JAGS 49:732–736, 2001© 2001 by the American Geriatrics Society 0002-8614/01/$15.00

The Association of Bone Mineral Density and Depression in an Older Population

John Robbins, MD, MHS,* Calvin Hirsch, MD,* Rachel Whitmer, MS,* Jane Cauley, DrPH,

†

and Tamara Harris, MD, MS,

‡

for the Cardiovascular Health Study

OBJECTIVE:

To evaluate the association between bonemineral density (BMD) and measurements of depression inan older population.

DESIGN:

Population-based, cross-sectional study.

SETTING:

Study subjects were participants in the Car-diovascular Health Study (CHS), a longitudinal, long-term, follow-up study, at the University of CaliforniaDavis (Sacramento, California) and the University of Pitts-burgh (Pittsburgh, Pennsylvania) clinical centers.

PARTICIPANTS:

A random sample of 1,566 Medicareenrollees age 65 and older enrolled in the CHS.

MEASUREMENTS:

Total hip BMD, measured usingdual energy x-ray absorptiometry (DEXA), after adjust-ment for multiple covariates, was compared with depres-sion evaluated with the Center for Epidemiological Studies10-item Depression Scale (CES-Dm). Risk factors for os-teoporosis were compared in depressed and nondepressedparticipants. Potential correlates were entered into a re-gression model. Depression scores were compared in nor-mal, osteopenic, and osteoporotic individuals.

RESULTS:

Sixteen percent of participants were clinicallydepressed; 9% had BMDs in the osteoporotic range. MeanBMD was 40 mg/cm

2

lower in those with clinical depres-sion. High CES-Dm scores were associated with lowerBMD (

P

�

.001) when adjusted for body mass index(BMI), age, kilocalories of activity, estrogen use, gender,race, smoking and drinking. When stratified by race, thisremained true for all Caucasians (

P

�

.01), all AfricanAmericans (

P

�

.05), and when stratified by race and gen-der the association remained only for all Caucasian women(

P

�

.001). In women and Caucasian men there was an in-crease in depression scores among individuals with os-teoporotic-range BMDs.

CONCLUSIONS:

A significant association was foundbetween BMD and depressive symptoms after adjustmentfor osteoporosis risk factors. In Caucasians, depressivesymptoms were associated with both osteoporotic and os-teopenic levels of BMD. Causality cannot be ascribed,with only one measurement of BMD. We postulate thatthere may be an unmeasured third factor, such as an en-dogenous steroid, that is responsible for both low BMDand depression.

J Am Geriatr Soc 49:732–736, 2001.

Key words: human; older; depression; bone mineral density;

osteoporosis

A

s the population ages, osteoporosis is taking on majorimportance as a medical condition. With increasing

numbers of individuals living into their 80s and 90s, wecan expect to see an increase in the incidence of os-teoporotic fractures. More than 10 years ago, Riggs andMelton estimated that there were 1.2 million osteoporoticfractures in the United States annually.

1

As the populationages, demographics are driving the incidence steadily higher.A number of commonly accepted risk factors for osteo-porosis and fracture are mentioned in recent literature.

2–4

Of important note, depression is not listed among thesefactors and may not even have been considered. A recentlypublished paper from the Study of Osteoporotic FracturesGroup suggests that depression is a risk factor for fracturebut not for low bone mineral density (BMD).

5

There arevery limited data linking depression with osteoporosis.One case-control study compared 24 women hospitalizedfor depression with 24 matched controls and found lowermean BMD in the depressed patients.

6

There are also tworeports of decreased BMD in patients being treated formajor depression.

7,8

Liu et al. reported the association ofthe use of antidepressant medication and hip fracture.

9

This association may have been related to depression, notthe antidepressant medication use, but this hypothesis wasnot considered by the authors. However, that study makesuse of a case-control methodology without matching forBMD or depression. The authors suggest that the antide-pressants themselves, rather than the underlying depres-sion, were causally related to the hip fractures.

From the *Department of Medicine, University of California Davis, Sacra-mento, California;

†

University of Pittsburgh, Pittsburgh, Pennsylvania; and

‡

National Institute of Aging, Bethesda, Maryland.Supported by Contracts N01-HC-85079—N01-HC-85086 from the Na-tional Heart, Lung, and Blood Institute, and Georgetown Echo RC—HL 35129 JHU MRI RC–HL 15103.Address correspondence to John A. Robbins, MD, Division of General Medicine, Patient Support Services Building, 4150 V Street, Suite 2400, Sacramento, CA 95817.

