Epidemiology/Health Services/Psychosocial Research

O R I G I N A L A R T I C L E

History of Depression Increases Risk of

Type 2 Diabetes in Younger Adults
LAUREN C. BROWN, BSCPHARM1 STEPHEN C. NEWMAN, MD, MA, MSC1,4 people with depression are more likely to
SUMIT R. MAJUMDAR, MD, MPH, FRCPC2,3 JEFFREY A. JOHNSON, PHD1,2 abuse alcohol and smoke cigarettes com-
pared with individuals without depres-
sion. These behaviors can potentially
increase the risk of developing type 2 di-
abetes. However, the literature evaluating
OBJECTIVE — The purpose of this study was to assess the history of previous depression in depression as a risk factor for diabetes is
people with incident diabetes compared with people without diabetes. quite inconsistent.
Nine prospective observational stud-
RESEARCH DESIGN AND METHODS — We conducted a population-based nested
case-control study using the administrative databases of Saskatchewan Health to assess the study ies (15–23) suggest that depression and
objective. We identified cases of type 2 diabetes based on diagnostics codes and prescription depressive symptoms may be a risk factor
records for individuals over the age of 20 years. For each case subject, two control subjects were for the development of type 2 diabetes,
randomly selected from the nondiabetic population during the same index year. History of with relative risk estimates ranging from
depression, based on diagnostic codes and antidepressant prescription, was ascertained up to 3 1.3 to 3.0. The majority of these studies
years before index date. Simple and multivariate logistic regression analysis was used to estimate relied on self-reported measures to iden-
the odds ratio (OR) and 95% CIs, after adjusting for age, sex, and frequency of physician visits. tify depressive symptoms or depression
and diabetes. Also, some of the studies
RESULTS — Individuals with newly diagnosed diabetes (1,622 of 33,257; 4.9%) were 30% involved limited follow-up or relatively
more likely to have had a previous history of depression compared with people without diabetes
homogenous samples with limited gener-
(2,279 of 59,420; 3.8%). This increased risk remained after controlling for sex and number of
physician visits but was limited to subjects 20 –50 years of age (adjusted OR 1.23 [95% CI alizability. For example, Arroyo et al. (20)
1.10 –1.37]) and not in those aged ⱖ51 years (0.92 [0.84 –1.00]). evaluated women aged 45–72 years,
Kawakami et al. (18) only evaluated men,
CONCLUSIONS — Depression appears to increase the risk of developing diabetes by ⬃23% and Palinkas et al. (19) assessed risk of
in younger adults. This provides information regarding the temporality of the relationship diabetes in people aged ⱖ50 years, limit-
between diabetes and depression. ing the applicability of these studies to the
general population. In contrast, Saydah et
Diabetes Care 28:1063–1067, 2005 al. (24) observed no difference in the in-
cidence of diabetes between those who
reported high depressive symptoms or

D
epression is a highly prevalent dis- U.S. (13,14), adjustment for age, sex, and moderate symptoms compared with indi-
ease associated with substantial cardiovascular disease tends to reduce the viduals with no depressive symptoms.
morbidity and mortality (1– 6) and previous prevalence estimates from meta- Likewise, Kessing et al. (25) found no in-
is recognized as an important comorbid- analyses of published studies (11). creased risk of developing type 1 or 2 di-
ity for a number of chronic medical con- The majority of this literature, how- abetes in people with depression or
ditions (7–10). Diabetes is among many ever, has reported only cross-sectional as- bipolar disorder. However, these studies
chronic medical conditions that appear to sessments of comorbid diabetes and were also not without their limitations.
be adversely affected by comorbid de- depression. Little information is available The study conducted by Saydah et al. (24)
pression. Numerous reports have indi- on the temporal association between dia- relied on self-reported measures for both
cated that patients with diabetes are 1.5–2 betes and depression. Onset of depression depressive symptoms and identification
times more likely to have depression com- may result in increased weight gain (as a of diabetes. Kessing et al. (25) required
pared with people without diabetes (11– result of the disorder or in relation to an- patients to be rehospitalized to identify a
14). In the most recent estimates using tidepressant treatment) and decreased diagnosis of diabetes after a previous hos-
large administrative databases from the self-care measures such as exercise. Also, pital admission for an affective disorder,
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● excluding individuals who may have been
From the 1Department of Public Health Sciences, University of Alberta, Edmonton, Alberta, Canada; the diagnosed with diabetes outside of the
2
Institute of Health Economics, Edmonton, Alberta, Canada; the 3Department of Medicine, University of hospital.
Alberta, Edmonton, Alberta, Canada; and the 4Department of Psychiatry, University of Alberta, Edmonton, In summary, it is clear that diabetes
Alberta, Canada.
Address correspondence and reprint requests to Jeffrey A. Johnson, PhD, 1200-10405 Jasper Ave.,
and depression are important comorbid
Edmonton, Alberta, Canada T5J 3N4. E-mail: [email protected]. conditions, and comorbid depression is
Received for publication 22 October 2004 and accepted in revised form 7 February 2005. associated with worse outcomes in people
A table elsewhere in this issue shows conventional and Système International (SI) units and conversion with diabetes. It is not entirely clear, how-
factors for many substances. ever, whether those with a history of de-
© 2005 by the American Diabetes Association.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby pression are somehow predisposed to
marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. developing diabetes. The temporal rela-

