Brain, Behavior, and Immunity 36 (2014) 176–182

Contents lists available at ScienceDirect

Brain, Behavior, and Immunity

journal homepage: www.elsevier .com/locate /ybrbi

Phobic anxiety symptom scores and incidence of type 2 diabetes in USmen and women q

0889-1591/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.bbi.2013.10.025

q Funding sources had no role in the design or analysis of the study or thedecision to submit the manuscript for publication.⇑ Corresponding author.

E-mail address: mfarvid@hsph.harvard.edu (M.S. Farvid).

Maryam S. Farvid a,⇑, Lu Qi a,d, Frank B. Hu a,b,d, Ichiro Kawachi c, Olivia I. Okereke b,d,e, Laura D. Kubzansky c,Walter C. Willett a,b,d

a Department of Nutrition, Harvard School of Public Health, Boston, MA, United Statesb Department of Epidemiology, Harvard School of Public Health, Boston, MA, United Statesc Department of Social and Behavioral Sciences, Harvard School of Public Health, Boston, MA, United Statesd Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, United Statese Department of Psychiatry, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, United States

a r t i c l e i n f o

Article history:Received 6 August 2013Received in revised form 24 October 2013Accepted 25 October 2013Available online 31 October 2013

Keywords:Phobic anxietyType 2 diabetes

a b s t r a c t

Context: Emotional stress may be a risk factor for type 2 diabetes (T2D), but the relation between phobicanxiety symptoms and risk of T2D is uncertain.Objective: To evaluate prospectively the association between phobic anxiety symptoms and incident T2Din three cohorts of US men and women.Design, setting and patients: We followed 30,791 men in the Health Professional’s Follow-Up Study (HPFS)(1988–2008), 68,904 women in the Nurses’ Health Study (NHS) (1988–2008), and 79,960 women in theNurses’ Health Study II (NHS II) (1993–2011). Phobic anxiety symptom scores, as measured by theCrown–Crisp index (CCI), calculated from 8 questions, were administered at baseline and updated in2004 for NHS, in 2005 for NHS II, and in 2000 for HPFS. Incident T2D was confirmed by a validated sup-plementary questionnaire. We used Cox proportional hazards analysis to evaluate associations with inci-dent T2D.Results: During 3,099,651 person-years of follow-up, we documented 12,831 incident T2D cases. In mul-tivariate Cox proportional-hazards models with adjustment for major lifestyle and dietary risk factors,the hazard ratios (HRs) of T2D across categories of increasing levels of CCI (scores = 2 to <3, 3 to <4, 4to <6, P6), compared with a score of <2, were increased significantly by 6%, 10%, 10% and 13%(Ptrend = 0.001) for NHS; and by 19%, 11%, 21%, and 29% (Ptrend < 0.0001) for NHS II. Each score incrementin CCI was associated with 2% higher risk of T2D in NHS (HRs, 1.02, 95% confidence intervals: 1.01–1.03)and 4% higher risk of T2D in NHS II (HRs, 1.04, 95% confidence intervals: 1.02–1.05). Further adjustmentfor depression did not change the results. In HPFS, the association between CCI and T2D was not signif-icant after adjusting for lifestyle variables.Conclusion: Our results suggest that higher phobic anxiety symptoms are associated with an increasedrisk of T2D in women.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction

The prevalence of type 2 diabetes (T2D) is increasing at alarm-ing rates in the US and worldwide (Chen et al., 2011; Zabetianet al., 2013). In addition to well-known diabetic risk factors suchas diet, obesity, physical inactivity, age, race, and a family historyof T2D (Psaltopoulou et al., 2010; Venables and Jeukendrup,2009), recent studies have suggested a role of emotional stress inthe etiology of T2D (Kato et al., 2009; Pouwer et al., 2010; Rod

et al., 2009). The epidemiological studies support the concept thatdifferent forms of emotional stress, particularly depression, generalemotional stress, anxiety, anger, hostility and sleeping problems(Pouwer et al., 2010), contribute to an elevated risk of T2D. Anxietydisorders are the most prevalent mental disorders and lifetimeprevalence of either specific phobia or social phobia is over 12%in the U.S. (Kessler et al., 2005a,b).

Emotional stress may influence behavioral factors and therebyincrease the risk of T2D through unhealthy dietary intake, exces-sive alcohol consumption, smoking and low exercise levels (Rodet al., 2009; Kye and Park, 2012; Bonnet et al., 2005). Additionalevidence also suggests the association between phobic anxietysymptoms and increasing inflammatory biomarkers such as C-reactive protein, tumor necrosis factor a, leptin, soluble E-selectin

M.S. Farvid et al. / Brain, Behavior, and Immunity 36 (2014) 176–182 177

and soluble intercellular adhesion molecule (Brennan et al., 2009;Pitsavos et al., 2006), which are well-known risk factors for T2D(Mirza et al., 2012). Importantly, phobic anxiety is treatable; thus,any potential impacts on T2D incidence may be amendablethrough early identification and intervention.

An association between phobic anxiety symptoms and in-creased risk of coronary heart disease (CHD) in men and womenhas been previously reported in our and other cohorts (Albertet al., 2005; Kawachi et al., 1994; Watkins et al., 2010), to date,however, the relationship between phobic anxiety symptoms andT2D incidence has not been directly examined. Therefore, usingdata from three prospective cohorts, the Nurses’ Health Study(NHS), Nurses’ Health Study II (NHS II) and Health Professional’sFollow-up Study (HPFS), we examined the association betweenphobic anxiety symptom scores, as measured by Crown–Crispindex (CCI), and T2D incidence in women and men.

