|Home | About | Journals | Submit | Contact Us | Français|
Osteoporosis and depression may be associated through common physiologic systems or risk factors.
To assess the associations between depressive symptoms (Burnam’s scale) or antidepressant use and bone outcomes.
Prospective cohort study.
A total of 93,676 postmenopausal women (50 to 79 years old) enrolled in the Women’s Health Initiative Observational Study.
Self-reported fractures (n=14,982) (hip [adjudicated], spine, wrist, and “other”). Analyses included 82,410 women with complete information followed on average for 7.4 years. Bone mineral density (BMD) of the hip (n=4539), spine (n=4417), and whole body (n=4502) was measured at baseline and 3 years in women enrolled at 3 densitometry study sites.
Overall, there were no statistically significant associations between depressive symptoms or antidepressant therapy and 3-year change in BMD. In a subset of women not using antidepressants, there was a significant difference in whole-body BMD change between women with and without depressive symptoms (P=.05). Depressive symptoms (hazard ratio [HR] 1.08; 95% CI=1.02 to 1.14) and antidepressant therapy (HR=1.22; CI=1.15 to 1.30) independently increased risk of any fracture, the majority of which occurred at “other” anatomic sites. Antidepressant therapy increased the risk of spine fracture (HR=1.36; CI=1.14 to 1.63). No associations were observed between depressive symptoms or antidepressant therapy and hip or wrist fracture.
In this study of postmenopausal women, average age 64, we observed minimal association between depressive symptoms and 3-year changes in either BMD or fracture risk. Antidepressant use was not associated with changes in BMD, but was associated with increased risk of fractures at the spine and “other ” anatomic sites.
Osteoporosis and depression are common in women over age 50.1,2 Both conditions have multifaceted etiologies, but have physiologic processes in common that could provide biologic plausibility for an association. Corticosteroids, cytokines, and serotonin have been postulated to play a role in both depression3–6 and bone homeostasis.7–11 Alternatively, the link could be caused by factors associated with depression that affect bone health such as inactivity, nutrition, weight, calcium intake, or smoking. Antidepressants may also have a direct effect on bone through serotonin pathways12,13 or have a complex relationship through falls or depressive symptomatology and severity.
Among the first epidemiological evaluations of depression and bone health was a small case-control study by Michelson and colleagues.14 They observed an association between clinically assessed depression and lower bone mineral density (BMD) and higher urinary cortisol excretion in premenopausal women. Prospective studies are limited. In postmenopausal women, 2 longitudinal studies have reported an association between depressive symptoms and hip BMD 15,16 with 1 reporting no association at the spine.15 Longitudinal studies more consistently report a positive association between depressive symptoms and risk of nonvertebral or hip fractures.17–20 Only 1 study examined other anatomic sites and reported an increase risk of spine but not wrist fracture after adjustment for antidepressant use.17
This prospective study investigates: (1) the association between self-reported depressive symptoms and bone health (3-year changes in BMD and fracture risk) at multiple anatomic locations in a large cohort of postmenopausal women while controlling for covariates known to influence bone outcomes, and (2) whether the use of antidepressants independently affects these associations.
Participants were 93,676 postmenopausal women, aged 50 to 79 years, enrolled in the Women’s Health Initiative (WHI) Observational Study (OS) between 1994 and 1998.21 Bone mineral density (BMD) measurements were collected on a subset of women (n=6,441) enrolled at the 3 designated bone densitometry clinics.22 Details have been previously described.23–26 Women were excluded if it was unlikely that they could participate for at least 3 years, including “serious emotional problems, mental illness, or too much stress, that would make it hard for women to be in a research study”.24 Therefore, women with more severe depression were likely excluded from the study. Human subjects review committees at each participating institution approved the study.22
Baseline depressive symptoms were assessed by self-report using Burnam’s 8-item scale for depressive disorders (major depression and dysthymia).27 This scale combines 6 questions about frequency of depressive symptoms from the Center of Epidemiologic Studies Depression Scale (CES-D), with 2 questions about symptom duration from the Diagnostic Interview Schedule (DIS). The distribution of scores was highly skewed, suggesting a binomial distribution. Therefore, as has been done in other studies,28 a cut point greater than or equal to 0.06 was used to dichotomize the continuous score.
To measure antidepressant use, participants were asked to bring all current medications to baseline interviews.29 Medication used for 2 or more weeks was recorded along with duration of use. We evaluated selective serotonin reuptake inhibitors (SSRIs) in subanalyses.
