Glucose Indices, Health Behaviors, and Incidence of Diabetes in Australia: The Australian Diabetes, Obesity and Lifestyle Study

The AusDiab baseline study methods and response rates are described in detail elsewhere (17,18). In brief, the baseline study was a cross-sectional, national, population-based survey of 11,247 adults aged ≥25 years in 1999–2000. Over 85% of the sample was from an Australian, New Zealand, or British background. A stratified cluster sample was drawn from 42 randomly selected census collector districts across Australia. Information was collected using a brief household interview, followed by a biomedical examination. The response rate to the baseline biomedical examination was 55%. In 2004–2005, all living eligible participants were invited to attend follow-up. Those who were considered ineligible included those who refused further contact (n = 128), were deceased (n = 310), had moved overseas or into a nursing facility classified for high care, or had a terminal illness (n = 21).

At baseline and follow-up, the physical examination included blood samples, anthropometric measurements, and questionnaires. All participants except for those currently receiving treatment for diabetes or who were pregnant underwent a standard 75-g OGTT (19). In 1999–2000, fasting plasma glucose (FPG) and 2-h plasma glucose (2hPG) levels were determined by a glucose oxidase method using an Olympus AU600 automated analyser (Olympus Optical, Tokyo, Japan), and in 2004–2005 a spectrophotometric-hexokinase method utilizing a Roche Modular (Roche Diagnostics, Indianapolis, IN) was used. To compare results from the two assays, 195 stored baseline FPG samples and 171 2hPG samples were analyzed on the Roche assay used at follow-up. The median (10th–90th percentile) difference between the pairs of FPG values was 0.2 mmol/l (0.1–0.4), and the median (10th–90th percentile) difference between the pairs of 2hPG values was 0.2 mmol/l (0.0–0.4), with higher values from the baseline assay. Total glycated hemoglobin analysis used high-performance liquid chromatography (Bio-Rad Variant Hemoglobin Testing System; Bio-Rad, Hercules, CA) with standardized conversion to A1C values. Blood pressure was measured using Dinamap or a standard mercury sphygmomanometer with appropriate adjustments made as previously described (20). The study was approved by the International Diabetes Institute Ethics Committee.

Glucose tolerance status was classified according to the 1999 World Health Organization criteria (19). Diabetes was classified on the basis of FPG ≥7.0 mmol/l or 2hPG ≥11.1 mmol/l or current treatment with insulin or oral hypoglycemic agents. For those not reporting treatment for diabetes, FPG <7.0 mmol/l and 2hPG ≥7.8 mmol/l but <11.1 mmol/l indicated IGT, FPG 6.1–6.9 mmol/l and 2hPG <7.8 mmol/l indicated IFG, and both FPG <6.1 mmol/l and 2hPG <7.8 mmol/l indicated normal glycemia. An incident case of diabetes was defined as an individual who had normal glycemia, IFG, or IGT at baseline but had developed diabetes at follow-up. Incident cases of IFG were defined as people who had normal glycemia at baseline but had developed IFG at follow-up. Incident cases of IGT were defined as people who had normal glycemia or IFG at baseline but had developed IGT at follow-up.

Hypertriglyceridemia was defined as serum triglycerides ≥2.0 mmol/l, and abnormal HDL cholesterol was defined as HDL cholesterol <1.03 and <1.29 mmol/l for men and women, respectively. Hypertension was defined as having systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg or reporting antihypertensive medication use.

Data on education, smoking, leisure-time physical activity, and family history of diabetes were collected by an interviewer-administered questionnaire. Education was classified into four categories: 1) university/further education, 2) completed secondary school, 3) some secondary school, or 4) primary school/never attended school. Smoking history (current smoker, past smoker, or never smoked) was collected using a questionnaire on tobacco use, which has been validated in Australian adults (21). Total leisure-time physical activity reported for the previous week (none, insufficient: 1–149 min; sufficient: ≥150 min) was measured using the Active Australia questionnaire, which is the standard instrument for population surveillance and has acceptable reliability (22,23). Family history of diabetes was defined as a mother or father being diagnosed as having diabetes.