2. Research design and methods

2.1. Study population

We used data from 3 prospective cohort studies: NHS (startedin 1976; n = 121,704; age ranged at baseline: 30–55 years; enrolledfrom 11 US states), NHS II (established in 1989; n = 116,643; ageranged at baseline: 24–43 years; enrolled from 14 US states) andHPFS (initiated in 1986; n = 51,529; age ranged at baseline: 40–75 years; enrolled from 50 US states). In all the 3 cohorts, question-naires were administered at baseline and biennially thereafter tocollect and update information on lifestyle practices and occur-rence of chronic diseases. Information on phobic anxiety symp-toms was first obtained on the 1988 questionnaire in NHS(n = 103,614), on the 1993 questionnaire in NHS II (n = 87,238)and on the 1988 questionnaire in HPFS (n = 48,834); this servedas the baseline populations for our analyses. Participants were ex-cluded if they had T2D, cancer, CHD or stroke at baseline(n = 16,255 in NHS, n = 5935 in NHS II and n = 7354 in HPFS), miss-ing information on T2D diagnosis date (n = 3355 in NHS, n = 937 inNHS II, and n = 1524 in HPFS), age (n = 48 in NHS and n = 182 inNHS II), or phobic anxiety symptom score data (n = 15,052 inNHS, n = 224 in NHS II, and n = 9165 in HPFS). After exclusions,data from 68,904 women in NHS, 79,960 women in NHS II and30,791 men in HPFS were available for the analysis. The study pro-tocol was approved by the institutional review boards of Brighamand Women’s Hospital and Harvard School of Public Health (Bos-ton, Massachusetts, United States).

2.2. Assessment of T2D

In the three cohorts, participants who reported a new diagnosisof T2D on any of the biennial questionnaires were sent supplemen-tary questionnaires regarding symptoms, diagnostic tests, andhypoglycemic therapy. A case of T2D was confirmed if at leastone of the following items was reported on the supplementaryquestionnaire according to the National Diabetes Data Group crite-ria (National Diabetes Data Group, 1979): (1) one or more classicsymptoms (excessive thirst, polyuria or frequent urination, weightloss, or hunger) plus fasting plasma glucose concentrationsP140 mg/dl or random plasma glucose concentrations P200 mg/dl; (2) P2 elevated plasma glucose concentrations on differentoccasions (fasting concentrations P140 mg/dl, random plasmaglucose concentrations P200 mg/dl, and/or concentrationsP200 mg/dl after P2 h shown by oral glucose tolerance testing)in the absence of symptoms; or (3) treatment with hypoglycemicmedication (insulin or oral hypoglycemic agent). For cases diag-nosed in 1998 and later, the fasting plasma glucose threshold

was lowered to 126 mg/dl according to the American DiabetesAssociation criteria (Gavin et al., 1997). The validity of the supple-mentary questionnaire has been demonstrated previously in theNHS (Manson et al., 1991; Field et al., 2001). Only cases confirmedby the supplemental questionnaires were included in the currentanalysis.

2.3. Assessment of phobic anxiety symptom scores: the Crown–Crispindex (CCI)

The phobic anxiety scale of the CCI measures personality symp-toms of phobic anxiety (Crown and Crisp, 1966). It is a standard-ized, self-rating inventory of 8 questions on common symptomsof phobic anxiety with two (0, 2) to three (0, 1, 2) levels of possibleresponses to each question. The scores range from 0 to 16, withhigher scores given to higher levels of phobic anxiety symptoms(Crown and Crisp, 1966). The CCI has been validated in psychiatricoutpatient clinic settings and it could discriminate patients withanxiety disorders and agoraphobia from healthy controls and pa-tients with other forms of psychopathology (e.g., depressive, obses-sive–compulsive) (Crown and Crisp, 1966; Mavissakalian andMichelson, 1981). The validity of the phobic anxiety scale of CCIin the NHS population has been tested through assessing the asso-ciation with tranquilizer medication (Albert et al., 2005). For thosewith missing data on 1 or 2 questions, the total score was dividedby the fraction of questions answered and then rounded to thenearest whole number. The CCI was completed by participants inthe NHS in 1988, in NHS II in 1993 and HPFS in 1988 and updatedin 2004 for NHS, in 2005 for NHSII, and in 2000 for HPFS. TheSpearman correlation between CCI scores in 1988 and 2004(NHS), 1993 and 2005 (NHS II) and 1988 and 2000 (HPFS) werehigh (r = 0.75, p < 0.0001 in NHS, r = 0.70, p < 0.0001 in NHS II,and r = 0.75, p < 0.0001 in HPFS), an indication that scores reliablyrepresent long-term phobic anxiety symptom levels.

2.4. Assessment of covariates

In the biennial follow-up questionnaires, we inquired and up-dated information on risk factors for chronic diseases, such as bodyweight, cigarette smoking, physical activity, aspirin use, multivita-min use, family history of diabetes, menopausal status and post-menopausal hormone use (NHS and NHS II only), marital statusand routine screening for physical exam. Dietary information(including alcohol) was assessed using a validated semi-quantita-tive food frequency questionnaire every 4 year starting from1986 (NHS), 1991 (NHS II) and 1986 (HPFS) (Feskanich et al., 1993).

We calculated a diabetes dietary score composed of quintilevalues of polyunsaturated fat to saturated fat ratio, trans fat (in-verted), cereal fiber, and glycemic load (inverted) by standardizingand summarizing the respective continuously scaled dietary vari-ables. This method was described in detail elsewhere (Hu et al.,2001). In women, we defined depression status (yes/no) as having:Mental Health Index-5 652 [from the Medical Outcomes ShortForm-36, scaled from 0 to 100; higher scores indicate better men-tal health; included on questionnaires in 1992, 1996, and 2000 inthe NHS and in 1993, 1997 and 2001 in the NHS II], self-reportedregular antidepressant use, and/or self-reported physician-diag-nosed depression. In men, we defined depression status (yes/no)based on self-reported regular antidepressant use and/or self-reported physician-diagnosed depression.