Areal BMD measures (g/cm2) were obtained at the hip, posterior-anterior lumbar spine (L2-L4 region), and the whole body at baseline and 3 years using dual-energy x-ray absorptiometry (DXA) Hologic QDR-2000 Series Bone Densitometers. Quality assurance was maintained through central review of scans, cross-calibration of scanners, and corrections for longitudinal changes.30
Any, hip (nonpathological), spine (thoracic or lumbar), wrist/lower arm (radius/ulna or 1 or more carpal bones) and “other” (tailbone, pelvis, femur shaft, knee, lower leg, ankle, foot, heel, clavicle, shoulder, upper arm, elbow, or hand [metacarpal]) inpatient and outpatient fractures were self-reported annually.31 As spine x-rays were not performed, morphometric vertebral fractures could not be assessed; only clinically noted vertebral fractures are included. Hip fractures were centrally adjudicated by hospital record review. Traumatic fractures were not excluded.
During the baseline visit, demographic data, reproductive and medical history, lifestyle, and behavioral factors were self-reported using standardized questionnaires.21,23 The physical function score (lowest function  to highest ) was calculated from the 10-item SF-36 physical functioning subscale.
Distributions of participants’ baseline characteristics were calculated in strata defined by depressive symptom status and use of antidepressant therapy.
Multivariable linear regression methods were used to assess the associations between 3-year changes in BMD and either depressive symptoms or antidepressant use. One model was developed for all anatomic sites. To identify confounding baseline variables, known determinants of BMD32 (Table 1) were included in models and variables that changed the point estimate by less than 10% were removed.33 Using this process, the baseline variables of BMD, age (years), height (cm), weight (log kg), ethnicity (African-American, Latino, white, or other), physical function score, oral hormone therapy (opposed or unopposed; never, past, and current use), current smoking, and years since menopause were retained in the model. Owing to missing data on BMD, depression, and other variables, complete case analyses for hip, spine, and total body included 4,539, 4,417, and 4,502 participants, respectively.
Semi-parametric Cox survival models were used to calculate hazard ratios for fracture occurrence. Length of follow-up (average 7.4 years) was calculated from day of enrollment to day of fracture, and observations were censored at the last follow-up visit or death (n=5996). The proportional hazards assumption was tested, and stratified estimates (stratum-specific baseline hazard with equal coefficients across strata) were calculated if needed. All analyses used multivariable models, developed as described above, and included baseline factors of age (years), height, weight (log kg); ethnicity (African-American, Latino, white, or other); years since menopause, physical function score, exercise, current smoking, ever having cardiovascular disease (CVD), analgesic/narcotic use, and history of previous fracture. After accounting for missing data and loss to follow-up (n=507), complete case analysis included 82,410 women.
Use of antidepressants can be a marker for more severe depressive symptoms or can reduce depressive symptoms and potentially obscure associations. To remove possible medication effects in our analyses of depressive symptoms, we performed subanalyses restricted to women not using antidepressants at baseline. Similarly, antidepressants are usually prescribed for depressive symptoms. To control for variability in bone outcomes that might be explained in part by depressive symptoms, we controlled for depressive symptoms in our analyses of antidepressant use. As falls could be a mediating factor between depressive symptoms17 or antidepressant use34 and fracture, we controlled for the number of falls (0, 1, 2, or 3+) in the year before baseline in a separate subanalysis. All analyses were performed using STATA version 9 software (College Station, TX, USA).
Of the women in the full cohort (n=93,676), 11% (n=10,368) had depressive symptoms at enrollment (Burnam score ≥0.06); 8% (n=7,212) were using antidepressants and nearly half of these were using SSRIs (Table 1). On average, the women were 64 years old and 15 years beyond the age of menopause. Compared to women without depressive symptoms, at baseline women with depressive symptoms were less likely to be married or living with a partner or to have graduated from college. They exercised less, took less calcium supplementation, had lower physical function, were more likely to be non-white, and were more likely to have arthritis or CVD. Compared to women not receiving antidepressant therapy, women using antidepressants on average had lower physical function, exercised less, and were more likely to be using hormone therapy.
Of the women enrolled at the 3 BMD centers (n=6,441), 14% (n=877) had depressive symptoms, 8% (n=508) were receiving antidepressant therapy, and 3% were using SSRIs (n=220). Women in the BMD subset were similar to the full cohort in age, BMI, and physical function, but were less likely to be white (77% vs 84%). Approximately 80% of the women had 3-year BMD measures. Women who were missing 3-year hip BMD measures had a higher probability of baseline depressive symptoms (mean Burnam score 0.07 vs 0.05).