Five-year cumulative incidence was calculated and standardized to the 1998 Australian population using the direct method (7). In brief, age- and sex-specific 5-year cumulative incidences of diabetes were applied to the equivalent age and sex strata from the Australian population of 1998, who were free of diabetes. The 1998 Australian diabetes-free population was calculated by applying age- and sex-specific diabetes prevalence estimates from the baseline AusDiab data to the 1998 Australian population and subtracting the diabetic population from the total population to give a nondiabetic population. The incidence of IFG and IGT were age and sex standardized using the same method but applying the relevant IFG- and IGT-free populations rather than the diabetes-free population. Annual incidence (% per year) was calculated from the 5-year cumulative incidence by applying the following formula: [−ln(1 − S)]/t; where S is the proportion of new cases over t years and t equals the time of follow-up. Baseline characteristics of the attendees (n = 6,537) were compared with those of the nonattendees (n = 4,710) with an independent t test, Mann-Whitney U test, or Pearson's χ² test, where appropriate. The incidence of diabetes is restricted to those who did not have diabetes at baseline (n = 5,842) but developed diabetes at follow-up.

Risk factors for diabetes were identified by multivariate logistic regression. The covariates included in the model were age, sex, waist circumference, smoking status, FPG, 2hPG, A1C, hypertension, physical activity, education level, family history of diabetes, total cholesterol, HDL cholesterol, and triglycerides. Triglycerides, 2hPG, FPG, and A1C were log transformed. Since waist circumference was a better predictor of diabetes than BMI, it was used as the measure of adiposity in the model. The risk of diabetes associated with age was calculated per 10 years, and for waist circumference, FPG, 2hPG, and A1C, the risk of diabetes was estimated per 1 SD. In models comparing the incremental value of 2hPG and A1C to FPG, the test procedure was based on maximum likelihood estimators, Akaike's information criterion (AIC statistic), and the area under the receiver-operating characteristic curve. Analyses were conducted using Stata statistical software version 9.2 (StataCorp, College Station, TX).

Of 10,788 participants eligible for testing in 2004–2005, 6,400 (59.3%) attended the full physical examination. A further 137 (1.3%) participants had blood and urine tests only, and another 2,261 (21.0%) completed a telephone questionnaire only. Thus, the response rate for the follow-up was 60.6% (6,537 of 10,788).

Compared with those who did not attend (n = 4,710), attendees (n = 6,537) were significantly less likely to be hypertensive, to have a lower level of education attainment, or to be smokers and had lower 2hPG, A1C, and smaller waist circumferences at baseline. Furthermore, a significantly larger proportion of nonattendees had abnormal triglyceride or HDL cholesterol levels at baseline compared with attendees (data not shown).

Of the population at risk of diabetes (n = 5,842), 54.3% were female and 88.5% were of Australian, New Zealand, or British background. The mean (range) for age and BMI was 50.9 years (25–88) and 26.7 kg/m² (14.5–59.4), respectively. The baseline characteristics of those who developed diabetes and those who did not are described in Table 1. Table 2 describes the incidence of diabetes, IFG, and IGT. At follow-up, there were 224 new cases of diabetes. More men (4.4%) than women (3.4%) developed diabetes (P = 0.09). The incidence of diabetes among those with IFG was significantly higher in women than in men, while the opposite applied to those with IGT at baseline. Among those with IGT, the excess risk of diabetes among men (adjusted for age and waist circumference) was attenuated and became nonsignificant with adjustment for baseline FPG. Among those with IFG, the excess risk of diabetes among women (adjusted for age) was reduced and was no longer significant after adjustment for baseline waist circumference and was further attenuated, following adjustment for baseline 2hPG (data not shown).

When the fasting glucose values from follow-up were adjusted via a regression equation (derived from comparing the 195 samples from baseline that were retested at follow-up), 244 new cases of diabetes (compared with 224 new cases of diabetes observed in the main analysis) occurred over follow-up. The annual incidence of self-reported diabetes was 0.5% (95% CI 0.4–0.6) in the 6,537 (5,842 free of diabetes) attendees in the main analysis and 0.5% (0.4–0.7) in the 2,261 (1,990 free of diabetes) nonattendees who only completed health questionnaires. After adjusting for age and sex, the odds of self-reporting incident diabetes remained similar in these two groups (odds ratio 1.0 [95% CI 0.7–1.4]).

Table 3 shows that waist circumference, current smoking, primary school level of education only/never attended school, hypertension, family history of diabetes, FPG, and high triglycerides were positively associated with incident diabetes. An increased risk of diabetes was observed in those who did not achieve the public health recommendations for physical activity (the inactive and the insufficiently active), compared with those who achieved the recommended levels (sufficiently active: ≥150 min/week). Low HDL cholesterol was only associated with incident diabetes in adjusted models, which did not include triglycerides. Each of the three measures of glycemia had significant relationships with incident diabetes that were similar to each other in magnitude. These relationships remained significant among those with normal glycemia and among those with IFG or IGT (data not shown), with similar and significant results in men and women.