2.5. Statistical analyses

Person-years for each participant were calculated from the re-turn date of the phobic anxiety questionnaire (1988 in NHS,1993 in NHS II and 1988 in HPFS) to the date of diagnosis of

178 M.S. Farvid et al. / Brain, Behavior, and Immunity 36 (2014) 176–182

T2D, death, loss to follow-up, or the end of the follow-up period(June 1, 2008 for NHS, June 1, 2011 for NHS II and January 30,2008 for HPFS), which came first. The CCI score was calculatedand categorized into 5 groups as follows:<2 (the reference group);2 to <3; 3 to <4; 4 to <6; and 6 or above (the highest phobic anxietysymptoms group) (Okereke and Grodstein, 2012). We used Coxproportional-hazards models to estimate the hazard ratios (HRs)and 95% confidence intervals (CI) of developing T2D across the 5categories of the CCI; the comparison group was participants withCCI <2. In addition to adjustment for age, we further adjusted forthe following covariates: Model 1: race, marital status, husbandeducation (in NHS and NHS II), family history of diabetes, currentaspirin use, menopausal status and hormone use (in NHS andNHS II), smoking status, alcohol intake, physical activity, and Mod-el 2: further adjustment for dietary factors (energy intake, diabetesdietary score and coffee). In additional analyses, we further ad-justed for body mass index (BMI) to examine the degree to whichthe association between T2D and CCI was mediated by BMI (Model3). Due to known high degree of association between anxiety anddepression (Byers et al., 2010) and the relationship betweendepression and T2D incidence (Arroyo et al., 2004), we also con-ducted a secondary analysis additionally adjusted for depressionstatus. Because phobic anxiety symptoms might affect physicalexam screening, we conducted an additional analysis in whichwe adjusted for routine screening for physical exam.

To better represent the long-term CCI scores and to minimizewithin-person variation, we created cumulative averages of CCIscores from baseline to second measurement. We stopped updat-ing the CCI scores when the participants reported a diagnosis ofstroke, CHD and cancer that might lead to changes in CCI scores.In sensitivity analyses, however, we also examined cumulativeaverage of CCI scores that was continually updated even after thedevelopment of stroke, CHD, or cancer.

The above time-varying covariates included in the multivariateanalyses were updated every 2 or 4 years. The last value was car-ried forward for one 2-years cycle to replace missing values. Ifthe last value was also missing, then a missing value indicatorwas created (Huberman and Langholz, 1999). To examine whetherassociations between CCI scores and T2D risk were modified byestablished diabetes risk factors, a cross-product term betweenCCI scores and risk factor was included in the multivariate model.P values for tests for interactions were obtained from a likelihoodratio test comparing the model with the interactions terms to themodel with only the main effects.

To estimate the proportion of the association between phobicanxiety symptoms and diabetes explained by other covariates,SAS macro %MEDIATE (publicly available at http://www.hsph.har-vard.edu/faculty/spiegelman/mediate.html) was applied accordingto the methods described by Lin et al. (1997). For all statisticalanalyses, two-sided p < 0.05 was considered to be statistically sig-nificant. Data were analyzed with SAS 9.3 (SAS Institute Inc., CaryNC).

3. Results

During the 3,099,651 person-years of follow-up, a total of12,831 incident T2D cases (6143 in NHS, 4396 in NHS II and2292 in HPFS) were documented in these 3 cohorts. The CCI scoresranged from 0 to 16. The means of CCI scores were higher in wo-men than men (NHS: 3.02 ± 2.36, NHS II: 2.56 ± 2.22, and HPFS:2.04 ± 1.91 (p < 0.0001)) and a lower proportion of men than wo-men had scores in the highest group of CCI scores (P6) (NHS:14.8%, NHS II: 10.6% and HPFS: 5.7%).

Selected demographic and lifestyle characteristics and potentialconfounders were compared across levels of CCI scores in the 3

cohorts (Table 1). In these cohorts, men and women with CCIscores P6 were more likely to have a larger BMI, to be currentsmokers and to have depression. These participants were also morelikely to have a family history of diabetes and history of hyperten-sion and hypercholesterolemia. CCI scores were also associatedwith use of beta-blockers. Participants with high CCI scores alsotended to have lower levels of physical activity, lower routinescreening for physical exam, and lower diabetes dietary score; theyalso had a higher total energy intake. Women with high CCI scorestended to have lower levels of alcohol consumption but men withhigher score of CCI had higher levels of alcohol consumption.

The age-adjusted HRs of T2D across categories of CCI score (2 to<3, 3 to <4, 4 to <6, P6), compared with a score of <2, were in-creased by 11%, 21%, 28% and 48% (Ptrend < 0.0001) for NHS and31%, 34%, 57% and 102% (Ptrend < 0.0001) for NHS II. In multivariateCox proportional-hazards models with adjustment for demo-graphic, behavior, family history and other biologic factors, relativeto reference group, the HRs of T2D across categories of CCI scoreswere somewhat attenuated but remained significantly increasedby 6%, 10%, 10% and 13% (Ptrend = 0.001) for NHS; and by 19%,11%, 21%, and 29% (Ptrend < 0.0001) for NHS II (Model 3, Table 2).When CCI score was modeled as a continuous variable, afteradjustment for demographic, behavior, family history and otherbiologic factors, the HRs for T2D was 2% higher for each CCI scoreincrement in NHS and 4% higher for each CCI score increment inNHS II (Table 2, Model 3). Additional adjustment for depressionor routine screening for physical exam did not appreciably attenu-ate the HR’s (data not shown). The proportions of the associationbetween phobic anxiety symptom scores and T2D explained byBMI and physical activity were reported in Table 3. In NHS, BMI ex-plained 41.9% (30.5–53.3) and physical activity explained 18.4%(13.1–23.7) of the association between phobic anxiety symptomscores and T2D. In NHS II, BMI explained 43.7% (35.5–51.9) andphysical activity explained 13.7% (10.8–16.6) of the association be-tween phobic anxiety symptom scores and T2D.