DIABETES CARE, VOLUME 28, NUMBER 5, MAY 2005 1063

Depression and diabetes

tionship has important implications for Selection of case and control years before the diabetes index date. Indi-
the mechanisms whereby depression subjects viduals identified as having depressive
might predispose to diabetes and man- Individuals eligible for inclusion in this episodes, diagnosed ⬎3 years before dia-
agement of diabetes risk in individuals di- study were residents of Saskatchewan, el- betes identification, were considered to
agnosed with depression. We therefore igible for prescription drug benefits dur- have ongoing depression if they had a dis-
conducted a nested case-control study, ing the study period, and aged ⬎20 years. pensation for an antidepressant medica-
using a large population-based adminis- Two study groups were identified: people tion within 3 years before index date.
trative dataset, to assess whether people with diabetes (case subjects) and people
with a history of previous depressive epi- without diabetes (control subjects). Data analysis
sodes were at increased risk compared Case subjects were identified based All study groups were described in terms
with those without such a history to de- on the established case definition of the of age, sex, and number of physician visits
velop type 2 diabetes. National Diabetes Surveillance System in the year before diabetes index date. We
(28 –30) within the diabetes index period. initially estimated the unadjusted OR and
Subjects were identified as having diabe- 95% CI of the association between de-
tes if they had two or more physician ser- pression and new-onset diabetes using
RESEARCH DESIGN AND
vice claims for diabetes (ICD-9 code 250) simple logistic regression, with case ver-
METHODS
within a 2-year period or one or more sus control subject status as the depen-
hospitalizations with a diabetes code as dent variable and a history of previous
Saskatchewan Health databases the primary, secondary, or other diagno- depressive episodes as the main indepen-
Saskatchewan Health databases include sis. Dispensation of oral antidiabetic dent variable. We then used multivariate
information on most residents (99%) of agents (see APPENDIX) was used to restrict logistic regression, including age, sex, and
the province of Saskatchewan (popula- the diabetes cohort to type 2 diabetic pa- number of physician visits in the year be-
tion ⬃1 million) (26,27). Individuals not tients. Diabetes index date was identified fore index date, to adjust for comorbidity
covered by Saskatchewan Health include as the date of first dispensation for an oral and medical surveillance bias in case sub-
those with federally funded health care, antidiabetic agent or the date the National jects (32). We tested for the possibility of
such as members of the Royal Canadian Diabetes Surveillance System criteria statistically significant interaction terms
Mounted Police and Canadian Forces were met, whichever came first. Women between depression status and all avail-
(26). About 90% of the covered popula- with services claims for gestational diabe- able covariates. We prespecified that we
tion is eligible for prescription drug ben- tes (ICD-9 648.8) were excluded. would only consider as important those
efits. Those ineligible include registered Control subjects (i.e., did not meet interaction terms that achieved a level of
Indians who receive prescription benefits the definition for diabetes during the statistical significance of P ⬍ 0.10.
through a federal program. Data from washout or index periods) were identified Age was initially included in the anal-
four different data files were used in this by randomly selecting two subjects from ysis as a continuous variable, along with a
study: the health registration, outpatient the nondiabetic population for each dia- quadratic term to adjust for the nonlinear-
prescription drug, medical services, and betic subject within the same index year. ity of the relationship between age and
hospital separation. The data files are Control subjects were assigned the same diabetes onset. In this continuous and
linkable based on personal health num- index date as their respective case subject quadratic form, we found a highly signif-
bers and provide demographic informa- and were not matched on any demo- icant interaction between age and depres-
tion, prescription drug usage, and diag- graphic characteristics. sion. To better evaluate this interaction,
nostic codes for outpatient visits and hos- we recoded age into a categorical variable
pital stays. Prior depression grouped by decade. In this form, it was
To identify depression using the adminis- found that the rate of depression was rel-
trative databases, we used a composite atively equivalent in those aged 20 –50
Study periods case definition previously validated in the years and in those aged ⱖ51 years. There-
administrative databases of Saskatchewan fore, to further simplify the interpretation
Incident cases of diabetes and randomly
Health by West et al. (31). Based on this of our results, we included age in the fi-
selected control subjects were identified
definition, we considered the following to nal regression model as a dichotomous
between 1 January 1992 and 31 Decem-
be an episode of depression: a prescrip- variable split at the age of 51 years. All
ber 2000 (i.e., index period). Study index tion for an antidepressant medication and analyses were conducted using SPSS ver-
date was defined as date diabetes was any one of three ICD-9 codes for depres- sion 12.0.
identified, and randomly selected control sive disorders (i.e., 296, 309, or 311)
subjects were given the same index date as from the physician services records RESULTS — The data for the study
their respective diabetes case subjects. To within a 6-month reference period (i.e., comprised 92,677 people, of whom
ensure we identified people with new- ⫾3 months). The criteria evaluated by 33,257 were case subjects. The mean age
onset diabetes, individuals who met the West et al. for identifying people with of the sample was 52 years (median 51
case subject definition from 1 January depression in the Saskatchewan Health years [range 20 –95]), and 49% of the
1989 to 31 December 1991 were ex- administrative databases provide a sensi- subjects were women, with an average of
cluded (i.e., a 3-year diabetes washout pe- tivity of 71% and a specificity of 85%. We 8.2 (median 5.0) physician visits in the
riod). Exposure period was defined as a identified individuals with at least one de- year before study index date. The mean
3-year period before study index date. pressive episode at least 1 year and up to 3 age of case subjects was 61.3 years (range