JAGS JUNE 2001–VOL. 49, NO. 6

BMD AND DEPRESSION

733

In the current report, we tested the hypothesis that de-pression is associated with low BMD independent of otherrisk factors in a population-based sample of older menand women. The present investigation permits us to exam-ine further the relationship of depression to osteoporosisin a population-based cohort.

METHODS

The Cardiovascular Health Study (CHS) was initiated in1989–1990 with the recruitment of 5,201 individuals age65 to 100 randomly selected from Medicare enrollees infour geographic areas of the United States. Participantswere selected to match the age and gender distribution ofthe population. For a complete discussion of the recruit-ment of the population see the referenced articles describ-ing the study.

10,11

To better reflect the racial distribution ofthe older population of the United States, 687 AfricanAmericans were recruited using a similar methodology in1993–1994. Using the CHS data allowed us to include in-formation collected 2 years before the BMD measurement.This strengthens the potential to identify the possiblecausal relationship.

MEASURES

Dependent Variable, BMD

Two of the four CHS sites, the Pittsburgh, Pennsylvania,and Sacramento, California, centers, performed dual en-ergy x-ray absorptiometry (DEXA) on 1,591 participantsduring 1995–1996. (A complete data set was availableonly on 1,566 participants.) Total hip BMDs were ob-tained using Hologic QDR-2000 densitometers (Hologic,Inc., Waltham, MA) and read blindly at the University ofCalifornia San Francisco reading center using Hologicsoftware, version 7.10. BMD of the total hip was chosenas the most representative and reproducible measure ofBMD. The coefficient of variation for the total hip BMDwas

�

0.75%.

Independent Variables

Because factors causally related to the development of os-teoporosis should be measured when they have the poten-tial to influence bone formation or breakdown, we useddata obtained in the study year 1993–1994 to assess all in-dependent variables and correlated them with measures ofBMD obtained in the study year 1995–1996. The studyyear 1993–1994 was also the first year that data wereavailable on both cohorts.

Depression

All CHS participants were screened for depressive symp-toms with a modified 10-item version of the Center forEpidemiological Studies Depression Scale (CES-Dm).

12,13

Each item has a possible score of 0 to 3. The cut-off scorefor depression CES-Dm is

�

10, versus

�

20 in the full-length 20-item Center for Epidemiological Studies Depres-sion Scale (CES-D). The CES-Dm has demonstrated valid-ity as a screening tool for depression in older adults.

14

Energy Expenditure

Estimated kilocalories expended in 1 week were chosen asthe summary variable for physical activity. Kilocalories

were derived algorithmically using the validated physicalactivity questionnaire from the Minnesota Heart Survey.

15

Other Covariates

Alcohol consumption and smoking were based on self-reported information. Each was used as a continuous vari-able: drinks per week and pack-years of smoking. All par-ticipants were included in this analysis even if they did notsmoke or drink.

Current estrogen use in women was based on self-report and confirmed by reviewing the medications broughtto the clinic by the participants.

16

Current estrogen use wasused in the model because it was more likely to be a con-founder in the association of depression and BMD. (Thus,both depression and estrogen use were measured duringthe study year 1993–1994.) Past use of estrogens was lesslikely to have an effect on current depression. Past estro-gen use would tend to increase BMD but it would not bein the causal pathway between depression and BMD if es-trogen use had been discontinued. Body mass index (BMI)was calculated by dividing weight (kg) by height (metersquared) measured at the clinic visit. Use of oral steroidsand antidepressant medications as reported by the partici-pants were also investigated. The dietary information thatwas collected was not felt to be sensitive enough to be usedas a measure of calcium intake.

Statistical Methods

Our statistical analyses involved four modeling steps.First, in our exploratory analyses of the data, we used

t

-tests (two-tailed) to examine whether total hip BMD wassignificantly lower among those with clinical depression,defined as having a CES-Dm score

�

10.We stratified all subsequent analyses by race and gen-

der, because normal mean BMD varies between AfricanAmericans and Caucasian Americans, as well as betweenmen and women. Using a multiple linear regression model(Statistical Analysis System, version 6), we assessed whetherdepression was associated with total hip BMD, while con-trolling for possible confounders known to be associatedwith both BMD and depression: BMI, age, kilocalories ex-pended, current estrogen use, smoking, and reported alco-hol consumption in ounces per week. Some variables thatdid not reach significance when related to BMD in strati-fied analysis were retained in all the models because theywere significant for some groups.