In HPFS, the significant association between CCI scores and dia-betes incidence in age-adjusted model (Ptrend = 0.001) disappearedafter adjustment for potential confounders or mediators in Models1–3 (Table 2). However, BMI and physical activity were not signif-icant contributors of phobic anxiety symptom scores and T2D asso-ciation (BMI: 122.8%; 95% CI: �51.7 to 297.4 and physical activity:67.1%; 95% CI: �27.7 to 161.9) (Table 3). In addition, results did notdiffer when we adjusted additionally for depression or routinescreening for physical exam (data not shown).

In sensitivity analysis, in all the three cohorts, when we contin-ued updating a participant’s CCI scores even after a diagnosis ofcancer, stroke or CHD, risk estimates were similar to those ob-tained when we stopped updating CCI scores at these diagnoses(data not shown).

We also examined whether the association between CCI scoresand T2D risk differed by levels of T2D risk factors including smok-ing, alcohol intake, BMI, physical activity, family history of diabe-tes, and beta-blocker use. In all three cohorts, none of the Pvalues for interaction were statistically significant (data notshown).

4. Discussion

In these 3 large prospective cohorts’ studies of US men and wo-men, with 18–20 years of follow-up, we observed that a higherphobic anxiety symptom score was associated with a greater riskof developing T2D in women. Associations remained statisticallysignificant after adjustment for depression and numerous potentialconfounders and appeared to be partly mediated through bodyweight and physical activity. Age-adjusted associations were

Table 1Characteristics of participants according to CCI score in NHS, NHS II and HPFS cohorts.a

CCI score

<2 2 to <3 3 to <4 4 to <6 P6

Women (NHS)Participants (n) 21,251 12,577 10,569 14,336 10,171Age (years) 54.0 ± 7.1 53.8 ± 7.1 53.9 ± 7.1 53.8 ± 7.2 54.1 ± 7.1BMI (kg/m2) 25.0 ± 4.6 25.1 ± 4.6 25.4 ± 4.7 25.5 ± 4.8 25.9 ± 5.0Physical activity (MET-h/wk) 17.0 ± 23.1 16.0 ± 21.3 15.6 ± 21.9 15.0 ± 22.0 13.2 ± 19.1Family history of diabetes (%) 27 28 28 29 31Postmenopausal hormone use (%) 20 20 20 19 19History of hypertension (%) 21 23 25 27 30History of hypercholesterolemia (%) 20 22 23 23 25Current smoker (%) 17 19 19 20 21Multivitamin supplement use (%) 40 40 38 39 37Current aspirin use (%) 64 65 65 66 64Beta-blocker use (%) 7 8 8 9 10Depression (%) 24 27 27 29 34Race [white (%)] 98 98 98 98 97Marital status [married (%)]b 75 75 75 75 73Physical exam [routine screening (%)] 86 83 83 81 78Total energy intake (kcal/d) c 1748 ± 521 1768 ± 521 1775 ± 519 1788 ± 523 1787 ± 540Alcohol (g/d) c 6.4 ± 10.5 6.5 ± 10.6 6.1 ± 10.3 6.4 ± 11.0 6.1 ± 11.3Diabetes dietary score c 12.1 ± 2.4 12.0 ± 2.5 12.0 ± 2.4 11.9 ± 2.4 11.7 ± 2.5

Women (NHS II)Participants (n) 31,063 15,353 11,445 13,602 8497Age (years) 38.0 ± 4.7 38.0 ± 4.6 38.0 ± 4.6 38.2 ± 4.6 38.3 ± 4.6BMI (kg/m2) 24.8 ± 5.2 25.1 ± 5.6 25.2 ± 5.6 25.4 ± 5.8 26.1 ± 6.3Physical activity (MET-h/wk)d 22.3 ± 28.5 21.2 ± 27.8 20.2 ± 25.7 19.9 ± 26.4 18.3 ± 25.0Family history of diabetes (%) 33 34 34 36 37Postmenopausal hormone use (%) 6 6 7 7 8History of hypertension (%) 6 7 8 8 10History of hypercholesterolemia (%) 15 16 18 19 22Current smoker (%) 9 11 11 13 15Multivitamin supplement use (%) 44 45 45 43 43Current aspirin use (%) 8 9 10 11 11Beta-blocker use (%)e 4 5 5 6 7Depression (%) 15 19 23 27 42Race [white (%)] 97 97 97 96 94Marital status [married (%)] 79 81 81 81 80Physical exam [routine screening (%)]f 74 73 70 69 66Total energy intake (kcal/d) c 1769 ± 536 1791 ± 537 1800 ± 547 1808 ± 550 1819 ± 572Alcohol (g/d) c 3.2 ± 5.9 3.2 ± 6.0 3.3 ± 6.5 3.2 ± 6.3 2.8 ± 6.1Diabetes dietary score c 12.2 ± 2.7 12.0 ± 2.7 12.0 ± 2.7 11.9 ± 2.7 11.6 ± 2.7

Men (HPFS)Participants (n) 14,868 5816 4056 4281 1770Age (years) 54.3 ± 9.4 54.4 ± 9.5 54.9 ± 9.5 55.2 ± 9.5 56.0 ± 9.8BMI (kg/m2) 25.3 ± 3.0 25.4 ± 3.1 25.5 ± 3.1 25.7 ± 3.2 25.9 ± 3.5Physical activity (MET-h/wk) 31.1 ± 36.4 29.0 ± 32.3 28.7 ± 32.4 27.0 ± 31.7 24.3 ± 31.2Family history of diabetes (%) 19 20 22 22 22History of hypertension (%) 16 18 20 21 26History of hypercholesterolemia (%) 17 19 20 21 24Current smoker (%) 8 9 9 9 10Multivitamin supplement use (%) 40 42 41 42 43Current aspirin use (%) 35 39 40 39 43Beta-blocker use (%) 6 6 8 8 9Depression (%)g 14 19 19 22 26Race [white (%)] 96 96 95 94 92Marital status [married (%)] 90 90 90 91 89Physical exam [routine screening (%)] 65 63 63 62 60Total energy intake (kcal/d)c 1991 ± 609 2017 ± 620 2038 ± 630 2032 ± 626 2040 ± 656Alcohol (g/d)c 11.3 ± 15.0 11.6 ± 15.3 11.7 ± 15.4 11.9 ± 15.9 12.3 ± 17.0Diabetes dietary scorec 12.1 ± 2.7 12.0 ± 2.7 12.0 ± 2.7 11.9 ± 2.7 11.8 ± 2.7