Women with depressive symptoms had slightly higher average hip BMD measures at baseline (T score −0.78 versus −0.90) but similar spine (−0.64 vs −0.68) and whole body (−0.78 vs −.083) BMD compared to those without depressive symptoms. We did not observe a statistically significant association between depressive symptoms and mean change in 3-year BMD at the hip or spine (Table 2). In women not using antidepressants, there was a minimal, but statistically significant 3-year difference in whole-body BMD change associated with depressive symptoms(−0.0036 g/cm2; P=.05).
At baseline, women using antidepressants had higher mean BMD at the hip (T score −0.74 vs −0.90), spine (−0.38 vs −0.71), and whole body (−0.63 vs −0.85) compared to non-users. We did not observe any statistically significant associations between the use of antidepressants and 3-year changes in hip, spine, or whole-body BMD, after controlling for depressive symptoms and other potential confounding variables (Table 3).
During follow-up, women experienced 14,982 fractures: 1,132 (1%) hip fractures; 1,607 (2%) clinical vertebral fractures; 2,867 (3%) wrist fractures; and 11,128 (12%) fractures at “other” skeletal sites. The average time from enrollment to the first fracture occurrence was 4.4 years: hip 4.9 years, spine 5.2 years, wrist 4.7 years, and “other” locations 4.3 years.
There were no statistically significant differences in adjusted risk of hip, spine, or wrist fracture between women with and without depressive symptoms (Table 4). Depressive symptoms were associated with a minimal increased risk of any fracture (hazard ratio [HR] 1.08; 95% CI=1.02 to 1.14), the majority of which occurred at “other” anatomic sites (HR 1.16; 95% CI=1.09 to 1.23).
We observed no association between antidepressant therapy and risk of hip or wrist fracture after controlling for depressive symptoms and other covariates (Table 5). Compared to women not using antidepressants, women using antidepressant medication had increased adjusted risk of any fracture (HR=1.22; 95% CI=1.15 to 1.30), clinical spine fracture (HR=1.36; 95% CI=1.14 to 1.63) and fractures at “other” locations (HR=1.24; 95% CI=1.15 to 1.32). Associations at these sites persisted, although the magnitude of the association decreased, after controlling for number of falls in the last 12 months (data not shown). This increase in risk was similar for women using antidepressants for less than or greater than 6 months at baseline. The risk estimates for SSRIs were similar in magnitude across anatomic sites although statistical significance was not achieved at the hip.
In this large prospective study, after adjusting for major covariates, we found no association between 3-year bone loss at the hip, spine, or whole body and self-reported depressive symptoms in our primary analyses. In a subanalysis of women not using antidepressants, we observed a slight difference in whole-body BMD change favoring women without depressive symptoms compared to those with symptoms.
To date, the evidence supporting an influence of depression on changes in BMD in postmenopausal women has not been uniformly strong. A few small cross-sectional and case-control studies reported an inverse association between clinically diagnosed depression and femoral or spinal BMD,14,35,36 whereas a larger cross-sectional study found no association.37
A recent 1-year longitudinal study of 320 postmenopausal white women (average age 56 years) did not observe an association between depressive symptoms and BMD loss at the more trabecular spine.15 A second study including older white women (average age 76, n=4,177) reported an inverse relationship between depressive symptoms and total hip BMD.16 Both studies reported an association between depressive symptoms and bone loss at the more cortical femoral neck region, but not at the trochanter. These studies are consistent with our findings in that we saw a decrease in whole-body BMD, a measure dominated by cortical bone. The disparity between the findings may partly be explained by differences in population ages, race, and the division of hip BMD by cortical and trabecular subregion.
After controlling for depressive symptom status, we observed no association between antidepressant use and 3-year changes in BMD at the hip, spine, or in whole body. There is limited literature regarding the association between antidepressant therapy and BMD change. A recent cohort study followed 2,722 older women (average age 78.5; white only) for an average of 5 years and reported decreases in hip BMD associated with SSRI use, but not tricyclic antidepressants (TCAs), after controlling for depressive symptoms and in analyses that removed women with more severe depressive symptoms.38 Rate of bone loss was not different in women who reported using antidepressants at only 1 versus both time points. The younger age of participants and shorter follow-up in our study could contribute to the difference in findings.