After the models were developed, we made a number ofother calculations to try to add clinical meaning to the find-ings. First, we calculated the proportion of BMD explained bydepression in a post hoc fashion using the variables that en-tered the model for Caucasian women. To examine the cau-sality in the model more closely, we examined mean level ofCES-Dm score by normal, osteopenic, and osteoporotic lev-els of BMD in all four groups using a linear regression modeladjusting for age. This was done to give perspective to theresults, permit them to be displayed graphically, and deter-mine whether depression scores were elevated in those withsubclinical, nonsymptomatic levels of bone deterioration.Normal, osteopenic, and osteoporotic bone levels were foundby using the World Health Organization Criteria: normal(T-score

��

1), osteopenic (T-score

��

1and

��

2.5), osteo-porotic (T-score

��

2.5).

17

The T-score is defined as one

734

ROBBINS ET AL

. JUNE 2001–VOL. 49, NO. 6 JAGS

standard deviation in BMD compared with the gender-matched population at estimated peak BMD, age 29 to 30.We used the reference values provided by the Third NationalHealth and Nutrition Examination Survey (NHANES III).

18

RESULTS

Descriptive and Bivariate Analysis

The characteristics of the population are outlined in Table1. The means and standard deviations for each of the co-variates, stratified by race and gender, are shown. Sixteenpercent of the cohort met the criteria for clinical depres-sion, defined as a CES-Dm

�

10. Table 2 shows the resultsof bivariate analyses between participants with CES-Dm

�

10 and those with CES-Dm

�

10. The use of oral steroidor antidepressant medication was very small in this co-hort, 2% and 3%, respectively. Bivariate associations be-tween these medications and BMD could not be demon-strated (

P

�

.05). They are not included in Table 2 orfurther analysis. Age, BMD, alcohol consumption, andkilocalories expended were all significantly different be-tween the depressed and nondepressed groups. The clini-cally depressed group had an average BMD 40 mg/cm

2

lower than the nondepressed group.

Regression Model

The factors used in the stratified models with CES-Dm arelisted in Table 3, along with their parameter estimates (

�

)and

P

-values. The models are presented stratified for raceand gender. The model for “all participants” containsterms for gender and race. They are not included in thestratified models.

Depression was negatively associated with total hipBMD in the full cohort (

P

�

.001), and among Caucasians(

P

�

.01) and African Americans (

P

�

.05). After stratifi-cation by gender and race, a highly significant negative as-sociation between depressive symptoms and BMD was ob-served for Caucasian women (

P

�

.001), but the associationwas reduced to nonsignificant (

P

�

.05) for all others(Caucasian men

P

�

.9, African-American women

P

�

.6, African-American men

P

�

.2). In Caucasian women,every one-unit increase in CES-Dm score corresponded toa 0.3 mg/cm

2

decrease in total hip BMD. In an unadjustedmodel, depressive symptoms explained 13% of the ob-served variability in total hip BMD. After adjustment for

the covariants, depressive symptoms continued to explain2% of this variability.

Tricotomized BMD Versus Depression

In this older cohort, 25% of the men and approximately13% of the women met standardized criteria for os-teoporosis by having a T-score of

��

2.5 on total hipBMD. Figure 1 shows the mean CES-Dm score for normal,osteopenic, and osteoporotic individuals, adjusted for ageand stratified by race and gender. For Caucasian women,least squares means analysis revealed that the mean CES-Dm score for osteoporotic women was significantly higherthan for osteopenic and normal women (

P

�

.005 and

P

�

.02, respectively). This trend did not achieve statistical sig-nificance for Caucasian men or African Americans.

DISCUSSION

Prior work has demonstrated a relationship between de-pression and BMD in patients with major depression orthose receiving medical treatment for depression. To ourknowledge, this is the first demonstration of an associa-tion between BMD and a measure of depression in a popu-lation-based sample of normal older individuals. These re-sults demonstrate that multiple factors, including age,BMI, and caloric expenditure, are related to a measure ofdepressive symptoms. This is to be expected. A relativelyunique finding is that after adjusting for multiple acceptedrisk factors for osteoporosis, depressive symptoms, as as-sessed by the CES-Dm, showed a strong, inverse correla-tion with total hip BMD, measured 2 years later. Similarbut weaker associations were seen when the cohort wasstratified by gender and race; Caucasian women still showeda significant association. The cross-sectional design of thisstudy prevents the determination of a causal relationship be-tween depression and osteoporosis. However, the 2-year lagbetween the assessment of depression and the measure-ment of BMD is consistent with the hypothesis that de-pression may contribute to a decline in BMD. Depressionpreceded the measurement of BMD and thus may play acausal role in reducing bone density.