a All characteristics are from questionnaire in 1988 in NHS, 1993 in NHS II and 1988 in HPFS, except if otherwise indicated.b Variables are obtained from 1992 questionnaire.c Variables are obtained from 1986 questionnaire in NHS, 1991 questionnaire in NHS II and 1986 questionnaire in HPFS.d Variables are obtained from 1991 questionnaire.e Variables are obtained from 2001 questionnaire.f Variables are obtained from 1989 questionnaire.g Variables are obtained from 2000 questionnaire.

M.S. Farvid et al. / Brain, Behavior, and Immunity 36 (2014) 176–182 179

evident in men but these were less robust with addition of covar-iates. Our results did not support a significant role for phobic anx-iety symptoms in T2D incidence in men independent of knownT2D risk factors.

We found a gender difference in the relation between phobicanxiety symptoms and T2D risk. Association between phobic anx-iety and risk of CHD in men and women has been previously re-ported in our and other cohorts (Albert et al., 2005; Kawachi

Table 2Hazard ratio and 95% confidence intervals of type 2 diabetes by CCI score in NHS, NHS II and HPFS cohorts.

CCI Score

<2 2 to <3 3 to <4 4 to <6 P6 Ptrend Per 1 CCI score

Women (NHS, 1988–2008)Cases/person-yrs 1657/387,987 1087/228,899 972/189,333 1361/252,662 1066/171,867Age-adjusted model 1 1.11 (1.03–1.20) 1.21 (1.12–1.31) 1.28 (1.19–1.37) 1.48 (1.37–1.60) <0.0001 1.06 (1.05–1.07)Multivariate-adjusted model 1 1 1.08 (1.00–1.16) 1.14 (1.05–1.23) 1.17 (1.09–1.26) 1.26 (1.16–1.36) <0.0001 1.04 (1.03–1.05)Multivariate-adjusted model 2 1 1.07 (0.99–1.16) 1.13 (1.04–1.22) 1.16 (1.08–1.24) 1.23 (1.13–1.33) <0.0001 1.03 (1.02–1.04)Multivariate-adjusted model 3 1 1.06 (0.98–1.14) 1.10 (1.01–1.18) 1.10 (1.03–1.19) 1.13 (1.05–1.22) 0.001 1.02 (1.01–1.03)

Women (NHS II, 1993–2011)Cases/person-yrs 1411/547,821 876/262,598 627/186,708 844/215,869 638/128,290Age-adjusted model 1 1.31 (1.20–1.42) 1.34 (1.22–1.47) 1.57 (1.44–1.71) 2.02 (1.84–2.22) <0.0001 1.11 (1.09–1.12)Multivariate-adjusted model 1 1 1.26 (1.16–1.37) 1.22 (1.11–1.34) 1.34 (1.23–1.46) 1.55 (1.41–1.70) <0.0001 1.06 (1.05–1.08)Multivariate-adjusted model 2 1 1.24 (1.14–1.35) 1.21 (1.10–1.32) 1.32 (1.21–1.44) 1.50 (1.37–1.66) <0.0001 1.06 (1.05–1.07)Multivariate-adjusted model 3 1 1.19 (1.09–1.29) 1.11 (1.01–1.22) 1.21 (1.11–1.32) 1.29 (1.17–1.41) <0.0001 1.04 (1.02–1.05)

Men (HPFS, 1988–2008)Cases/person-yrs 1064/263,089 451/101,661 333/67,955 317/68,342 127/26,570Age-adjusted model 1 1.10 (0.98–1.23) 1.21 (1.07–1.37) 1.15 (1.02–1.31) 1.21 (1.00–1.45) 0.001 1.04 (1.02–1.06)Multivariate-adjusted model 1 1 1.04 (0.93–1.16) 1.12 (0.99–1.27) 1.05 (0.93–1.19) 1.01 (0.84–1.22) 0.38 1.01 (0.99–1.03)Multivariate-adjusted model 2 1 1.03 (0.92–1.15) 1.11 (0.98–1.26) 1.04 (0.91–1.17) 0.98 (0.82–1.18) 0.59 1.01 (0.98–1.03)Multivariate-adjusted model 3 1 1.01 (0.90–1.13) 1.09 (0.97–1.24) 0.97 (0.85–1.10) 0.90 (0.74–1.08) 0.46 0.99 (0.97–1.01)

Multivariate-adjusted model 1: age (year), race (white/non-white), marital status (married/non-married), family history of diabetes, current aspirin use, smoking status(never, past, current 1–14/d, current 15–24/d, current P25), alcohol intake (none, .01–4.9, 5–14.9, P15 g/day), physical activity (<3, 3–8.9, 9–17.9, 18–26.9, P27 METs/wk).Among women, we also adjusted for menopausal status and hormone use (premenopausal, postmenopausal never users, postmenopausal past users, postmenopausal currentusers) and husband’s education (<high school, some high school, high school grade or collage grade, graduate school, not applicable or missing).Multivariate-adjusted model 2: model 1 plus energy intake (quintile), diabetes dietary score (quintile), coffee (quintile), Multivariate-adjusted model 3: model 2 plus BMI(<23, 23–25, 25–29.9, 30–34.9, P35 kg/m2).

Table 3The mediation proportion for the effect of phobic anxiety symptoms on type 2diabetes risk explained by BMI and physical activity in NHS, NHS II and HPFS cohorts.