To determine the additional value of 2hPG and A1C to other risk factors, these parameters were incorporated into to a model comprising FPG, waist circumference, age, sex, triglycerides, hypertension, education level, family history of diabetes, physical activity category, and smoking status. The sequential addition of 2hPG and A1C significantly improved the maximum likelihood ratio (2hPG: χ² = 131.8, P < 0.001) (A1C: χ² = 48.1, P < 0.001) and reduced the AIC statistic of the model (data not shown). The basic model with FPG yielded an area under the curve of 0.82 (95% CI 0.79–0.85), which increased by 0.05 and 0.01 by the addition of 2hPG and A1C, respectively. The differences in area under the curve upon the addition of 2hPG (P < 0.001) and A1C (P = 0.003) were statistically significant at each point.

In a model including all three measures of glycemia (per 1 SD), as well as other independent risk factors, each measure was a significant, independent predictor of diabetes (A1C: odds ratio 2.9 [95% CI 2.1–3.9], FPG: 2.3 [1.7–3.2], and 2hPG: 3.0 [2.4–3.8]).

Despite the obvious importance of accurate diabetes data, there are no national, population-based studies reporting the incidence of diabetes in a developed country. There are, however, two previous studies reporting the incidence of diabetes in high-risk Australians. The incidence of diabetes in Australian Aborigines (24) and in older (aged ≥49 years) Australians from the Blue Mountains Eye Study (25) was 9.9% over 8 years and 9.3% over 10 years, respectively. The data from both of these studies equate to slightly higher incidences than we report here for AusDiab, in keeping with the higher risks of these populations. Compared with other recent studies reporting incidence of diabetes using an OGTT, our study has a slightly lower incidence than has been reported for South African Indians (26), a very similar incidence than has been reported in Italians (5), and a much lower incidence than that reported in high-risk populations such as Mauritians (16) and Pima Indians (7).

The incidence of diabetes was higher for those with IGT than for those with IFG. However, as shown in Table 3, the relationship between baseline FPG and incident diabetes was slightly stronger than was the relationship between baseline 2hPG and incident diabetes. This apparent inconsistency is due to the fact that the World Health Organization classifications include all those who have combined IFG and IGT within the IGT group, while IFG includes only those with isolated IFG and a completely normal 2hPG.

The incidence of diabetes in those with IFG was higher in women than in men, while at baseline, the prevalence of IFG was higher in men (18). Conversely, the incidence of diabetes in those with IGT was higher in men than in women, but the prevalence of IGT was slightly higher in women. It appears that although more women had IGT, men with IGT were more likely to convert to diabetes. In contrast, while more men had IFG, women with IFG were more likely to convert to diabetes. While this may appear counterintuitive, it is explained by sex differences in glucose levels. Among those with IGT, men have a higher FPG, leading to a higher risk of progressing to diabetes. Among those with IFG, women have a higher 2hPG, leading to a higher risk of progression to diabetes. We have previously reported these glucose differences (27) and observed that women have a larger difference between FPG and 2hPG than do men. This in itself may be due to the use of a fixed glucose load (75 g) in men and women, despite women being shorter than men.

Traits of the metabolic syndrome were associated with an increased risk of diabetes, as has been shown previously (5,13,14,28). Relationships with HDL cholesterol were less important in our study than in other studies (25).

The effect of age in the model of incident diabetes was attenuated after adjustment for other age-related covariates. This finding is consistent with other reports. Cugati et al. (25) showed that age was only weakly associated with incident diabetes in a cohort of older Australians. Similar to the findings in this study, family history of diabetes was also associated with incident diabetes (7,25). Consistent with data from other large recent prospective studies, smokers were found to be at increased risk of diabetes compared with nonsmokers (29,30).

In terms of other lifestyle factors, education level, a likely surrogate for socioeconomic status, was associated with increased risk of diabetes. Those having only a primary school level of education or who reported having no education were almost two times more likely to develop diabetes, and there was a significant trend across categories of education attainment, suggesting that risk of diabetes increased with lower levels of education. This finding confirms early research showing that a low level of education predicted diabetes in Mexicans Americans (31).

