Ten-year incidence of diabetes in older Australians: the Blue Mountains Eye Study

The BMES was conducted in an older community within a geographically defined area, west of Sydney. The study was approved by the Western Sydney Area Human Research Ethics Committee and was conducted in adherence to the tenets of the Declaration of Helsinki.

Details of the BMES methods were previously reported.14,15 The study population was representative of the Australian population for demographic and socioeconomic status, although slightly older on average. After a private door-to-door census of 38 Australian Bureau of Statistics (ABS) Census Collection Districts of two postcodes in the Blue Mountains region of New South Wales (2780 and 2782), all permanent, non-institutionalised residents with birthdates before 1 January 1943 were invited to participate. Of the 4433 eligible residents, 3654 (82.4%) participated in baseline examinations (BMES I) during 1992–1994. A comparison of characteristics of participants and non-participants was previously reported.15 Surviving baseline participants were invited to attend follow-up examinations after 5 years (1997–1999) and 10 years (2002–2004).

All three examinations used similar survey methods. A detailed questionnaire was administered by trained interviewers and included demographic information, eye and general medical history, and medications used. All participants were asked to undertake a fasting blood test within a month of the interview to assess fasting blood glucose, lipids, creatinine and fibrinogen concentrations. Blood was collected in fluoride/ oxalate tubes and centrifuged on site within 1 hour to separate plasma. Samples were then refrigerated before transportation on ice by courier within 4 hours of collection to the Institute of Clinical Pathology and Medical Research, Westmead Hospital, Sydney, for blood glucose analysis by the hexokinase method.

Criteria for a diagnosis of diabetes were:

• self-reported diabetes history and current use of diabetic medications, or

• a fasting plasma glucose concentration ≥ 7.0 mmol/L, according to the World Health Organization diabetes classification.3

IFG was defined as:

• fasting plasma glucose concentration 5.6–6.9 mmol/L among participants not using diabetes medications, as recommended by the American Diabetes Association (ADA).16

Diabetes incidence was defined in participants who were free of diabetes at baseline but who developed diabetes, diagnosed before or at the 5- or 10-year follow-up examination. Similarly, IFG incidence was defined in participants who were free of diabetes or IFG at baseline but who developed IFG, diagnosed before or at the 5- or 10-year follow-up examination.

The 2003 WHO/International Society of Hypertension guidelines17 were used to classify participants as:

• pre-hypertensive if systolic blood pressure (BP) was 120–139 mmHg or diastolic BP was 80–89 mmHg;

• hypertensive stage I if systolic BP was 140–159 mmHg or diastolic BP was 90–99 mmHg; and

• hypertensive stage II or higher if the individual was previously diagnosed as hypertensive and was using antihypertensive medications or had a systolic BP ≥ 160 mmHg or diastolic BP ≥ 100 mmHg at examination.

Obesity was defined as body mass index (BMI) ≥ 30 kg/m2. Participants were classified as smokers if they currently smoked or had stopped smoking less than 12 months before examination.

Total cholesterol, triglycerides and high density lipoprotein (HDL) were measured from fasting blood samples using a 747 Biochemistry Analyzer (Hitachi, Tokyo, Japan). Hypercholesterolaemia was defined as a serum cholesterol level ≥ 5.2 mmol/L, hypertriglyceridaemia as serum triglyceride ≥ 2.0 mmol/L, and abnormal HDL cholesterol as ≤ 1.0 mmol/L.

Renal function was estimated based on serum creatinine level and creatinine clearance (using the Cockcroft–Gault formula). Adjusted creatinine clearance was then calculated as creatinine clearance × 1.73/body surface area.18 Renal impairment was considered present if the adjusted creatinine clearance was < 60 mL/min/1.73 m2. All participants were asked about a history of angina, acute myocardial infarction, stroke or gout.

The metabolic syndrome was defined according to the new International Diabetes Federation definition19 as central obesity (defined as BMI > 30 kg/m2) plus any two of the following four factors:

• serum triglyceride level ≥ 1.7 mmol/L, or specific treatment for this lipid abnormality;

• serum HDL cholesterol < 1.03 mmol/L in men and < 1.29 mmol/L in women, or specific treatment for this lipid abnormality;

• systolic BP ≥ 130 mmHg or diastolic BP ≥ 85 mmHg, or treatment of previously diagnosed hypertension; or

• fasting plasma glucose ≥ 5.6 mmol/L.

A history of diabetes in the participant’s family was sought during the study interview. First-degree relatives included parents and siblings, while second-degree relatives included both maternal and paternal relatives. Participants with unknown family history of diabetes in both parents were considered to have no family history of diabetes.