1064 DIABETES CARE, VOLUME 28, NUMBER 5, MAY 2005

 Brown and Associates

remained after controlling for potential

confounding variables including age, sex,
and number of physician visits before
study index date. Our analyses indicate
that this increased risk lies mainly in the
population of people aged 20 –50 years.
This is an important observation, as the
onset of depression typically occurs be-
tween age 20 –30 years (33,34).
Figure 1—History of depression in peo-
ple with and without diabetes stratified
There are several potential mecha-
by age in decades. nisms at play in this observed relation-
ship. It is possible that depression itself
contributes to diabetes. For example, in-
20 –95), and 45% of them were women, significant interaction between depres- dividuals with depression are more likely
whereas the mean age for control subjects sion and age (P ⫽ 0.002) (Table 1). To to have body weight changes and are less
was 46.9 years (range 20 –94), and 51% further explore the interaction term be- likely to engage in healthy behaviors such
were women. Case subjects had an aver- tween age and depression and still recog- as exercise, which increases the risk of de-
age of 11.5 physician visits (range 8 –223) nize the nonlinearity for age, we explored veloping diabetes (33). Alternatively,
and control subjects an average of 6.3 the relationship with age divided into de- many of the medications used to treat de-
(range 4 –170) in the year before index cades. We observed that the relationship pression cause weight gain and sedation,
date. Of 3,901 individuals found to have between history of depression and dia- such as tricyclic antidepressants and se-
depression, 2,160 episodes occurred betes was more similar in people aged lective serotonin reuptake inhibitors,
within 3 years before index date. For the 20 –50 years and individuals aged ⱖ51 which could also contribute to develop-
1,741 individuals with ongoing depres- years (Fig. 1). ing diabetes. In either case, the onset of
sion, onset was an average of 5.9 (SD 2.1) We therefore conducted multivariate depression or its treatment may unmask a
years before diagnosis of diabetes (range analyses that adjusted for sex and physi- tendency for the development of diabetes
3–11). cian visits after stratifying our sample, perhaps earlier than it otherwise would
A history of depression was more which was dichotomized based on age. have manifested. Individuals who were
common in people with new-onset diabe- People with diabetes who were aged aged ⱖ51 years may have already passed
tes (1,622 of 33,257; 4.9%) when com- 20 –50 years were more likely to have had the period of increased risk of developing
pared with the control subjects (2,279 of a prior depressive episode compared with diabetes, such that a previous history of
59,420; 3.8%); the overall unadjusted OR those without diabetes (adjusted OR 1.23 depression did not alter the risk of subse-
for this association was 1.29 (95% CI [95% CI 1.10 –1.37]) (Table 2). Individ- quently developing diabetes. That is, hav-
1.20 –1.37). For people 20 –50 years of uals aged ⱖ51 years with diabetes were ing depressive episodes may accelerate
