OBESITY AND DEPRESSION SYMPTOMS IN THE BEAVER DAM OFFSPRING STUDY POPULATION

Research Article

OBESITY AND DEPRESSION SYMPTOMS IN THE BEAVER

DAM OFFSPRING STUDY POPULATION

Wenjun Zhong, M.S.,



Karen J. Cruickshanks, Ph.D.,

Carla R. Schubert, M.S.,

F. Javier Nieto, M.D. Ph.D.,

Guan-Hua Huang, Ph.D.,

Barbara E.K. Klein, M.D. M.P.H.,

and Ronald Klein, M.D. M.P.H.

Background: Depression and obesity are both important public health problems.

However, it is not clear whether obesity contributes to depression. Our study aims

to evaluate the association between obesity and possible depression. Methods:

During the Beaver Dam Offspring Study examination, participants’ body

weight and height were measured with a Detecto 758C digital scale with height

bar, and depression symptoms were measured with the Center for

Epidemio-logical Studies-Depression (CES-D) Scale. Other relevant information, such as

demographic factors, lifestyle factors, comorbidities, and use of antidepressants,

was also collected during the examination. There were 2,641 participants

included in the analyses. Results: Obesity was associated with possible depression

measured by CES-D Scale (OR 5 1.6, 95% CI: 1.3–2.0) after controlling for

age and gender. The association remained similar after further adjustments.

Obesity was significantly associated with all four domains measured by CES-D

Scale after controlling for age and sex, with the largest effect on ‘‘Somatic

Complaints’’ domain (b .15, 95% CI: 0.0836–0.223). The association with

‘‘Interpersonal Difficulties’’ was not significant after further adjustments.

Conclusions: Obesity was associated with a higher risk of possible depression

and had different influences on specific domains of depression symptoms

measured by CES-D Scale. These findings suggest the need for longitudinal

studies on the effects of obesity on specific depression symptoms. Depression and

Anxiety 27:846–851, 2010.

Key words: obesity; depression; epidemiology

INTRODUCTION

D

epression symptoms are a common source of

distress and dysfunction, and have a great impact on

the quality of life. Major depressive disorder affects

approximately 14.8 million American adults, or about

6.7% of the US population, aged 18 and older in a

given year.

It is estimated that by the year 2020,

unipolar major depression will be the second leading

cause of global disease burden.

Obesity is another

important public health problem. Since the mid-1970s,

the prevalence obesity has increased sharply: data from

two National Health and Nutrition Examination

Survey (NHANES) showed that among adults aged

20–74 years, the prevalence of obesity increased from

15.0% (in the 1976–1980 survey) to 32.9% (in the

2003–2004 survey).

Obesity is a major risk factor for cardiovascular

disease, but its relation to psychological diseases

remains unclear. Data from 41,654 respondents in

the National Epidemiologic Survey on Alcohol and

Published online 28 January 2010 in Wiley Online Library (wiley onlinelibrary.com).

DOI 10.1002/da.20666

Received for publication 19 October 2009; Revised 18 December 2009; Accepted 18 December 2009

The authors disclose the following financial relationships within the past 3 years: Contract grant sponsor: National Institutes of Health; Contract grant number: AG021917.

Correspondence to: Wenjun Zhong, M.S., Population Health Sciences, University of Wisconsin-Madison, Madison, WI. E-mail: wzhong@wisc.edu

1_{University of Wisconsin-Madison, Madison, Wisconsin} 2

National Chiao Tung University, University Road, Hsinchu, Taiwan

r

Related Conditions showed that obesity defined by

self-reported height and weight was associated with

increased risk for any mood disorder, major depressive

disorder, and dysthymic disorder, in both men and

women (odds ratios ranged from 1.35 to 1.88).

Other

studies reported the positive association between

obesity and depression.