Estrogen therapy could easily affect both depressionand BMD and thus was included as a covariate in the mul-tivariate models. It has been postulated that there aremodest but persistently elevated plasma cortisol levels indepressed patients that may be sufficient to accelerate bone

Table 1. Descriptive Characteristics of Participants

Age Mean (SD)

BMI Mean (SD)

Pack/Years Mean (SD)

Drinks/WeekMean (SD)

Estrogen*% Using

KilocaloriesMean (SD)

CES-Dm Mean (SD)

Hip BMDgm/cm

2

Mean (SD)

All (n

�

1552) 74.32 (4.79) 26.86 (4.51) 17.54 (24.75) 2.88 (8.23) 18 1,589 (1,882) 5.29 (4.85) 0.83 (0.18)Caucasian (n

�

536) 75.19 (4.87) 26.47 (3.55) 23.47 (28.54) 3.92 (6.51) 2,097 (2,272) 4.43 (4.12) 0.92 (0.16)Caucasian (n

�

716) 74.35 (4.40) 26.40 (4.77) 14.69 (22.80) 2.07 (4.53) 20 1,413 (1,659) 5.58 (5.06) 0.72 (0.13)African-American

(n

�

113) 73.02 (5.11) 26.98 (4.03) 16.24 (21.13) 6.28 (23.91) 1,280 (1,403) 5.64 (4.73) 1.04 (0.17)African-American

(n

�

187) 72.52 (5.09) 29.70 (5.22) 12.22 (18.18) 0.77 (2.50) 11 997 (1,286) 6.41 (5.63) 0.85 (0.15)

*

Women only.CES-Dm

�

Center for Epidemiological Studies 10-item Depression Scale; SD

�

standard deviation; BMI

�

body mass index; BMD

�

bone mineral density.

JAGS JUNE 2001–VOL. 49, NO. 6

BMD AND DEPRESSION

735

loss.

6

Unfortunately, endogenous cortisol levels were notdetermined in the CHS. Exogenous corticosteroid use wasnot found to be significantly related to BMD in this cohort.

The association between depression and BMD ap-pears stronger for Caucasians than for African Americansand stronger for women than men. It is not surprising thatthe strongest association was found in Caucasian women,who had the highest prevalence of osteoporosis and, thus,the greatest power to show a statistically significant rela-tionship. Despite the smaller number of African Americansin the CHS cohort who had DEXA scans (n

�

297), andtheir relatively higher mean total hip BMD, African-Amer-ican participants showed the same association of depres-sive symptoms and BMD before the stratification by gen-der further decreased the sample size, and thus the power.This power problem is easily noted for BMD and estrogenuse where the betas are similar for African Americans andCaucasians, but the

P

-values are only significant for Cau-casians. The relationship of BMD to depression is likelynot to be limited only to Caucasian women, but larger sam-ple sizes will be required to demonstrate the association.

This study has a number of important limitations. It ispossible that the CES-Dm is not a comparable measure ofdepression in African Americans and Caucasians. AlthoughRoberts showed that the internal consistency and factorstructure for the CES-D was similar among African Amer-icans and Caucasians, other researchers have questionedthis.

19

The World Health Organization–defined cutpointsfor increased risk of fracture are based on data from Cau-casian women. Reference values for African Americans are

available from the Third National Health and NutritionExamination Survey (NHANES III), but it is unknownwhether fracture risk is equivalent in African Americansand Caucasians for a given T-score. Bone size and struc-ture may be different in African Americans, but DEXAmeasures “density” per cross-sectional area, not volume orstructure. T-score evaluations are included in the resultsbecause they are used in clinical practice but may be oflimited value in this context.