age, the unadjusted OR for the association no more or less likely to have had a prior the onset of diabetes in at risk individuals.
between depression and diabetes was episode of depression compared with A recent report of higher rates of depres-
1.76 (1.58 –1.95), while it was 1.07 those without diabetes (0.92 [0.84 – sion in people with impaired glucose tol-
(0.98 –1.17) for people aged ⱖ51 years. 1.00]) (Table 2). erance supports this notion (35).
Multivariate analyses to control for The strengths of this study include its
number of physician visits (five or more), CONCLUSIONS — The association population-based case-control design,
age as a continuous and quadratic vari- between depression and diabetes is be-
the large sample size, and its detailed data
able, and sex resulted in an adjusted OR coming well recognized, but the temporal
on medication use. Compared with cur-
of 1.47 (95% CI 1.14 –1.90) (Table 1). relationship is less well understood (11–
rently available longitudinal literature
Without the quadratic term for age, the 25). In this large population-based study,
evaluating the risk of diabetes in people
adjusted OR for depression was 1.60 we observed an increased risk of develop-
with depression or depressive symptoms
(1.26 –2.04). In the full multivariate ing diabetes in people with a previous ep-
(15–24), our study had a much larger
model, however, there was statistically isode of depression. This relationship
sample size, was population-based, and
had many more cases of diabetes, which
Table 1—Adjusted OR for new-onset diabetes with history of depression allows for more confidence in our results.
Also, our study used previously validated
␤ SE P OR (95% CI)
measures, which included ICD-9 diag-
nostic codes and medications to identify
Prior depression 0.384 0.131 0.003 1.47 (1.14–1.90) individuals with major depressive disor-
Sex (male) 0.440 0.016 ⬍0.001 1.55 (1.50–1.60) der and diabetes, whereas the majority of
Age 0.179 0.003 ⬍0.001 1.20 (1.19–1.20) currently available studies used self-
Age2 ⫺0.001 0.000 ⬍0.001 0.99 (0.99–0.99) reported measures to identify diabetes
Age ⫻ prior depression ⫺0.007 0.002 0.002 0.99 (0.99–0.99) and depressive symptoms. Lastly, only
Number of physician visits* 1.052 0.017 ⬍0.001 2.86 (2.77–2.96) one study controlled for the possibility of
*Dichotomous variable based on median split of five physician visits in the year prior to study index date. surveillance bias influencing the identifi-

DIABETES CARE, VOLUME 28, NUMBER 5, MAY 2005 1065

Depression and diabetes

Table 2—Unadjusted and adjusted risk of developing diabetes stratified by age

Aged 20–50 years Aged ⱖ51 years

Unadjusted Adjusted* Unadjusted Adjusted*
OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Prior depression 1.76 (1.58–1.95) 1.23 (1.10–1.37) 1.07 (0.98–1.17) 0.92 (0.84–1.00)
Sex (male) 1.15 (1.10–1.21) 1.64 (1.54–1.71) 1.44 (1.39–1.50) 1.59 (1.54–1.66)
Number of physician visits† 3.13 (2.98–3.29) 3.53 (3.35–3.72) 2.45 (2.35–2.55) 2.61 (2.50–2.71)
*Adjusted for other variables in the table. †Dichotomous variable based on median split of five physician visits in the year prior to study index date.

