SEARCH

SEARCH BY CITATION

Keywords:

  • AusDiab;
  • diabetes;
  • epidemiology;
  • longitudinal cohort;
  • metabolic syndrome;
  • risk prediction

Abstract.

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References

Objective.  To compare the ability of the metabolic syndrome (MetS), a diabetes prediction model (DPM), a noninvasive risk questionnaire and individual glucose measurements to predict future diabetes.

Design.  Five-year longitudinal cohort study. Tools tested included MetS definitions [World Health Organization, International Diabetes Federation, ATPIII and European Group for the study of Insulin Resistance (EGIR)], the FINnish Diabetes RIsk SCore risk questionnaire, the DPM, fasting and 2-h post load plasma glucose.

Setting.  Adult Australian population.

Subjects.  A total of 5842 men and women without diabetes ≥25 years. Response 58%. A total of 224 incident cases of diabetes.

Results.  In receiver operating characteristic curve analysis, the MetS was not a better predictor of incident diabetes than the DPM or measurement of glucose. The risk for diabetes among those with prediabetes but not MetS was almost triple that of those with MetS but not prediabetes (9.0% vs. 3.4%). Adjusted for component parts, the MetS was not a significant predictor of incident diabetes, except for EGIR in men [OR 2.1 (95% CI 1.2–3.7)].

Conclusions.  A single fasting glucose measurement may be more effective and efficient than published definitions of the MetS or other risk constructs in predicting incident diabetes. Diagnosis of the MetS did not confer increased risk for incident diabetes independent of its individual components, with an exception for EGIR in men. Given these results, debate surrounding the public health utility of a MetS diagnosis, at least for identification of incident diabetes, is required.


Introduction

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References

The metabolic syndrome (MetS), a clustering of abnormalities thought to be related to one or both of insulin resistance and increased central adiposity, has increased in prominence over the last decade because of an increasing interest in its underlying cause as well as the possible utility of the concept in clinical practice. A major justification for a clinical definition of the MetS is to identify those at high risk of cardiovascular disease (CVD) and type 2 diabetes, leading to lifestyle or pharmacological intervention among those who might not otherwise be treated. Studies supporting the ability of the MetS to predict incident type 2 diabetes are based on cohorts in Finland, the United States, Scotland and China [1–9]. Those studies that used published definitions of the MetS include a population-based cohort of middle-aged men from Finland, a clinical trial population of middle-aged men from Scotland and a population-based cohort of 1734 from the San Antonio Heart Study (USA) [1, 2, 4, 5, 9]. Only the (U.S.) National Cholesterol Education Program (ATPIII) definition of the MetS has been compared with the San Antonio Heart Study Diabetes Prediction Model (DPM). A thorough assessment of whether the MetS is an improvement on other available risk constructs, including noninvasive risk questionnaires and the simple measurement of fasting glucose, in the prediction of incident diabetes has been lacking.

The national, population-based and longitudinal Australian Diabetes, Obesity and Lifestyle (AusDiab) study was conducted between 2000 and 2005. This study provides an opportunity to further address the question of whether the MetS is superior to published diabetes risk prediction devices, and in addition, to evaluate whether it provides predictive power for diabetes beyond that of its component parts and other known risk factors.

Research design and methods

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References

Survey procedures

The AusDiab study baseline methods and response have been described in detail elsewhere [10]. In brief, the AusDiab sample was drawn from 42 randomly selected Census Collector Districts across Australia, with the primary aim to determine the prevalence of diabetes in Australia. All noninstitutionalized, usual residents aged ≥25 years in each of the 42 districts was eligible for inclusion, In this stratified, clustered, population-based survey, 11 247 adults (5049 men) participated in 2000 (response 55.3% of those completing household interview). Five years later in 2005, 6537 of these participants were followed up (response 58%), with the primary aim of estimating the incidence of diabetes. Comparisons of responders and nonresponders to both the baseline and follow-up surveys have been reported previously [10, 11]. The annual incidence of self-reported diabetes was identical (0.5%) for those who attended for physical testing and for an additional 2261 nonattendees (1990 free of diabetes at baseline) who only completed self-reported health questionnaires. After adjusting for age and sex, the odds of self-reporting incident diabetes remained similar in these two groups [OR 1.0 (95% CI 0.7–1.4)]. After exclusion of those who could not be classified according to diabetes status for both baseline and follow-up (n = 247), a total of 6290 were available for analysis, of whom 448 had diabetes at baseline. Analyses involving incident diabetes were therefore conducted on 5842 participants.