We observed no increased risk of hip, wrist, or self-reported spine fractures associated with depressive symptoms in postmenopausal women. We did observe a slight increased risk of “any” fractures, most of which occurred at “other” skeletal sites. Two other studies, which did not adjust for antidepressant use, have reported elevated risk of hip fracture associated with only the highest level of depressive symptoms or mental distress.19,20 Another study has reported that mental distress, present at 2 different time periods, increased the odds of nonvertebral fractures; their sample size did not allow for stratification by use of “nerve” medicine.18 Similar to our results, Whooley et. al. found an increased risk of self-reported, nontraumatic, nonvertebral fractures in older women (n=7,414) with depressive symptoms after adjusting for antidepressant use, along with no statistically significant associations for hip or wrist fractures when assessed separately.17 In contrast to our findings, these authors observed higher odds of vertebral fracture in women. Their method for identifying spine fractures compared baseline and follow-up lateral radiographs. In our study, fractures were initially self-reported, which precluded assessment of morphometric spine fractures.
We observed an association between antidepressant therapy and increased risk of any fracture, clinical spine fracture and fractures at “other” skeletal locations independent of depressive symptom status. Similarly, 2 other large cohort studies have reported an increase in nontraumatic, nonvertebral fractures associated with antidepressant39 or SSRI40 therapy.
In contrast to our findings, a number of case-control studies using automated data have reported increased risk of hip fracture with antidepressant use.41–46 This study design can be more vulnerable to bias owing to unmeasured covariates.47 One cohort study has also reported a 65% increase risk of hip fracture with antidepressant therapy.39 Compared to this study, our study analyzed a greater number of hip fractures (1,094 vs 288) and a larger percent of women received antidepressant therapy (8% vs 6%); this difference may be because of the younger average age of our study population (64 vs 77 years) or their exclusion of traumatic fractures.
The following points should be considered when interpreting the study results. Our study was well powered; we had 80% power to detect 3-year changes in hip and whole-body BMD of 0.008 g/cm2 and a 0.01 g/cm2 change at the spine. However, certain characteristics inherent to our study could partially explain our null BMD findings. The women in our study were heavy (mean BMI 27 kg/m2; SD 5.9) and a disproportionate number of minority women were recruited at the densitometry sites. In addition, the WHI exclusion criteria included serious emotional problems and mental illness so it is likely the women in our study had fewer depressive symptoms than the general population. However, the range in the Burnam depressive symptom score appeared adequate (0.00 to 0.96) to provide variability. We also used a measure of self-reported depressive symptoms, not a clinical diagnosis of depression. An ancillary study assessed the reliability of the Burnam instrument in a subset of WHI OS volunteers and reported 74% sensitivity and 87% specificity among women with DSM-IV defined current major depression and dsythmia.28 Such measurement error could bias our results toward the null. However, another study of younger women with low levels of depressive symptoms excluded women with low BMD and still reported an association between femoral neck BMD and Beck Depression Inventory score (BDI).15 These authors repeated their analysis replacing the BDI with the Burnam measure and found persistence of the association with changes in femoral neck BMD (p=0.09). Finally, depression may be associated with study withdrawal or with incomplete information. We did observe higher baseline depressive scores in women with incomplete BMD data.
Our hip fractures were centrally adjudicated; however, all other types of fractures were self-reported so morphometric vertebral fractures could not be assessed. Local adjudication of a sample of self-reported WHI fractures confirmed 80% of other fractures (n=3,011), but only 52% of spine (n=302) fractures.25,31 Vertebral fractures are also reported to be underdiagnosed.48,49 This misclassification, if nondifferential, could have reduced the magnitude of the association seen in our study at this anatomic site. Alternatively, depression has been associated with illness behavior and increased utilization of health services, which could result in differential higher reporting in this group, potentially increasing ascertainment and the magnitude of association.50–53
As this is an observational study, although we adjusted for numerous covariates, there is always the potential for residual confounding as a result of imperfect measurement of variables or unmeasured covariates, especially in the area of confounding by indication of antidepressant use. We also used baseline exposure data, which does not account for depressive symptoms or antidepressant use after baseline.