cation of diabetes in people with depres- in the estimates of the risk of developing mechanisms through which these ill-
sion (16). diabetes when physician visits were in- nesses are related is required.
As with other studies based solely on cluded in the adjusted analyses. Previous
administrative databases, however, there studies, particularly those with greater es-
are several limitations that must be recog- timates of risk, neglected or were unable Acknowledgments — This study was funded
by a grant from the Canadian Institutes of
nized. It is likely that our figures under- to control for potential surveillance bias,
Health Research (CIHR). This work was also
estimate the prevalence of all and it is clearly an important factor in un- supported by a New Emerging Team (NET)
symptomatic conditions, including “un- derstanding these associations (15–25). Grant to the Alliance for Canadian Health Out-
diagnosed” diabetes, as patients with As we were not able to investigate comes Research in Diabetes (ACHORD). The
milder presentations are less likely to seek whether depression increases the risk of ACHORD NET Grant is sponsored by the Ca-
treatment or be admitted to the hospital other chronic medical conditions during nadian Diabetes Association, the Heart and
and therefore would not be captured in our analysis, the potential exists that dia- Stroke Foundation of Canada, The Kidney
the databases. Also, due to the stigma as- betes is not isolated in its associated in- Foundation of Canada, the CIHR Institute of
sociated with mental illness, many people creased risk in people with depression. It Nutrition, Metabolism and Diabetes, and the
are reluctant to seek treatment for depres- may be that prior depression results in CIHR Institute of Circulatory and Respiratory
Health.
sive disorders. This has likely resulted in an unmasking of other chronic medical
S.R.M. holds a Population Health Investiga-
an underestimation of individuals with conditions; such relationships should be tor Award through the Alberta Heritage Foun-
depression. We have no reason to believe, explored. Regardless, our analysis dem- dation for Medical Research (AHFMR) and is a
however, that this underestimation onstrated an increased risk of diabetes in New Investigator of the Canadian Institute of
would be systematically different between people with treated depression that may Health Research. J.A.J. is a Health Scholar with
case and control subjects. require increased vigilance in this popu- the AHFMR and a Canada Research Chair in
An additional drawback to the use of lation, particularly in those aged ⬍51 Diabetes Health Outcomes. This study is
administrative data is the lack of clinical years. This would suggest that the risk of based in part on deidentified data provided by
data. This limits our ability to investigate diabetes in young people with depression the Saskatchewan Department of Health. The
whether severity or other comorbidities should be considered over and above interpretation and conclusions contained
herein do not necessarily represent those of
associated with depression were present. other usual risk factors, such as family
the Government of Saskatchewan or the
We also recognize the potential for sur- history, sedentary lifestyle, and BMI. Saskatchewan Department of Health.
veillance bias in the recognition and diag- Finally, with this nested case-control
nosis of diabetes in this study. For design, we are unable to determine the APPENDIX —Oral antidiabetic agents:
example, someone with depression may corollary risk in the association between metformin, glyburide, gliclazide, rosiglita-
be more likely to have clinical work-ups diabetes and depression. That is, does zone, repaglenide, pioglitazone, nateglin-
such as a random or fasting blood glucose having type 2 diabetes increase the risk of ide, acarbose, acetohexamide, tolbutamide,
level upon diagnosis of depression com- developing depression? Addressing this and chlorpropamide.
pared with someone without medical ill- question would require a cohort study of
ness, increasing the chance of diabetes new-onset diabetes. Such a study, in com-
being diagnosed. To address this poten- bination with the results of this nested References
tial confounding, we controlled for the case-control analysis, would provide im- 1. Parikh SV, Lam RW, the CANMAT De-
number of physician visits in the year be- portant information on the temporal pression Workgroup: Clinical guidelines
fore study index date (32). We felt that relationship between diabetes and for the treatment of depressive disorders.
this variable would control to some extent depression. Can J Psychiatry 46 (Suppl. 1):1S–90S,
for both comorbidity and medical surveil- We observed that younger individu- 2001
2. Remick RA: Diagnosis and management
lance, as one would expect someone with als with a history of depression had a of depression in primary care: a clinical
multiple medical conditions or a severe greater risk of developing new-onset type update and review. CMAJ 167:1253–
single condition to have a higher number 2 diabetes compared with individuals 1260, 2002
of physician visits compared with an in- who had never been depressed. Further 3. Weissman MM, Bruce ML, Leaf PJ, Florio
dividual without illness. We observed this investigation into the relationship be- LP, Holzer C: Affective disorders. In Psy-
to be the case, with a considerable change tween depression and diabetes and the chiatric Disorders in America. LN Robins,