In this epidemiological observational study, 2% of thevariability in total hip BMD in Caucasian women age 65to 100 was explained by the CES-Dm. Additional studieswith more comprehensive depression inventories are neededto assess whether this association is robust. The addition ofmore extensive measures of depression to ongoing or futurestudies of osteoporosis should be straightforward and rela-tively inexpensive. Ideally, BMD should be measured at twopoints in time to see whether depressive symptoms are actu-ally associated with a greater fall in BMD when comparedwith nondepressed individuals. If the association holds trueusing other instruments and repeated measures, further in-vestigation of a causal link between depression and os-teoporosis will be warranted. These should include a searchfor a causal mechanism such as endogenous steroid produc-tion or measures of inflammation.

Participating Institutions and Principal StaffForsyth County, NC—Bowman Gray School of Medi-

cine of Wake Forest University:

Gregory L. Burke, SharonJackson, Alan Elster, Curt D. Furberg, Gerardo Heiss,

Table 2. Comparison of Variables Among Those Clinically Depressed (CES-Dm

�

10) and Nondepressed

AgeMean (SD)

BMI*Mean (SD)

Pack/YearsMean (SD)

Drinks/Week*Mean (SD)

Estrogen

†

*% Using

Kilocalories**Mean (SD)

Hip BMD*gm/cm

2

Mean (SD)

Depressed 74.87 (5.56) 27.70 (5.09) 19.10 (26.54) 2.13 (5.03) 13 1167 (1562) 0.79 (0.18)Nondepressed 74.21 (4.61) 26.70 (4.37) 17.22 (24.37) 3.02 (8.73) 20 1673 (1929) 0.83 (0.18)

*

P

�

.05;

**

P

� .001.†Women only.CES-Dm � Center for Epidemiological Studies 10-item Depression Scale; SD � standard deviation; BMI � body mass index; BMD � bone mineral density.

Table 3. Relationship of Depression and BMD mg/cm2, Adjusted for Covariates

Parameter Estimates (�s)

Women Men

Risk Factor All Caucasian African American Caucasian African American

CES-D �3.81*** �2.92** �0.79 �2.52 3.47BMI 13.98*** 13.37*** 11.86*** 15.61*** 17.85***Age �4.24*** �5.57*** �6.21** �5.11*** �4.18Kilocalories 0.01* 0.0008 �0.006 0.006* �0.009Estrogen† 62.4*** 62.0Pack/years 0.35* 0.01 �0.21 �0.20 �1.25Alcohol 2.27*** 2.56** 1.79 3.05** �0.73

*P � .05; **P � .01; ***P � .001.†Only used in models with women.BMD � bone mineral density; CES-D � Center for Epidemiological Studies Depression Scale; BMI � body mass index.

736 ROBBINS ET AL. JUNE 2001–VOL. 49, NO. 6 JAGS

Dalane Kitzman, Margie Lamb, David S. Lefkowitz, MaryF. Lyles, Cathy Nunn, Ward Riley, John Chen, BeverlyTucker; Forsyth County, NC—Wake Forest UniversityECG Reading Center: Farida Rautaharju, Pentti Rautaha-rju; Sacramento County, CA—University of California,Davis: William Bonekat, Charles Bernick, Michael Buono-core, Mary Haan, Calvin Hirsch, Lawrence Laslett, Mar-shall Lee, John Robbins, William Seavey, Richard White;Washington County, MD—The Johns Hopkins University:M. Jan Busby-Whitehead, Joyce Chabot, George W. Com-stock, Adrian Dobs, Linda P. Fried, Joel G. Hill, Steven J.Kittner, Shiriki Kumanyika, David Levine, Joao A. Lima,Neil R. Powe, Thomas R. Price, Jeff Williamson, MoysesSzklo, Melvyn Tockman; Washington County, MD—MRIReading Center, The Johns Hopkins University: NormanBeauchamp, R. Nick Bryan, Douglas Fellows, MelanieHawkins, Patrice Holtz, Naiyer Iman, Michael Kraut,Cynthia Quinn, Grace Lee, Carolyn C. Meltzer, LarrySchertz, Earl P. Steinberg, Scott Wells, Linda Wilkins,Nancy C. Yue; Allegheny County, PA—University ofPittsburgh: Diane G. Ives, Charles A. Jungreis, LaurieKnepper, Lewis H. Kuller, Elaine Meilahn, Peg Meyer,Roberta Moyer, Anne Newman, Richard Schulz, VivienneE. Smith, Sidney K. Wolfson; Echocardiography ReadingCenter (Baseline)—University of California, Irvine: HodaAnton-Culver, Julius M. Gardin, Margaret Knoll, TomKurosaki, Nathan Wong; Echocardiography Reading Cen-ter (Follow-Up)—Georgetown Medical Center: John Gott-diener, Eva Hausner, Stephen Kraus, Judy Gay, Sue Liven-good, Mary Ann Yohe, Retha Webb; Ultrasound ReadingCenter—New England Medical Center: Daniel H. O’Leary,Joseph F. Polak, Laurie Funk; Central Blood Analysis Lab-