HRs (95% CI) per 1 CCIscore

Mediation proportion(95% CI)

Women (NHS, 1988–2008)Base model* 1.04 (1.03–1.05)Base model + physical

activity1.03 (1.02–1.04) 18.4% (13.1–23.7%)

Base model + BMI 1.02 (1.01–1.03) 41.9% (30.5–53.3%)

Women (NHS II, 1993–2011)Base model* 1.07 (1.06–1.08)Base model + physical

activity1.06 (1.05–1.07) 13.7% (10.8–16.6%)

Base model + BMI 1.04 (1.03–1.05) 43.7% (35.5–51.9%)

Men (HPFS, 1988–2008)Base model* 1.02 (0.99–1.04)Base model + physical

activity1.00 (0.98–1.03) 67.1% (�27.7–161.9%)

Base model + BMI 1.00 (0.98–1.02) 122.8% (�51.7–297.4%)

* Base model adjusted for age (year), race (white/non-white), marital status(married/non-married), family history of diabetes, current aspirin use, smokingstatus (never, past, current 1–14/d, current 15–24/d, current P25), alcohol intake(none, .01–4.9, 5–14.9, P15 g/day), energy intake (quintile), diabetes dietary score(quintile), coffee (quintile), Among women, we also adjusted for menopausal statusand hormone use (premenopausal, postmenopausal never users, postmenopausalpast users, postmenopausal current users) and husband’s education (<high school,some high school, high school grade or collage grade, graduate school, not appli-cable or missing).

180 M.S. Farvid et al. / Brain, Behavior, and Immunity 36 (2014) 176–182

et al., 1994; Watkins et al., 2010). To our knowledge, there havebeen no prior studies in which the relationship between phobicanxiety symptoms and risk of T2D has been directly examined inboth genders with similar measures and long-term follow-up.However, gender differences in the relationship between otherforms of psychological stress and T2D have been shown. In linewith our results, after 11 years of follow-up, Norberg et al. (2007)demonstrated that work stress and low emotional support in-creased the risk of T2D among middle-aged women, but not inmen. In contrast, Kato et al. (2009), in a community-based,

prospective cohort study, found that the risk of T2D increased withan increasing mental stress level, among men but not women. Theyassessed mental stress only based on three levels of response (low,medium and high) to the question, ‘‘How much stress do you feelin your daily life?’’. In contrast to mental stress, they reported thatthe association between type A behavior (based on a validatedinstrument) and incidence of T2D was significant only among wo-men. Also, Eriksson et al. (2008) found that self-reported psycho-logical distress, measured by five items (anxiety, apathy,depression, fatigue and insomnia) in the questionnaire, was notassociated with later onset of T2D in middle-aged women butwas related to higher risk of T2D in middle-aged men. Psycholog-ical distress was measured by questioning how often during thelatest 12 months they had been troubled by anxiety, apathy,depression, fatigue and insomnia. In a cohort study in short termfollow-up, Atlantis et al. (2012) found higher risk of diabetes inDanish men and women with depression and anxiety. However,the results were not separately reported for men and women.

We could not determine why our observed associationsbetween phobic anxiety symptoms and T2D were stronger inwomen than in men. It is possible that the differential associationis partly related to the different distribution of phobic anxietysymptoms between men and women. Furthermore, it has beenhypothesized that phobic anxiety symptoms may influence T2Drisk through inflammation (Brennan et al., 2009; Pitsavos et al.,2006) and activation of the hypothalamic–pituitary–adrenal(HPA) axis/hypercortisolism (Gragnoli, 2012). In addition, cortico-tropin-releasing hormone (CRH) plays an essential role in the HPAaxis and the stress response and both estrogen receptor a and bcan stimulate CRH gene expression (Chen et al., 2008). Thus, itis possible that involvement of sex hormones in the regulationof the stress response and the HPA axis may partly determinewho is more likely to develop new onset T2D under chronic stress(Chen et al., 2008), thereby distinguish effects among men andwomen.

Similar other emotional stress, phobic anxiety may influenceon behavioral factors, such as sedentary lifestyles, smoking, drink-ing alcohol and overeating that may also help to explain the

M.S. Farvid et al. / Brain, Behavior, and Immunity 36 (2014) 176–182 181

effects of phobic anxiety on the onset of diabetes (Rod et al.,2009; Kye and Park, 2012; Bonnet et al., 2005). In these cohorts,participants with higher phobic anxiety symptom scores had lar-ger BMI, higher total energy intake, lower levels of physical activ-ity and a lower diabetes dietary score and were more likely tosmoke. There was a gender difference in alcohol consumption:with higher scores on the phobic anxiety symptom scale, womentended to have lower levels while men had higher levels of alco-hol consumption. While risk in women was somewhat attenuatedby inactivity, obesity, alcohol and dietary intake, and smokingstatus, the association was also maintained even after adjustingfor these factors. Furthermore, we did not observe the significantinteractions with BMI, physical activity, alcohol intake, family his-tory of diabetes, smoking and beta-blocker use. Nevertheless, inmen, we found that adjustment for alcohol intake, BMI and life-style factors substantially decreased the HRs, indicating phobicanxiety symptoms may be related to diabetes through poorhealth-related behaviors.

The large sample size and long follow-up provided sufficientpower to examine the association with T2D over a wide rangeof phobic anxiety symptom scores in men and women. Further-more, the prospective nature and high rate of follow-up minimizethe potential for recall and selection bias compared with otherobservational study designs because phobic anxiety symptomswere assessed prior to the event (T2D). The detailed, standardizedand updated information on diet, physical activity, smoking andBMI in the 3 cohorts allowed us to have information on a widerange of health and lifestyle-related confounders and control forthem throughout the follow-up. Furthermore, we excluded partic-ipants with CHD, stroke and cancer at baseline and also stoppedupdating CCI scores after developing these diseases during thefollow-up that could be associated with mood disorders. How-ever, including these participants in the analysis did not changethe results.

