Data were from the Whitehall II study (9
). In the 2003 and 2004 data collection phase, 4,228 men and women aged 50–74 years completed a depression questionnaire and, if without known diabetes, underwent an oral glucose tolerance test (OGTT). Venous blood samples were taken after at least 8-h fasting, before OGTT, and at 2-h postadministration of a 75 g glucose solution. Blood glucose was measured using the glucose oxidase method (10
) on a YSI MODEL 2300 STAT PLUS Analyzer (YSI Corporation, Yellow Springs, OH; mean coefficient of variation [CV] 1.4–3.1%) (11
). Diabetes was defined by a fasting glucose ≥7.0 mmol/l, a 2-h postload glucose ≥11.1 mmol/l, reported doctor-diagnosed diabetes, or use of diabetes medication (12
). In nondiabetic participants, we classified moderate hyperglycemia as IFG (fasting glucose 5.6–6.9 mmol/l) and impaired glucose tolerance (IGT) (2-h postload glucose 7.8–11.0 mmol/l) (12
). A1C was measured in whole blood with a calibrated high-performance liquid chromatography system (CV 0.8%).
Depressive symptoms were assessed with the Center for Epidemiologic Studies Depression scale (CES-D) summary score (13
), a measure that has been validated among diabetic patients (14
). Ethnicity (Caucasian or non-Caucasian), BMI (weight in kilograms divided by height in meters squared [kg/m2
]), waist circumference (cm), systolic and diastolic blood pressure (mmHg), HDL and LDL cholesterol (mmol/l), triglycerides (mmol/l), current smoking (yes/no), alcohol consumption (none, 1–3 units per day, or >3 units per day), and physical inactivity (<2.5 h moderate and >1 h vigorous exercise per week) were measured according to standardized protocols.
Statistical analyses are based on between 3,945 (93% of the 4,228) and 4,228 (100%) of the eligible participants because there were missing data for some of the various glucose parameters. Values for triglycerides were log transformed prior to analyses because of skewed distribution. We used linear regression analysis to model the associations of IFG and IGT with the CES-D score. The shape of the associations between fasting glucose, postload glucose, A1C, and CES-D were studied by treating all these measures as continuous variables. To test for curvilinear trends, a squared term of the glycemic measures was added to an equation containing the linear term. The models were adjusted for age, sex, ethnicity, and clinical characteristics (BMI, waist circumference, systolic and diastolic blood pressure, HDL and LDL cholesterol, triglycerides, smoking status, alcohol consumption, and physical activity). Subsidiary analyses examined these associations in subgroups and in the whole cohort with glucose levels measured repeatedly (in 1997–1999 and in 2003–2004) and with a dichotomized CES-D score (<16 vs. ≥16). The statistical tests were performed with STATA version 10.1.