3, suggesting that these variables only partially link depression with fracture.

Depression is a risk factor for falls in older individuals,^7-12 and our results indicate that women with depression had a 40% increased odds of subsequent falls. This greater frequency of falls among women with depression appeared to explain part, but not all, of the relationship between depression and fracture. Depression also leads to physical decline in older individuals.^6,40 Whether greater physical decline among women with depression is responsible for the association between depression and fracture remains to be determined.

In a post hoc attempt to determine the aspects of depression that may confer an increased risk of fracture, we analyzed the rate of fracture associated with each of the 15 items on the GDS. Feelings of hopelessness, worthlessness, and dissatisfaction had the strongest relationships with fracture. Lacking energy, feeling in poor spirits, and preferring to stay at home were not associated with fracture. Perhaps women who lacked energy and preferred to stay at home had less exposure to environmental conditions that may lead to fracture.

Our results suggest that older women with depression have BMD that is similar to those for women without depression. Three previous studies^13-15 found that women with depression had lower BMD than women without depression. One study¹³ reported that 80 inpatients with major depression had a 15% lower spine BMD than 57 healthy controls. Another study¹⁴ found that 68 psychiatric inpatients, including 21 patients with major depression, had spine and hip BMD 1 to 2 SDs below normal. A third study¹⁵ reported that 24 women with past or current major depression had lower spine and hip BMD than 24 matched controls. Perhaps the association between depression and BMD is restricted to younger patients, or to those with more severe or longer-lasting depression than the women in our study. It is also possible that lower bone density among the inpatients with depression in previous studies may have resulted from the relative immobilization associated with hospitalization.⁴¹

In a post hoc analysis, we found that BMD was slightly lower among women with depression with high BMI, but not among those in the lower 2 tertiles of BMI. This raises the possibility that hypercortisolemia may explain the association between depression and BMD found in previous studies. Both obesity and depression are associated with hypercortisolism, which is a risk factor for low BMD in patients with endogenous or exogenous corticosteroid excess.^42-46 However, only 1 study,¹⁴ which examined the association between plasma cortisol levels and BMD in patients with milder hypercortisolism,has found that increased plasma cortisol levels are associated with low BMD in patients with major depression.

Several limitations deserve comment. First, because we determined incident vertebral fractures by comparing repeated spine films with those obtained at the first visit, some vertebral fractures may have occurred before the depression scale was administered. Second, we adjusted for antidepressants and sedatives/hypnotics taken "for anxiety, nerves, or to relax muscles," but some women who were using antidepressants may not have responded affirmatively to this question. Third, because we did not do a clinical interview for depression, we can only conclude that depressive symptoms, and not necessarily the clinical diagnosis of depression, are associated with falls and fracture. Fourth, we did not measure other potential links in the pathway between depression and fracture such as cortisol levels, indexes of bone quality, or markers of bone turnover. Finally, we studied only elderly white women, whose characteristics may differ from those of other populations.

The poorer health and functional status among women with depression in our study highlights the severe disabilities associated with depression.^5,40,47,48 Although depression is common among elderly individuals, most cases go unrecognized or undertreated.^1,49 Results from this study identify depression as a risk factor for fracture in older women, but offer only a partial explanation for this association. If treatment for depression were effective in reducing the risk of falls and fractures, and in improving recovery following fractures, then better diagnosis and treatment of depression could substantially decrease fracture-related morbidity and mortality.

Accepted for publication August 15, 1998.

Dr Whooley is supported by a Research Career Development Award from the Department of Veterans Affairs Health Services Research and Development Service. This study was supported by grants AG05394, AG05407, AR35582, AR35583, AR35584, and NS36016 from the Public Health Service.

We are indebted to Li-Yung Lily Lui for her assistance with data analysis.

Reprints: Mary A. Whooley, MD, General Internal Medicine Section, San Francisco Veterans Affairs Medical Center, 4150 Clement St, 111A1, San Francisco, CA 94121 (e-mail: whooley@itsa.ucsf.edu).

University of California, San Francisco (Coordinating Center): S. R. Cummings (principal investigator), M. C. Nevitt (coinvestigator), D. G. Seeley (project director), D. M. Black (study statistician), H. K. Genant (director, central radiology laboratory), C. Arnaud, D. Bauer, W. Browner, L. Christianson, M. Dockrell, C. Fox, R. Gore, S. Harvey, M. Jaime-Chavez, L. Laidlaw, R. Lipschutz, L. Lui, G. Milani, L. Palermo, R. San Valentin, K. Stone, H. Tabor, D. Tanaka, and C. Yeung.

University of Maryland, Baltimore: J. C. Scott (principal investigator), R. Sherwin (coinvestigator), M. C. Hochberg (coinvestigator), J. Lewis (project director), E. Peddicord (clinic coordinator), A. Bauer, C. Boehm, G. Cullum, L. Finazzo, M. E. Flaks, T. Ford, D. Harris, B. Hohman, E. Oliner, T. Page, J. Schlossberg, C. Shaffer, A. Trimble, and S. Trusty.

University of Minnesota, Minneapolis: K. Ensrud (principal investigator), P. Schreiner (coinvestigator), C. Bell (project director), E. Mitson (clinic coordinator), C. Bird, D. Blanks, S. Estill, S. Fillhouer, S. Fincham, J. Griffith, J. Hansen, F. Imker-Witte, K. Jacobson, K. Kiel, K. Knauth, N. Nelson, E. Penland-Miller, and M. Riley-Alves.

University of Pittsburgh, Pittsburgh, Pa: J. A. Cauley (principal investigator), L. H. Kuller (coprincipal investigator), M. Vogt (coinvestigator), L. Harper (project director), L. Buck (clinic coordinator), C. Bashada, D. Cusick, G. Engleka, A. Githens, M. Gorecki, K. McCune, D. Medve, M. Nasim, C. Newman, S. Rudovsky, and N. Watson.

The Kaiser Permanente Center for Health Research, Portland, Ore: E. Harris (principal investigator, project director), W. M. Vollmer, E. Orwoll, H. Nelson (coinvestigators), K. Crannell (project administrator, clinic coordinator), J. Bender, A. Doherty, K. Easter, M. Erwin, F. Heinith, J. Kann, K. Redden, C. Romero, K. Snider, and C. Souvanlasy.