However, a recent

systema-tic review stated that the association between obesity

and depression was not consistent across studies;

_the

evidence from these studies was considered weak due to

quality issues, such as the use of self-reported body

mass index (BMI), invalid measurement of depression,

lack of description of the sampling process, low

response rate, loss of follow-up, and residual

con-founding. The aim of our study was to evaluate the

association between obesity and depression symptoms

in a large cohort, using standardized protocols to

measure and define obesity and depression.

MATERIALS AND METHODS

STUDY POPULATION

The Beaver Dam Offspring Study (BOSS) is a cohort study of adult children of participants in the Epidemiology of Hearing Loss Study (EHLS), which was designed to investigate sensory changes across generations and to provide important information on how genetic and changing environmental risk factors affect health. The description of the EHLS study can be found in earlier publica-tions.[10]_{In 2005, the adult children of the original EHLS population}

were invited to participate in the BOSS examination. Of the 4,965 offspring identified, 3,285 (66.2%) participated, 731 (14.7%) refused, 23 (0.5%) had died, and 926 (18.7%) failed to complete an examination or questionnaire. Participants were slightly older than nonparticipants (mean age 48 versus 46 years at the time of recruitment) and more likely to be women (54.6 versus 44.4%). More than 99% of the participants were reported as non-Hispanic white. This study was approved by the University of Wisconsin-Madison Health Sciences Institutional Review Board and all participants provided written informed consent.

MEASUREMENTS

The Center for Epidemiological Studies-Depression (CES-D) Scale was used to measure depression symptoms in our popula-tion.[11] The scale has been used for screening for depression in research and clinical settings.[12,13] It is composed of 20-item questions, which measure depression symptoms in four domains (factors): Depression Affect, Somatic Complaints/Activity Inhibition, Positive Affect, and Interpersonal Difficulties (Table 1). During the examination, participants were instructed to complete the CES-D form by indicating how often they experienced each symptom in the past week. The response is a four-point scale ranging from 0 to 3 indicating the frequency: ‘‘rarely or none of the time, oro1 day,’’ ‘‘some or little of the time, or 1–2 days,’’ ‘‘occasionally or a moderate amount of the time, or 3–4 days,’’ and ‘‘most of the time or 5–7 days,’’ except for questions 4, 8, 12, and 16, for which the scale is reversed. Higher total scores indicate worse depressive symptoms. For the four factors (domains), higher factor scores indicate worse depression symptoms other than the domain ‘‘Positive Affect.’’ A total score of 415 for the 20 questions is the usual cutoff for possible mild to major depression.

Participants were examined and interviewed by staff trained and certified in the study protocol. Body weight and height were measured with a Detecto 758C digital scale (Cardinal Scale Manufacturing Co., Webb City, MO) with height bar. The BMI was calculated as the ratio of body weight (kilogram) and square of height (meter). Overweight and obese is defined as BMI Z25 and Z30 kg/m2, respectively. Blood pressure was measured with a Dinamap Procare 100 (GE Medical Systems, Milwaukee, WI) after the participant rested at least 5 min. The measurement was taken three times at 1 min intervals and the average of the last two measurements was used in the analyses. Blood samples were collected from participants, and hemoglobin A1C was assessed at the Collaborative Studies Clinical Laboratory (Fairview-University Medical Center, Minneapolis, MN).

We selected potential confounders based on those reported in earlier studies.[9] _{Social Economic Status (SES) and demographic factors,}

including education levels (o12 years, 12 years, 13–15 years, and 161 years), family income, and marital status (married, single, and others) were included as well as comorbidites and lifestyle factors. Hyperten-sion was defined as a diagnosis of hypertenHyperten-sion and current antihypertensive medications or measured blood pressure Z140 mmHg (systolic) or Z90 mmHg (diastolic). Cardiovascular disease (CVD) was defined as self-reported doctor diagnosed myocardial infarction, stroke, or angina. Diabetes was defined as a self-report of doctor diagnosed diabetes or measured A1C46.1%. Sleep apnea was defined as a self-report of doctor diagnosed sleep apnea. Hearing was measured by audiometry and hearing loss was defined as pure tone average of 500, 1,000, 2,000 and 4,000 Hz425 dB in either ear.[14]_{Age-related Macular}