1066 DIABETES CARE, VOLUME 28, NUMBER 5, MAY 2005

 Brown and Associates

DA Regier, Eds. New York, Free Press, pression as a risk factor for incident dia- Osei W, Nichol J: Health databases in
p. 53– 80, 1991 betes: findings from the National Health Saskatchewan. In Pharmacoepidemiology.
4. Bland R, Newman S, Orn H: Period prev- and Nutrition Examination Epidemio- 3rd ed. Strom BL, Ed. Chichester, U.K.,
alence of psychiatric disorders in Ed- logic Follow-Up Study, 1971–1992. Am J Wiley, p. 325–345, 2000
monton. Acta Psychiatrica Scandinavia 77 Epidemiol 158:416 – 423, 2003 27. Johnson JA, Majumdar SR, Simpson SH,
(Suppl. 338):33– 42, 1988 16. Eaton WW, Armenian H, Gallo J, Pratt L, Toth EL: Decreased mortality associated
5. Beaudet MP: Depression. Health Rep 7:11– Ford DE: Depression and the risk for on- with metformin use compared to sulfo-
24, 1996 set of type II diabetes. Diabetes Care 19: nylurea monotherapy in type 2 diabetes
6. Offord DR, Boyle MH, Campbell D, 1097–1102, 1996 mellitus. Diabetes Care 25:2244 –2248,
Goering P, Lin E, Wong, Racine YA: One- 17. Hill Golden S, Paton Sanford C, Williams 2002
year prevalence of psychiatric disorder in JE, Nieto FJ, Ford DE, Brancati FL, Yeh 28. Blanchard JF, Ludwig S, Wajda A, Dean
Ontarians 15 to 64 years of age. Can J HC: Depressive symptoms and the risk of H, Anderson K, Kendall O, Depew N: In-
Psychiatry 41:559 –563, 1996 type 2 diabetes. Diabetes Care 27:429 –
cidence and prevalence of diabetes melli-
7. Barefoot JC, Schroll M: Symptoms of de- 435, 2004
tus in Manitoba, 1986 –1991. Diabetes
pression, acute myocardial infarction, and 18. Kawakami N, Takatsuka N, Shimizu H,
total mortality in a community sample. Ishibashi H: Depressive symptoms and Care 19:807– 811, 1996
Circulation 93:1976 –1980, 1996 occurrence of type 2 diabetes among Jap- 29. Hux JE, Ivis F, Flintoff V, Bica A: Diabetes
8. Crockett AJ, Cranston JM, Moss JR, Alpers anese men. Diabetes Care 22:1071–1076, in Ontario: determination of prevalence
JH: The impact of anxiety, depression, 1999 and incidence using a validated adminis-
and living alone in chronic obstructive 19. Palinkas LA, Lee PP, Barrett-Connor E: A trative data algorithm. Diabetes Care 25:
pulmonary disease. Qual Life Res 11:309 – prospective study of type 2 diabetes and 512–516, 2002
316, 2002 depressive symptoms in the elderly: the 30. Simpson SH, Jacobs P, Corabian P, John-
9. Everson SA, Roberts RE, Goldberg DE, Rancho Bernardo Study. Diabet Med 21: son JA: The cost of major co-morbidities
Kaplan GA: Depressive symptoms and 1185–1191, 2004 in a cohort of Saskatchewan residents
increased risk of stroke mortality over a 20. Arroyo C, Colditz GA, Hu FB, Speizer FE, with diabetes, 1996. CMAJ 168:1661–
29-year period. Arch Int Med 158:1133– Ryan LM, Manson J, Kawachi I: Depres- 1667, 2003
1138, 1998 sive symptoms and risk of type 2 diabetes 31. West SL, Richter A, Melfi CA, McNutt M,