At both baseline and 5-year follow-up, an oral glucose tolerance test (OGTT) was performed on all nonpregnant participants, except those taking insulin or oral hypoglycaemic drugs. In 2000, fasting plasma glucose (FPG) and 2 h post load plasma glucose (2hPG), fasting serum high-density lipoprotein cholesterol (HDL-C), triglycerides, total cholesterol and urinary albumin and creatinine were determined enzymatically using the Olympus AU600 analyser and insulin was assayed by radioimmunoassay (Linco Research Inc., St. Charles, MO). Insulin levels were only measured in those aged 35 and over (88.1% of the study population). In 2005, FPG and 2hPG were measured using a Roche Modular (Roche Diagnostics, Indianapolis, IN, USA) with a spectrophotometric-hexokinase method. Laboratory analysis methods for glucose were found to be comparable across the 2000 and 2005 surveys [11]. Retesting of stored baseline glucose samples after 5 years yielded a correlation coefficient of 0.97 (after removal of the 10% of samples collected on a single day in which 38% of retest results were <4.0 mmol L−1). Diabetes, impaired fasting glucose (IFG) and impaired glucose tolerance (IGT) were classified according to the World Health Organisation (WHO) criteria [12]. Height, weight, hip and waist circumference and blood pressure, were measured as previously described [10, 13, 14]. Body mass index (BMI) was calculated as weight (kg)/height (m2). Protocols were approved by the ethics committee of the International Diabetes Institute and Monash University’s Standing Committee on Ethics in Research involving Humans (SCERH). Written informed consent was obtained from all participants.

Definitions of the metabolic syndrome, the FINnish Diabetes RIsk SCore and the diabetes prediction model

Individuals were considered to have the MetS based on four published definitions [15–18], with the following exceptions: insulin resistance was defined as the lowest quartile of HOMA-S and the top quartile of fasting insulin in the nondiabetic population for the WHO and EGIR definitions respectively. Treatment for elevated tryglicerides or low HDL was not assessed, so was assumed to be absent for the International Diabetes Federation (IDF), EGIR and ATPIII definitions. The updated 2005 ATPIII criteria for elevated fasting plasma glucose were used (≥100 mg dL−1). Language spoken at home was used to define ethnic groups for obesity classification according to the IDF definition, with 1.2% of those surveyed classified as ‘non-Europid’. Data from the Australian Census (2006) show that of all Australians born in Asian countries, 81% speak a language other than English at home, suggesting that language spoken at home is a good proxy for ethnicity in this sample. Language spoken at home is likely to have high sensitivity for ethnicity. Those with missing values for urinary albumin were assumed to be negative for this variable (n = 572).

Those who could not be classified as either having or not having the MetS because of missing data were excluded from all analyses. The percentages unclassified for each MetS definition because of missing data were WHO (1.8%), ATPIII (0.3%), EGIR (1.6%) and IDF (0.4%). Despite fasting insulin only being measured in those aged over 35 years, 89% of those aged under 35 years in the sample could be classified according to the WHO and EGIR definitions (which both require insulin measurement) based on other components (i.e. where all other MetS abnormalities were absent, by definition MetS must be absent). The total number who could not be classified because of the absence of a fasting insulin measurement was 84. Insulin sensitivity (HOMA-S) was calculated using the homeostasis model [19].

The version of the FINnish Diabetes RIsk SCore (FINDRISC) [20] used differed from that published in that AusDiab lacks data on history of diabetes among relatives other than parents and occupational physical activity was excluded from total physical activity time in the validated Active Australia physical activity survey used in the AusDiab study because of the difficulties in measuring this component of physical activity [21, 22].

The DPM was developed in the San Antonio Heart Study to calculate the probability of development of diabetes over 7.5 years [23]. The model included age, sex, ethnicity, FPG, systolic blood pressure, HDL-C, BMI and family history of diabetes (in parents or siblings). In this analysis, the corrected equation used in a later publication has been used (P = 1/1 + e−x), where x = −13.415 + 0.028 (age in years) + 0.661 (1 if female, 0 if male) + 0.412 (if Mexican American, 0 if nonhispanic white) + 0.079 (fasting glucose in mg dL−1) + 0.018 (systolic blood pressure in mmHg) - 0.039 (HDL cholesterol in mg dL−1) + 0.070 (BMI in kg m−2) + 0.481 (1 if family history, 0 if no family history) [24]. In this equation, family history was defined as the presence of type 2 diabetes in a parent or sibling, however in AusDiab, only information on parental history of diabetes was collected. Participants (95.2%) were born in European countries or countries where a large proportion have Europid ancestry (e.g. USA, Canada, New Zealand), with no particular ethnicity highly represented in the remaining 4.8% [(n = 13 (0.2%) from Central and South American countries)]. For this reason, all participants were coded as ‘nonhispanic white’ for the ethnicity component of the equation. This model has also been validated for the prediction of diabetes in the Mexico City Diabetes Study [5].