Degeneration (AMD) was assessed by digital retinal images and a standardized grading system.[15] _{We included these two sensory}

TABLE 1. CES-D questionnaire items and the

abbreviations

20 questions Abbreviation 1. I was bothered by things that don’t usually

bother me

Bother 2. I did not feel like eating, my appetite was poor Eat 3. I felt that I could not shake off the blues even

with the help of my family or friends

Blue 4. I felt that I was just as good as other people Good 5. I had trouble keeping my mind on what I

was doing

Focus 6. I felt depressed Down 7. I felt everything I did was an effort Effort 8. I felt hopeful about the future Hope 9. I thought my life had been a failure Fail 10. I felt fearful Fear 11. My sleep was restless Sleep 12. I was happy Happy 13. I talked less than usual Quiet 14. I felt lonely Alone 15. People were unfriendly Aloof 16. I enjoyed life Enjoy 17. I had crying spells Cry 18. I felt sad Sad 19. I felt that people dislike me Shun 20. I could not get ‘‘going’’ Drag Four factors are: Depression Affect (sad, cry, down, blue, alone, fear,

fail); Somatic Complaints (effort, drag, focus, sleep, bother, eat, quiet); Positive Affect (enjoy, good, happy hope); Interpersonal Difficulties (shun, aloof)

disorders because hearing loss and AMD are associated with poorer quality of life and increased prevalence of symptoms of depres-sion.[16–19] _{Lifestyle factors included smoking status (never/former/}

current smokers), history of heavy alcohol use (ever drank four or more alcoholic beverages daily), and exercise (regular weekly exercise sufficient to work up a sweat). Participants were asked to bring all their current medications to the examination to identify participants who currently used antidepressive medications.

STATISTICAL ANALYSIS

The association between obesity and possible depression (CES-D score415) was assessed with logistic regression; taking antidepres-sants was analyzed as an alternative outcome in a supplementary analysis. Because some confounders, such as physical activity, could also act as mediators, they were included in the models sequentially. Covariates that had significant P-values or changed the effect of obesity modestly were included in the final model. Subgroup analysis was performed among participants without CVD or diabetes as a sensitivity analysis. These analyses were performed with SAS 9.1 (SAS Institute, Cary, NC).

The multiple indicators, multiple causes (MIMIC) model was used to assess the associations of obesity with specific domains of the CES-D Scale, controlling for the important confounders. Before fitting the MIMIC model, the confirmatory factor analysis (CFA) was performed to check if the CES-D 4-domain structure was applicable in our population. The MIMIC model was evaluated as good fit by checking the fit indicators[20,21]_{: (1) a comparative fit index (CFI)}

value greater than .9; (2) a Tucker–Lewis (TLI) index value greater than .9; (3) a root mean square error of approximation (RMSEA) value close to .06 or less; (4) a weighted root mean square residual (WRMR) values close to 1.0 or lower. The statistical significance of the factor loadings, and that the residual variances did not take negative values for any of the items, were also taken into account when evaluating the model fit. These analyses were conducted with Mplus Version 4.21.[22]

RESULTS

A total of 2,641 participants out of the total BOSS

population with complete CES-D and BMI data were

included in the analyses. The age range was 21–84

years (M: 49.2 years, SD: 9.9 years) where 45.1% were

male. The BMI range was 17.4–61.6 kg/m

(M:

30.1 kg/m

_{, SD: 6.6 kg/m}

_{). These people were similar}

to the BOSS population in terms of mean age (49.2

versus 48.7 years), male percentage (45.1 versus

45.4%), mean BMI (30.1 versus 30.2 kg/m

),

percent-age taking antidepressants (16.1 versus 15.9%) and

other aspects, such as SES factors and comorbid

conditions.