Potential limitations also need to be considered. Because of pre-dominant white educated US adults, we cannot determine whetherour findings are generalizable to other race or ethnic groups. Sev-eral T2D risk factors such as decreased physical activity and smok-ing could play the causal role in link between phobic anxiety andT2D risk, thus raising the possibility of statistical over-adjustmentand leading to an underestimate of the association between phobicanxiety and T2D incidence. Also, detection bias cannot be ignoredbecause phobic anxiety might affect testing for T2D. Participantswith higher score of CCI had lower routine screening for physicalexam and avoided seeing doctors. Another limitation of the studyis the long interval between two measurements of phobic anxietysymptoms, and therefore, the inability to account for changes inphobic anxiety in short time, which would tend to obscure associ-ations if the effect is of short duration. However, phobic disorderstend to be early in onset and chronic in nature. The age of onset forphobic disorders is typically in childhood or adolescence and con-sidered to be altered little during the life (Kessler et al., 2005b;Crown and Crisp, 1966; Kessler et al., 2010). Despite this limitation,CCI score has demonstrated strong relations with diabetes inci-dence in women.

In summary, our study in prospective cohorts suggests thathigher levels of phobic anxiety symptoms are associated with in-creased risk of T2D in women, independent of other known riskfactors; whereas phobic anxiety symptoms are not associated withT2D risk in men. Further studies are needed to elucidate the poten-tial mechanisms underlying these associations and also the differ-ences between men and women. Furthermore, if phobic anxietysymptoms are associated with the T2D incidence, at least inwomen; then, any potential impacts on T2D incidence may be pre-vented through early recognition, diagnosis and treatment of thesedisorders.

Acknowledgments

We thank the participants and staffs of the NHS, NHS II andHPFS for their valuable contributions. The study was supportedby the National Institutes of Health grants (R01CA050385,U19CA055075, R01DK058845, P30DK046200, MH092707-01,R01HL71981, R01HL60712 and U54CA155626). The study spon-sors were not involved in the study design and collection, analysisand interpretation of data, or the writing of the article or the deci-sion to submit it for publication. The authors were independentfrom study sponsors.

The authors’ responsibility were as follows: M.S.F., L.Q., F.B.H.and W.C.W.: designed the research; M.S.F.: analysis and wrotethe manuscript; and W.C.W.: had primary responsibility for the fi-nal content of the manuscript; and all authors: provided critical in-put in the writing of the manuscript and read and approved thefinal manuscript.

References

Albert, C.M., Chae, C.U., Rexrode, K.M., Manson, J.E., Kawachi, I., 2005. Phobicanxiety and risk of coronary heart disease and sudden cardiac death amongwomen. Circulation 111, 480–487.

Arroyo, C., Hu, F.B., Ryan, L.M., Kawachi, I., Colditz, G.A., Speizer, F.E., Manson, J.,2004. Depressive symptoms and risk of type 2 diabetes in women. DiabetesCare 27, 129–133.

Atlantis, E., Vogelzangs, N., Cashman, K., Penninx, B.J., 2012. Common mentaldisorders associated with 2-year diabetes incidence: the Netherlands Study ofDepression and Anxiety (NESDA). J. Affect. Disord. 142 (Suppl.), S30–S35.

Bonnet, F., Irving, K., Terra, J.L., Nony, P., Berthezène, F., Moulin, P., 2005. Depressivesymptoms are associated with unhealthy lifestyles in hypertensive patientswith the metabolic syndrome. J. Hypertens. 23, 611–617.

Brennan, A.M., Fargnoli, J.L., Williams, C.J., Li, T., Willett, W., Kawachi, I., Qi, L., Hu,F.B., Mantzoros, C.S., 2009. Phobic anxiety is associated with higher serumconcentrations of adipokines and cytokines in women with diabetes. DiabetesCare 32, 926–931.

Byers, A.L., Yaffe, K., Covinsky, K.E., Friedman, M.B., Bruce, M.L., 2010. Highoccurrence of mood and anxiety disorders among older adults: the NationalComorbidity Survey Replication. Arch. Gen. Psychiatry 67, 489–496.

Chen, X.N., Zhu, H., Meng, Q.Y., Zhou, J.N., 2008. Estrogen receptor-alpha and betabregulate the human corticotropin-releasing hormone gene through similarpathways. Brain Res. 1223, 1–10.

Chen, L., Magliano, D.J., Zimmet, P.Z., 2011. The worldwide epidemiology of type 2diabetes mellitus: present and future perspectives. Nat. Rev. Endocrinol. 8, 228–236.

Crown, S., Crisp, A.H., 1966. A short clinical diagnostic self-rating scale forpsychoneurotic patients. The Middlesex Hospital Questionnaire (MHQ). Br. J.Psychiatry 112, 917–923.

Eriksson, A.K., Ekbom, A., Granath, F., Hilding, A., Efendic, S., Ostenson, C.G., 2008.Psychological distress and risk of pre-diabetes and Type 2 diabetes in aprospective study of Swedish middle-aged men and women. Diabet. Med. 25,834–842.

Feskanich, D., Rimm, E.B., Giovannucci, E.L., Colditz, G.A., Stampfer, M.J., et al., 1993.Reproducibility and validity of food intake measurements from asemiquantitative food frequency questionnaire. J. Am. Diet. Assoc. 93, 790–796.

Field, A.E., Coakley, E.H., Must, A., Spadano, J.L., Laird, N., Dietz, W.H., Rimm, E.,Colditz, G.A., 2001. Impact of overweight on the risk of developing commonchronic diseases during a 10-year period. Arch. Intern. Med. 161, 1581–1586.