Data were analysed using SAS version 9.1 (SAS Institute Inc, Cary, NC, USA). Ten-year cumulative incidence rates were calculated using the Kaplan–Meier (product-limit) method, which allowed participants who were censored at the 5-year follow-up examination to contribute to the estimates. Multivariate-adjusted discrete logistic models (by means of the PHREG SAS procedure) were used to assess factors associated with incident diabetes, using three time points when presence or absence of the outcome event was recorded.20 Participants with incomplete data were excluded from analyses. Variables tested for association with diabetes incidence included fasting plasma glucose level, obesity, serum cholesterol level, serum HDL cholesterol level, serum triglycerides, current smoking, hypertension and family history of diabetes. Trends for increasing incidence with increasing age were tested for significance using the Mantel–Haenszel χ2 test. Odds ratios and 95% CIs are presented. Tests for interaction between sex and selected risk factors (family history of diabetes, HDL and obesity) showed they were not significant.

Of the 3654 baseline participants, 2335 (75.1% of survivors) returned to the 5-year follow-up examination, and 1952 (75.6% of survivors) to the 10-year follow-up examination. Including participants seen at either or both of these examinations, 2564 participants were followed up after baseline. Surviving participants who did not return at either the 5- or 10-year follow-up examinations were more likely to have been younger than 60 years of age and current smokers at baseline.

Exclusions were 163 participants with diabetes at baseline, 52 with missing diabetes data or fasting blood samples at baseline, and 226 with no fasting blood tests at either the 5- or 10-year examinations. Thus, 2123 participants (1242 women and 881 men) had complete data available and were included in the assessment of 10-year incident diabetes.

The 10-year incidence of diabetes in the study population estimated by the Kaplan–Meier method was 9.3% (n = 165). Of these 165 participants, 52 (31.5%) were diagnosed solely on the basis of an elevated fasting plasma glucose level at examination. The remaining 113 (68.5%) gave a history of physician-diagnosed diabetes and were receiving treatment (insulin, oral therapy or diet).

More women (incidence, 10.0%, n = 105) than men (8.2%, n = 60) developed diabetes, although this difference was not statistically significant (P = 0.20). The age-specific incidence of diabetes was 9.1%, 8.6%, and 11.2%, in the age groups < 60, 60–69 and 70 + years, respectively (P for age-related trend = 0.73). After multivariate adjustment, baseline plasma glucose level, low HDL cholesterol, presence of obesity, current smoking and family history of diabetes were all significantly associated with an increased risk of diabetes, but presence of hypertension was not (Box 1). Past history of angina, acute myocardial infarction, stroke or gout were also not found to be associated with diabetes incidence, nor was body surface area-adjusted creatinine clearance (data not shown).

The 10-year cumulative diabetes incidence was substantially higher among the 203 participants who had metabolic syndrome at baseline (28.1%) than among those without metabolic syndrome (7.6%). Box 2 shows the effect of metabolic syndrome traits on the incidence of diabetes. The risk of diabetes increased with a greater number of metabolic syndrome traits, in addition to a BMI > 30 kg/m2. A two-, four- and eightfold increased risk of diabetes was associated with the presence of two, three and four metabolic syndrome traits, respectively, in addition to a BMI > 30 kg/m2.

The 10-year incidence of diabetes was higher in participants with IFG at baseline (30.0%) than in those with normoglycaemia (6.5%). Of the 194 participants with IFG at baseline who did not develop diabetes, 94 (48.5%) were normoglycaemic at the 5-year examination, and 15 (7.7%) at the 10-year examination, using the new ADA definition. Of the 94 who were normoglycaemic at the 5-year examination, nine (9.6%) reverted to IFG at the 10-year examination. The corresponding return rates from IFG at baseline to normoglycaemia at follow-up examinations using the previous ADA definition (6.1–6.9 mmol/L) were also assessed: 34/50 (68%) became normoglycaemic at the 5-year examination, and two (4%) at the 10-year examination. Of the 34 who returned to normoglycaemia after 5 years, three (9%) developed IFG at the 10-year examination.

A history of diabetes in first-degree relatives, or a paternal history of diabetes, was also a significant predictor of incident diabetes (Box 3).

After excluding participants with diabetes or IFG at baseline and those with missing fasting blood samples, 1756 participants were considered at risk of incident IFG. The overall 10-year incidence of IFG was estimated as 15.8% (17.2% in men versus 14.7% in women; P = 0.11). The age-specific IFG incidence was 11.0%, 16.0%, and 18.6% in participants aged < 60, 60–69, and 70 + years, respectively (P for age-related trend = 0.03). Higher baseline plasma glucose level, current smoking, presence of obesity and stage II hypertension were all associated with increased IFG incidence (Box 4). However, family history of diabetes, low HDL cholesterol level, and high total cholesterol or triglyceride levels were not significantly associated with incident IFG, nor was history of gout, angina, stroke, or acute myocardial infarction, or the presence of renal impairment (data not shown).

Using the previous ADA definition (fasting plasma glucose, 6.1–6.9 mmol/L),21 the 10-year IFG incidence was 5.7%. Risk factors significantly associated with incident IFG of this definition included higher plasma glucose levels, stage II hypertension and hypertriglyceridaemia.