The range of the CES-D total score was 0–52

(M: 8.4, SD: 7.4). According to the cutoff, 14.1%

participants were considered to have possible mild

to major depression. The percent of overweight

and obese participants were 34 and 44%. Generally

speaking, comparing to those normal weight people,

they were older, more often men, had worse SES, and

were more likely to be smokers and report a history of

heavy drinking, and have more comorbid conditions

(Table 2).

Obesity was significantly associated with the possible

depression in sequentially adjusted models (Table 3).

After adjusting for important confounders, obese

participants had an OR of 1.5 (95% CI: 1.1–1.9,

P 5.002) to have possible depression compared to

those non-obese people (model 5). Furthermore,

adjusting for use of antidepressants did change the

result (data not shown). In the sensitivity analysis with

participants free of CVD and diabetes, the results were

similar (OR: 1.4, 95% CI: 1.1–1.8, P 5.01). When

antidepressants was used as an alternative outcome,

obesity was associated with a higher OR of taking

antidepressants (OR: 1.7, 95% CI: 1.3–2.1, Po.0001).

In addition, models using BMI as a continuous variable

showed similar results and each 5-unit increase in BMI

was associated with an OR of 1.1 (95% CI: 1.0–1.20,

P 5.003) for possible depression after adjusting for

important confounders.

Our CFA results showed that Radloff’s original

4-domain structure was applicable in our population.

The model fit indicators were: RMSEA 5 .055,

CFI 5 .93, TLI 5 .98, WRMR 5 2.1. The factors

loading were high (standardized factor loadings ranged

from .508–.956) and all were statistically significant. In

the MIMIC models, obesity had significant effects on

all four domains, with the strongest effect on somatic

domain after adjusting for age and sex: obese

partici-pants had an average .15 (95% CI: 0.09–0.22) units

higher score in the domain of ‘‘Somatic Complaints’’

(Table 4). When taking the SES factors into account,

the effect on the ‘‘Interpersonal Difficulties’’ was no

longer significant; the effects were similar when further

adjusting for other confounders.

DISCUSSION

Our data support that there is a positive association

between obesity and depression symptoms. Although

the effect size was small (OR: 1.4–1.6; Table 3), the

association was consistent in different analyses. It was

robust in sequential models with different confounders

and was consistent in the subgroup analysis restricted

to participants free of CVD and diabetes. A similar

association was present in analyses with use of

antidepressants as the outcome. The magnitude of this

association was consistent with some earlier studies,

which suggested that the ORs ranged from 1.4 to 1.9

for different psychiatric disorders.

One study

suggested that there might be a U-shape association

between BMI and depression

: being underweight or

overweight were both associated with depression

compared to normal weight. In our population, only

nine

(0.32%)

participants

were

underweight

(BMI

o18.5 kg/m

_{), so we were unable to examine}

the possible U-shape association.

The 4-factor structure of CES-D scale was originally

developed in the Caucasian population.

After that,

although varied CES-D factor structures were found in

different populations, such as different ethnic/cultural

groups, diseased groups,

the scale has been shown

to be suitable for many different groups.

We found

that the original 4-factor structure of the CES-D Scale

was applicable in a relatively heavier population: the

prevalence of overweight (including obesity) and

obesity in our study population were 78 and 44%,

respectively, whereas they were 66.3 and 32.2% among

US adults aged 20 years and over in 2003–2004,

according to the NHANES data (CDC website: http://

www. cdc. gov / nchs / products/pubs/pubd/hestats/over

weight/overwght_adult_03.htm).