Our findings of an increased diabetes risk among those who are physically inactive confirm our earlier cross-sectional results (32–34), as well as longitudinal and interventional data from other studies (2,4,15). These findings further reinforce the message that the adoption and maintenance of a physically active lifestyle must be a fundamental element of any approach to prevent diabetes.

The strength of the association between A1C and incident diabetes was noteworthy. Despite the fact that diabetes is defined by blood glucose, A1C was as good a predictor of diabetes as was 2hPG, suggesting a possible clinical role for the measurement of A1C (when the methods are better standardized) in the prediction of diabetes risk and not just in the management of those with established diabetes.

The strengths of the current study are 1) inclusion of men and women of a wide age range; 2) a large national, population-based sample of Australian adults; 3) diagnosis of diabetes using an objective measure, the OGTT; and 4) evaluation of a large array of candidate risk factors of incident diabetes. This study is also not without limitations. The response rate at baseline was 55% (17), and only 60% of all participants were followed-up. Therefore, it could be said that the results may not be reflective of the whole Australian population. This, however, is offset by the observation that the incidence of self-report diabetes among 1,990 participants who only completed questionnaires (and were not included in the main analysis) was very similar to the incidence of diabetes in 5,842 participants in the main analysis. It is unlikely that the remaining ∼20% of participants for whom we do not have any follow-up information would have had an appreciable impact on the overall incidence of diabetes. If anything, the differences between attendees and nonattendees suggests that a healthy cohort bias may have occurred in this study and that these findings could be a slight underestimate of the true incidence of diabetes in Australia. Another limitation relates to the number of cases of diabetes (n = 224); it is possible that some effects of covariates may have been missed due to insufficient power.

This study provides the first national, population-based data on the incidence of diabetes in a developed country and, as such, includes information of significant utility in both clinical and health-planning settings. The findings confirm the high risk of those with IFG and IGT, as well as the importance of physical activity, adiposity, and cigarette smoking as well as other previously identified risk factors, while also suggesting the potential value of A1C in the prediction of diabetes.

In Australia, the prevalence of diabetes doubled from 1981 to 2000 (18), and it is projected to increase further by the year 2025 (35). Obesity has been identified as a key driver of diabetes in this study, and the expected increases in the prevalence of obesity (34) are likely to further increase the incidence and prevalence of diabetes, contributing significantly to the total burden of ill health in Australia.

E.L.M.B. is supported by a National Health and Medical Research Council (NHMRC) (no. 379305)/National Heart Foundation Australia (no. PP 05M 2346) joint postgraduate scholarship. D.W.D. is supported by a Victorian Health Promotion Foundation Public Health Research Fellowship. A.J.C. is supported by a National Heart Foundation Australia postgraduate scholarship (no. PP 04M 1794). The AusDiab study, coordinated by the International Diabetes Institute, gratefully acknowledges the generous support given by the NHMRC (grant no. 233200); the Australian Government Department of Health and Ageing; Abbott Australasia; Alphapharm; AstraZeneca; Aventis Pharma; Bristol-Myers Squibb; City Health Centre, Diabetes Service, Canberra; Department of Health and Community Services, Northern Territory; Department of Health and Human services, Tasmania; the Department of Health, New South Wales; the Department of Health, Western Australia; the Department of Health, South Australia; the Department of Human Services, Victoria; Diabetes Australia; Diabetes Australia Northern Territory; Eli Lilly Australia; Estate of the Late Edward Wilson; GlaxoSmithKline; the Jack Brockhoff Foundation; Janssen-Cilag; Kidney Health Australia; Marian & FH Flack Trust; Menzies Research Institute; Merck Sharp & Dohme; Novartis Pharmaceuticals; Novo Nordisk Pharmaceuticals; Pfizer; Pratt Foundation; Queensland Health; Roche Diagnostics Australia; Royal Prince Alfred Hospital, Sydney; and Sanofi Synthelabo.

Also, for their invaluable contribution to the set-up and field activities of AusDiab, we are enormously grateful to A. Allman, B. Atkins, S. Bennett, A. Bonney, S. Chadban, M. de Courten, M. Dalton, T. Dwyer, H. Jahangir, D. McCarty, A. Meehan, N. Meinig, S. Murray, K. O’Dea, K. Polkinghorne, P. Phillips, C. Reid, A. Stewart, H. Taylor, T. Whalen, and F. Wilson.