Osteoporosis and resultant fractures are a growing public health concern. This sizable study adds to the body of knowledge regarding a common exposure, depression, and its influence on bone at a potentially vulnerable phase of life when the trajectory of bone metabolism is trending down. Thus, it is reassuring to observe minimal association of depressive symptoms or antidepressant therapy on 3-year changes in BMD in this group of generally healthy postmenopausal women. Our findings also provide additional evidence for an association between antidepressant medication and increased risk of fracture at some skeletal sites. Future randomized controlled trials of antidepressants should evaluate fractures as an outcome.
The authors thank Dr. Susan Ott, Department of Medicine University of Washington, for her review and suggestions on the manuscript.
Funding The WHI program is funded by the National Heart, Lung, and Blood Institute, U.S. Department of Health and Human Services. In addition, this analysis was funded in part from a Group Health Community Foundation grant.
WHI Investigators Program Office: (National Heart, Lung, and Blood Institute, Bethesda, Maryland) Elizabeth Nabel, Jacques Rossouw, Shari Ludlam, Linda Pottern, Joan McGowan, Leslie Ford, and Nancy Geller.Clinical Coordinating Center: (Fred Hutchinson Cancer Research Center, Seattle, WA)Ross Prentice, Garnet Anderson, Andrea LaCroix, Charles L. Kooperberg, Ruth E. Patterson, Anne McTiernan; (Wake Forest University School of Medicine, Winston-Salem, NC) Sally Shumaker; (Medical Research Labs, Highland Heights, KY) Evan Stein; (University of California at San Francisco, San Francisco, CA) Steven Cummings.Clinical Centers: (Albert Einstein College of Medicine, Bronx, NY) Sylvia Wassertheil-Smoller; (Baylor College of Medicine, Houston, TX) Jennifer Hays; (Brigham and Women’s Hospital, Harvard Medical School, Boston, MA) JoAnn Manson; (Brown University, Providence, RI) Annlouise R. Assaf; (Emory University, Atlanta, GA) Lawrence Phillips; (Fred Hutchinson Cancer Research Center, Seattle, WA) Shirley Beresford; (George Washington University Medical Center, Washington, DC) Judith Hsia; (Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance, CA) Rowan Chlebowski; (Kaiser Permanente Center for Health Research, Portland, OR) Evelyn Whitlock; (Kaiser Permanente Division of Research, Oakland, CA) Bette Caan; (Medical College of Wisconsin, Milwaukee, WI) Jane Morley Kotchen; (MedStar Research Institute/Howard University, Washington, DC) Barbara V. Howard; (Northwestern University, Chicago/Evanston, IL) Linda Van Horn; (Rush Medical Center, Chicago, IL) Henry Black; (Stanford Prevention Research Center, Stanford, CA) Marcia L. Stefanick; (State University of New York at Stony Brook, Stony Brook, NY) Dorothy Lane; (The Ohio State University, Columbus, OH) Rebecca Jackson; (University of Alabama at Birmingham, Birmingham, AL) Cora E. Lewis; (University of Arizona, Tucson/Phoenix, AZ) Tamsen Bassford; (University at Buffalo, Buffalo, NY) Jean Wactawski-Wende; (University of California at Davis, Sacramento, CA) John Robbins; (University of California at Irvine, CA) F. Allan Hubbell; (University of California at Los Angeles, Los Angeles, CA) Howard Judd; (University of California at San Diego, LaJolla/Chula Vista, CA) Robert D. Langer; (University of Cincinnati, Cincinnati, OH) Margery Gass; (University of Florida, Gainesville/Jacksonville, FL) Marian Limacher; (University of Hawaii, Honolulu, HI) David Curb; (University of Iowa, Iowa City/Davenport, IA) Robert Wallace; (University of Massachusetts/Fallon Clinic, Worcester, MA) Judith Ockene; (University of Medicine and Dentistry of New Jersey, Newark, NJ) Norman Lasser; (University of Miami, Miami, FL) Mary Jo O’Sullivan; (University of Minnesota, Minneapolis, MN) Karen Margolis; (University of Nevada, Reno, NV) Robert Brunner; (University of North Carolina, Chapel Hill, NC) Gerardo Heiss; (University of Pittsburgh, Pittsburgh, PA) Lewis Kuller; (University of Tennessee, Memphis, TN) Karen C. Johnson; (University of Texas Health Science Center, San Antonio, TX) Robert Brzyski; (University of Wisconsin, Madison, WI) Gloria E. Sarto; (Wake Forest University School of Medicine, Winston-Salem, NC) Denise Bonds; (Wayne State University School of Medicine/Hutzel Hospital, Detroit, MI) Susan Hendrix.
Conflict of interest statement None disclosed.