Although our study population was relatively heavier

than the national population, they did not have higher

CES-D scores or a higher prevalence of possible

depression. The mean CES-D score (8.4) and the

prevalence of possible depression (14.1%) were similar

to those reported in the original study of the CES-D

Scale,

in which the mean score ranged from 7.53 to

8.58 and the prevalence of possible depression ranged

from 15 to 19%. Other factors may have protected

them from having worse depression symptoms. In our

study, obesity is not the only factor associated with

depression; other factors, such as education, income,

and life styles, had important effects on depression

symptoms. Thus, factors such as better educational

levels (97.5% were high school graduates and 68.5%

had at least some college education) and good, regular

TABLE 3. ORs for the association between obesity

(BMIZ30 kg/m

) and possible depression (dependent

variable) in multivariable adjusted models

OR 95% CI P-value Model 1 1.6 1.3–2.0 o.0001 Model 2 1.5 1.2–1.9 .001 Model 3 1.4 1.1–1.8 .005 Model 4 1.4 1.1–1.8 .01 Model 5 1.5 1.1–1.9 .002 Model 1: age and gender adjusted; Model 2: further adjusted for demographics: marital status, family income, and education levels; Model 3: further adjusted for chronic diseases: CVD, diabetes, hypertension and sleep apnea, hearing loss and AMD; Model 4: further adjusted for life style factors 5 smoking and drinking habits, regular exercises; Model 5: age sex, education level, family income, marital status, CVD, sleep apnea, hearing loss and AMD, regular exercises adjusted.

TABLE 2. The study population according to BMI categories

Normal BMI (n 5 581) Overweight (n 5 887) Obese (n 5 1173)

Mean (SD) or % Mean (SD) or % Mean (SD) or % P Age (years) 45.7 (9.1) 48.6 (9.9) 50.6 (9.6) o.0001

Male 24.8 52 50.5 o.0001

BMI (kg/m2) categories 22.7 (1.6) 27.4 (1.4) 35.8 (5.4) NA CES-D score 7.8 (7.2) 7.7 (6.9) 9.1 (7.7) o.0001 Education levels o12 years 2.3 2.5 2.7 12 years 24.8 28.5 31.5 13–15 years 29.3 30.9 37.9 161 years 31.4 38.1 28.0 o.0001 Family income ($1,000) o40 19.5 20.7 24.4 o60 21.5 21.5 23.7 o75 17.9 18.7 20 o100 17 14.6 16.2 1001 24 24.4 15.7 o.0001 Marital status Married 73.3 74.3 72.2 Single 17.2 14.5 14.8 Others 9.5 11.2 13.0 .1 Smoking status Current smoker 21.2 18.4 32.0 Past smoker 23.1 27.0 15.1 .0003

Heavy drinking habit (%) 14.5 16.5 20.9 .001 Regular exercise habit (%) 68.5 64.1 55.1 o.0001

CVD prevalence 1.4 2.4 4.7 .0003

Diabetes prevalence 0.9 3.4 13.4 o.0001 Hypertension prevalence 15.0 29.3 51.7 o.0001 Diagnosed sleep apnea % 1.2 1.6 8.4 o.0001 Hearing impairment (%) 8.6 13.8 15.7 .0002 Any macular degeneration % 2.4 3.9 3.6 .3 Antidepressant medication 12.7 13.8 19.4 .0001

exercise habit (61% participants had regular weekly

exercise) may have helped offset the adverse effect of

obesity on depression symptoms.

By breaking down the depression symptoms

mea-sured by CES-D Scale into the four domains, we can

better understand how obesity may affect different

aspects of depression. Adjusting for age and sex, obesity

had significant influences on all four domains of

self-reported depression symptoms, with the largest effect

on the ‘‘Somatic Complaints.’’ This indicates that the

obese participants had more complaints on

psycho-somatic problems, which may be due to the

obesity-associated diseases and malfunction. After further

adjusting for SES factors and other factors, the effect

sizes of obesity were decreased a little bit, but were still

significant for all domains except for the ‘‘Interpersonal

Difficulties’’ domain. This suggests that obese people

may feel worse in the aspects of ‘‘Depression Affect,’’

‘‘Somatic Complaints,’’ and ‘‘Positive Affect,’’ but being

obese does not limit individuals from enjoying good

personal relationships when other conditions (such as

SES factors) were taken into consideration.