oratory—University of Vermont: Elaine Cornell, MaryCushman, Russell P. Tracy; Pulmonary Reading Center—University of Arizona, Tucson: Paul Enright; Coordinat-ing Center—University of Washington, Seattle: Alice Ar-nold, Annette L. Fitzpatrick, Richard A. Kronmal, BruceM. Psaty, David S. Siscovick, Will Longstreth, Patricia W.Wahl, David Yanez, Paula Diehr, Corrine Dulberg, BonnieLind, Thomas Lumley, Ellen O’Meara, Jennifer Nelson,Charles Spiekerman; NHLBI Project Office: Robin Boineau,Teri A. Manolio, Peter J. Savage, Patricia Smith.

REFERENCES1. Riggs BL, Melton LJ. Involutional osteoporosis. N Engl J Med 1986;314:

1676–1686.2. R Eastell. Treatment of postmenopausal osteoporosis. N Engl J Med 1998;

338:736–746.3. PD Ross. Osteoporosis: Frequency, consequences, and risk factors. Arch In-

tern Med 1996;156:1399–1411.4. Cummings SR, Vevitt MC, Browner WS et al. Risk factors for hip fracture in

white women. N Engl J Med 1995;332:767–773.5. Whooley MA, Kip KE, Cauley JA et al. Depression, falls and risk of fracture

in older women. Arch Intern Med 1999;159:484–490.6. Michelson D, Stratakis C, Hill L et al. Bone mineral density in women with

depression. N Engl J Med 1996;335:1176–1181.7. Schweiger U, Deuschle M, Korner A et al. Low lumbar bone mineral density

in patients with major depression. Am J Psych 1994;151:1691–1693.8. Halbreich U, Rojansky N, Palter S et al. Decreased bone mineral density in

medicated psychiatric patients. Psychosomatic Med 1995;57:485–491.9. Liu B, Anderson G, Mittmann N et al. Use of selective serotonin-reuptake in-

hibitors of tricyclic antidepressants and risk of hip fractures in elderly people.Lancet 1998;351:1303–1307.

10. Tell GS, Fried LP, Hermanson B et al. Recruitment of adults 65 years andolder as participants in the Cardiovascular Health Study. Ann Epidemiol1993;3:358–366.

11. Fried LP, Borhani NO, Enright P et al. The Cardiovascular Health Study: De-sign and rationale. Ann Epidemiol 1991;1:263–276.

12. Radloff LS. The CES-D scale: A self-report depression scale for research inthe general population. Appl Psychologic Measure 1977;1:385–401.

13. Andresen EM, Malmgren JA, Carter WB et al. Screening for depression inwell older adults: Evaluation of a short form of the CES-D (Center for Epide-miologic Studies Depression Scale). Am J Prev Med 1994;10:77–84.

14. Lewinsohn PM, Seeley JR, Roberts RE et al. Center for Epidemiologic Stud-ies Depression Scale (CES-D) as a screening instrument for depression amongcommunity-residing older adults. Psychol Aging 1997;12:277–287.

15. Taylor H, Jacobs DJ, Schucker B et al. A questionnaire for the assessment ofleisure time physical activities. J Chronic Dis 1978;31:741–755.

16. Psaty BM, Lee M, Savage PJ et al. Assessing the use of medications in the el-derly: Methods and initial experience in the Cardiovascular Health Study. JClin Epidemiol 1992;45:683–692.

17. Kanis JA, Melton III LJ, Christiansen C et al. The diagnosis of osteoporosis. JBone Miner Res 1994;9:1137–1141

18. Looker AC, Wahner HW, Dunn WL et al. Proximal femur bone mineral lev-els of U.S. adults. Osteoporos Int 1995;5:389–409.

19. Callahan CM, Wolinsky FD. The effect of gender and race on the measure-ment properties of the CES-D in older adults. Med Care 1994;32:341–356.

Figure 1. Mean CES-Dm score versus total hip BMD (age ad-justed). CES-Dm � Center for Epidemiological Studies 10-itemDepression Scale; BMD � bone mineral density.