Gavin III, J.R., Alberti, K.G.M.M., Davidson, M.B., DeFronzo, R.A., Drash, A., Gabbe,S.G., Genuth, S., Harris, M.I., Kahn, R., Keen, H., Knowler, W.C., Lebovitz, H.,Maclaren, N.K., Palmer, J.P., Raskin, P., Rizza, R.A., Stem, M.P., 1997. Report of theexpert committee on the diagnosis and classification of diabetes mellitus.Diabetes Care 20, 1183–1197.

Gragnoli, C., 2012. Depression and type 2 diabetes: cortisol pathway implicationand investigational needs. J. Cell. Physiol. 227, 2318–2322.

Hu, F.B., Manson, J.E., Stampfer, M.J., Colditz, G., Liu, S., Solomon, C.G., Willett, W.C.,2001. Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N. Engl. J.Med. 13 (345), 790–797.

Huberman, M., Langholz, B., 1999. Application of the missing-indicator method inmatched cse-control studies with incomplete data. Am. J. Epidemiol. 150, 1340–1345.

Kato, M., Noda, M., Inoue, M., Kadowaki, T., Tsugane, S., 2009. JPHC study group.psychological factors, coffee and risk of diabetes mellitus among middle-agedJapanese: a population-based prospective study in the JPHC study cohort.Endocr. J. 56, 459–468.

Kawachi, I., Colditz, G.A., Ascherio, A., Rimm, E.B., Giovannucci, E., Stampfer, M.J.,Willett, W.C., 1994. Prospective study of phobic anxiety and risk of coronaryheart disease in men. Circulation 89, 1992–1997.

182 M.S. Farvid et al. / Brain, Behavior, and Immunity 36 (2014) 176–182

Kessler, R.C., Chiu, W.T., Demler, O., Walters, E.E., 2005a. Prevalence, severity, andcomorbidity of twelve month DSM-IV disorders in the National ComorbiditySurvey Replication (NCS-R). Arch. Gen. Psychiatry 62, 617–627.

Kessler, R.C., Berglund, P., Demler, O., Jin, R., Merikangas, K.R., Walters, E.E., 2005b.Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in theNational Comorbidity Survey Replication. Arch. Gen. Psychiatry 62, 593–602.

Kessler, R.C., Ruscio, A.M., Shear, K., Wittchen, H.U., 2010. Epidemiology of anxietydisorders. Curr. Top. Behav. Neurosci. 2, 21–35.

Kye, S.Y., Park, K., 2012. Psychosocial factors and health behavior among Koreanadults: a cross-sectional study. Asian Pac. J. Cancer Prev. 13, 49–56.

Lin, D.Y., Fleming, T.R., De Gruttola, V., 1997. Estimating the proportion of treatmenteffect explained by a surrogate marker. Stat. Med. 16, 1515–1527.

Manson, J.E., Rimm, E.B., Stampfer, M.J., Colditz, G.A., Willett, W.C., Krolewski, A.S.,Rosner, B., Hennekens, C.H., Speizer, F.E., 1991. Physical activity and incidenceof non-insulin-dependent diabetes mellitus in women. Lancet 338, 774–778.

Mavissakalian, M., Michelson, L., 1981. The middlesex hospital questionnaire: avalidity study with American psychiatric patients. Br. J. Psychiatry 139, 336–340.

Mirza, S., Hossain, M., Mathews, C., Martinez, P., Pino, P., Gay, J.L., Rentfro, A.,McCormick, J.B., Fisher-Hoch, S.P., 2012. Type 2-diabetes is associated withelevated levels of TNF-alpha, IL-6 and adiponectin and low levels of leptin in apopulation of Mexican Americans: a cross-sectional study. Cytokine 57, 136–142.

National Diabetes Data Group, 1979. Classification and diagnosis of diabetesmellitus and other categories of glucose intolerance. Diabetes 28, 1039–1057.

Norberg, M., Stenlund, H., Lindahl, B., Andersson, C., Eriksson, J.W., Weinehall, L.,2007. Work stress and low emotional support is associated with increased riskof future type 2 diabetes in women. Diabetes Res. Clin. Pract. 76, 368–377.

Okereke, O.I., Grodstein, F., 2012. Phobic anxiety and cognitive performance over4 years among Community-Dwelling Older Women in the Nurses’ Health study.Am. J. Geriatr. Psychiatry 31 [Epub ahead of print].

Pitsavos, C., Panagiotakos, D.B., Papageorgiou, C., Tsetsekou, E., Soldatos, C.,Stefanadis, C., 2006. Anxiety in relation to inflammation and coagulationmarkers, among healthy adults: the ATTICA study. Atherosclerosis 185, 320–326.

Pouwer, F., Kupper, N., Adriaanse, M.C., 2010. Does emotional stress cause type 2diabetes mellitus? A review from the European depression in diabetes (EDID)research consortium. Discov. Med. 9, 112–118.

Psaltopoulou, T., Ilias, I., Alevizaki, M., 2010. The role of diet and lifestyle in primary,secondary, and tertiary diabetes prevention: a review of meta-analyses. Rev.Diabet. Stud. 7, 26–35.

Rod, N.H., Grønbaek, M., Schnohr, P., Prescott, E., Kristensen, T.S., 2009. Perceivedstress as a risk factor for changes in health behavior and cardiac risk profile: alongitudinal study. J. Intern. Med. 266, 467–475.

Venables, M.C., Jeukendrup, A.E., 2009. Physical inactivity and obesity: links withinsulin resistance and type 2 diabetes mellitus. Diabetes Metab. Res. Rev. 25,S18–S23.

Watkins, L.L., Blumenthal, J.A., Babyak, M.A., Davidson, J.R., McCants Jr., C.B.,O’Connor, C., Sketch Jr., M.H., 2010. Phobic anxiety and increased risk ofmortality in coronary heart disease. Psychosom. Med. 72, 664–671.

Zabetian, A., Keli, H.M., Echouffo-Tcheugui, J.B., Narayan, K.M., Ali, M.K., 2013.Diabetes in the middle east and north Africa. Diabetes Res. Clin. Pract. 101, 106–122.