The limitations of our study include that we cannot

determine the direction of the association due to its

cross-sectional design. Second, we may have missed

some confounders in our analyses. For example,

dementia may confound the association between

obesity and depression, especially among elderly

populations.

In our examination, the Mini-Mental

State Examination was administered to participants

older than 50 years; we only found three possible cases

of impaired cognition. Therefore, dementia is not

likely to be a confounder in our case. Finally, selection

bias may be of concern. Of the BOSS population, 81%

had completed the CES-D and included in our

analyses, and they were similar to the BOSS population

in terms of age, male percentage, mean BMI, and other

demographic factors. Thus, we consider that our

sample population was representative of the total

population and the likelihood of introducing selection

bias is low.

It is also important to realize that over-adjustment

may exist in the multivariable adjustment models

because some risk factors may be part of the causal

pathway. For example, less exercise may be a cause for

weight gain as well as a result from weight gain. When

adjusting for exercise, the effect size of BMI may be

incorrectly diluted. Thus, we performed the analyses

by sequentially adding confounders to the models.

Strengths of our study include: (1) a relatively large

sample size and a relatively heavier population, which

enable us to better evaluate the association of obesity

and depression; (2) a valid measurement of depression

symptoms as the CES-D Scale was tested and proved

to be applicable in our population; (3) the breakdown

of the depression symptoms into different domains for

analyses; (4) a comprehensive adjustment of potential

confounders, including sensory impairments, which

have been rarely investigated in earlier studies; and (5)

consistent results from different analyses.

CONCLUSION

Our study found a consistent association between

obesity and depression symptoms in a slightly heavier,

midlife population. Obesity had significant effects on

the domains of ‘‘Depressed Affect,’’ ‘‘Somatic

Com-plaints,’’ and ‘‘Positive Affects,’’ but not ‘‘Interpersonal

Difficulties’’ measured by the CES-D Scale after

including SES factors in the model. These results

suggest that obesity may contribute to depression

symptoms and may have different effects on various

aspects of depression symptoms. Longitudinal studies

are needed to determine the effect of obesity on specific

depression symptoms.

Acknowledgments. Funding for this research was

supported by National Institutes of Health AG021917

(K. J. C.).

REFERENCES

1. Kessler RC, Chiu WT, Demler O, Walters EE. Prevalence, severity, and comorbidity of twelve-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry 2005;62:617–627.

2. Murray CJ, Lopez AD. Evidence-based health policy—Lessons from the Global Burden of Disease Study. Science 1996;274: 740–743.

3. Ogden CL, Carroll MD, McDowell MA, Flegal KM. Obesity among Adults in the United States—No Change since

TABLE 4. Effects of obesity on the 4-domains of depression symptoms measured by CES-D scale: regression

coefficients from the MIMIC models

Depression affect Somatic complaints Positive affect Interpersonal difficulties Estimate 95% CI Estimate 95% CI Estimate 95% CI Estimate 95% CI Model 1 0.11 0.038 0.187 0.15 0.086 0.223 0.10 0.15 0.041 0.11 0.023 0.186 Model 2 0.09 0.015 0.167 0.12 0.051 0.193 0.07 0.125 0.017 0.07 0.015 0.153 Model 3 0.08 0.002 0.158 0.10 0.029 0.172 0.06 0.117 0.007 0.07 0.020 0.152 Model 1: age sex adjusted; Model 2: age sex, education level, family income, marital status adjusted; Model 3: Age sex, education level, family income, marital status, CVD, sleep apnea, hearing loss and AMD, regular exercises adjusted.

2003–2004. NCHS Data Brief No 1. Hyattsville, MD: National Center for Health Statistics; 2007.

4. Petry NM, Barry D, Pietrzak RH, Wagner JA. Overweight and obesity are associated with psychiatric disorders: results from the National Epidemiologic Survey on alcohol and related condi-tions. Psychosom Med 2008;70:288–297.