10. Krantz DS, Sheps DS, Carney RM, Natel- in women. Diabetes Care 27:129 –133, Nennstiel ME, Mauskopf JA: Assessing
son BH: Effects of mental stress in pa- 2004 the Saskatchewan Database for outcomes
tients with coronary artery disease: evi- 21. Kumari M, Head J, Marmot M: Prospec- research studies of depression and its
dence and clinical implications. JAMA tive study of social and other risk factors treatment. J Clin Epidemiol 53:823– 831,
283:1800 –1802, 2002 for incidence of type 2 diabetes in the 2000
11. Anderson RJ, Clouse RE, Freedland KE, Whitehall II Study. Arch Intern Med 164: 32. Reid R, Evans R, Barer M, Sheps S, Ker-
Lustman PJ: The prevalence of comorbid 1873–1880, 2004 luke K, McGrail K, Hertzman C, Pagliccia
depression in adults with diabetes: a 22. Everson-Rose SA, Torrens JI, Meyer PM, N: Conspicuous consumption: character-
meta-analysis. Diabetes Care 24:1069 – Kravitz HM, Powell LH, Bromberger JT, izing high users of physician services in
1078, 2001 Pandey D, Matthews KA: Depressive one Canadian province. J Health Serv Res
12. Egede LE, Zheng D, Simpson K: Comor- symptoms, insulin resistance, and risk of Policy 8:215–224, 2003
bid depression is associated with in- diabetes in women at midlife. Diabetes
33. Weissman MM, Bland RC, Canino GJ,
creased health care use and expenditures Care 27:2856 –2862, 2004
Faravelli C, Greenwald S, Hwu HG, Joyce
in individuals with diabetes. Diabetes Care 23. van der Akker M, Schuurman A, Metse-
PR, Karam EG, Lee CK, Lellouch J, Lepine
25:464 – 470, 2002 makers J, Buntinx F: Is depression related
13. Finkelstein EA, Bray JW, Chen H, Larson to subsequent diabetes mellitus? Acta Psy- JP, Newman SC: Cross-national epidemi-
MJ, Miller K, Tompkins C, Keme A, chiatr Scand 110:178 –183, 2004 ology of major depression and bipolar
Manderscheid R: Prevalence and costs of 24. Saydah SH, Brancati FL, Golden SH, disorder. JAMA 276:293–299, 1996
major depression among elderly claim- Fradkin J, Harris MI: Depressive symp- 34. Spaner D, Bland RC, Newman SC: Epide-
ants with diabetes. Diabetes Care 26:415– toms and the risk of type 2 diabetes in a miology of psychiatric disorders in Edm-
420, 2003 US sample. Diabetes Metab Res Rev 19: onton: major depressive disorder. Acta
14. Nichols GA, Brown JB: Unadjusted and 202–208, 2003 Psychiatr Scand Suppl 376:7–15, 1994
adjusted prevalence of diagnosed depres- 25. Kessing LV, Nilsson FM, Siersma V, 35. Timonen M, Laakso M, Jokelainen J,
sion in type 2 diabetes. Diabetes Care 26: Andersen PK: Increased risk of develop- Rajala U, Meyer-Rochow VB, Keinanen-
744 –749, 2003 ing diabetes in depressive and bipolar dis- Kiukaanniemi S: Insulin resistance and
15. Carnethon MR, Kinder LS, Fair JM, Staf- orders? J Psychiatr Res 38:395– 402, 2004 depression: cross-sectional study. BMJ
ford RS, Fortmann SP: Symptoms of de- 26. Downey W, Beck P, McNutt M, Stang MR, 330:17–18, 2005

DIABETES CARE, VOLUME 28, NUMBER 5, MAY 2005 1067