5. Heo M, Pietrobelli A, Fontaine KR, et al. Depressive mood and obesity in US adults: comparison and moderation by sex, age, and race. Int J Obes (Lond) 2006;30:513–519.

6. Herva A, Laitinen J, Miettunen J, et al. Obesity and depression: results from the longitudinal Northern Finland 1966 Birth Cohort Study. Int J Obes (Lond) 2006;30:520–527.

7. Ross CE. Overweight and depression. J Health Soc Behav 1994;35:63–79.

8. Johnston E, Johnson S, McLeod P, Johnston M. The relation of body mass index to depressive symptoms. Can J Public Health 2004;95:179–183.

9. Atlantis E, Baker M. Obesity effects on depression: systematic review of epidemiological studies. Int J Obes (Lond) 2008;32:881–891. 10. Cruickshanks KJ, Tweed TS, Wiley TL, et al. The 5-year

incidence and progression of hearing loss: the epidemiology of hearing loss study. Arch Otolaryngol Head Neck Surg 2003; 129:1041–1046.

11. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977; 1:385–401.

12. Zich JM, Attkisson CC, Greenfield TK. Screening for depression in primary care clinics: the CES-D and the BDI. Int J Psychiatry Med 1990;20:259–277.

13. Roberts RE, Vernon SW. The Center for Epidemiologic Studies Depression Scale: its use in a community sample. Am J Psychiatry 1983;140:41–46.

14. Zhan W, Cruickshanks KJ, Klein BEK, et al. Generational differences in the prevalence of hearing impairment in adults. Am J Epidemiol 2010;171:260–266.

15. Klein R, Klein BE, Linton KL. Prevalence of age-related maculopathy. The Beaver Dam Eye Study. Ophthalmology 1992; 99:933–943.

16. Dalton DS, Cruickshanks KJ, Klein BE, et al. The impact of hearing loss on quality of life in older adults. Gerontologist 2003;43:661–668.

17. Arlinger S. Negative consequences of uncorrected hearing loss: a review. Int J Audiol 2003;42:2S17–2S20.

18. Beaudreau SA, O’Hara R. Late-life anxiety and cognitive impairment: a review. Am J Geriatr Psychiatry 2008;16: 790–803.

19. Rovner BW, Casten RJ, Hegel MT, et al. Preventing depression in age-related macular degeneration. Arch Gen Psychiatry 2007; 64:886–892.

20. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling 1999;6:1–55.

21. Yu CY. Evaluating cutoff criteria of model fit indices for latent variable models with binary and continuous outcomes. PhD Dissertation. University of California at Los Angeles, Graduate School of Education and Information Studies; 2002. Website: [http://www.statmodel.com/download/Yudissertation.pdf]. 22. Muthen LK, Muthen BO. Mplus User’s Guide. Los Angeles:

Muthen & Muthen; 2004.

23. de Wit LM, van Straten A, van Herten M, et al. Depression and body mass index, a u-shaped association. BMC Public Health 2009;9:14.

24. Manson SM, Ackerson LM, Dick RW, et al. Depressive symptoms among American Indian adolescents: psychometric characteristics of the Center for Epidemiologic Studies Depres-sion scale (CES-D). Psychol Assess 1990;2:231–237.

25. Schroevers MJ, Sanderman R, Sonderen E, Ranchor AV. The evaluation of the Center for Epidemiologic Studies Depression (CES-D) scale: depressed and positive effect in cancer patients and health reference subjects. Qual Life Res 2000;9:1015–1029.

26. Shafer AB. Meta-analysis of the factor structures of four depression questionnaires: Beck, CES-D, Hamilton, and Zung. J Clin Psychol 2006;62:123–146.

27. Beaudreau SA, O’Hara R. Late-life anxiety and cognitive impairment: a review. Am J Geriatr Psychiatry 2008;16:790–803.