Statistical methods

Statistical analysis was conducted using stata 9.0 (College Station, TX, USA). Receiver operating characteristic (ROC) curves for incident diabetes were calculated for the FINDRISC score, the DPM and fasting and 2 h glucose values onto which sensitivity and false-positive rate for the MetS were also plotted. Area under the curve was calculated and compared using the roccomp command in Stata. Sensitivity and specificity for the MetS definitions tested were compared first using overall chi-square tests, then if P < 0.05, using chi-square tests with a single degree of freedom. Logistic regression was used to estimate odds ratios (95% CIs) for the development of diabetes between baseline and follow-up for the MetS, the FINDRISC score and the DPM, adjusting for each of the components of the respective MetS definitions as continuous variables, with the exception of treatment for hypertension or dyslipidaemia (dichotomous variables). Likelihood ratio tests (reporting the chi-square statistic) were used to compare the impact of adding fasting plasma glucose and the MetS respectively, to logistic regression models for incident diabetes, adjusted for age and sex. The DPM provides probabilities for the development of diabetes over 7.5 years. Annualized incidence was calculated using the formula −ln(1−S)/t, where S is the proportion of new cases (number of new cases at follow-up/number of cases at risk at baseline) and t is the time of follow-up. Odds ratios (OR) for the DPM are presented for a 10% increase in this probability.

Results

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References

The baseline characteristics of the study population, which was shown to be overweight and an average of 52 years of age, are shown in Table 1.

Table 1.   Baseline characteristics: the AusDiab study
 MaleFemaleTotal
  1. a=geometric mean.

  2. Data are percentages, means or geometric means (95% confidence interval).

n264232005842
Age (years)51.3 (50.8–51.8)50.6 (50.2–51.1)50.9 (50.6–51.2)
BMI (kg m−2)27.0 (26.9–27.1)26.4 (26.2–26.6)26.7 (26.5–26.8)
Waist circumference (cm)96.8 (96.4–97.2)84.1 (83.7–84.5)89.8 (89.5–90.2)
Weight (kg)83.8 (83.3–84.3)69.8 (69.3–70.3)76.1 (75.7–76.5)
Systolic blood pressure (mmHg)131.5 (130.9–132.2)124.8 (124.2–125.4)127.8 (127.4–128.3)
Diastolic blood pressure (mmHg)74.6 (74.2–75.0)66.2 (65.9–66.6)70.0 (69.7–70.3)
Serum HDL cholesterol (mmol L−1)1.3 (1.3–1.3)1.6 (1.6–1.6)1.4 (1.4–1.5)
Serum Triglycerides (mmol L−1)*1.4 (1.4–1.4)1.1 (1.1–1.2)1.3 (1.2–1.3)
Fasting plasma glucose (mmol L−1)5.5 (5.5–5.6)5.2 (5.2–5.3)5.4 (5.4–5.4)
2h Plasma glucose (mmol L−1)5.8 (5.8–5.9)6.0 (5.9–6.0)5.9 (5.9–6.0)
Fasting serum insulin (mU mL−1)*12.7 (12.5–12.9)12.1 (11.9–12.3)12.3 (12.2–12.5)
HOMA-S (units)a59.8 (58.7–60.9)63.8 (62.7–64.8)61.9 (61.2–62.7)
WHO MetS23.9 (22.2–25.5)14.2 (13.0–15.5)18.6 (17.6–19.6)
ATPIII MetS29.9 (28.1–31.6)21.2 (19.8–22.6)25.1 (24.0–26.2)
EGIR MetS18.6 (17.1–20.1)12.7 (11.5–13.8)15.3 (14.4–16.2)
IDF MetS36.3 (34.4–38.1)25.0 (23.5–26.5)30.1 (28.9–31.3)
Current smoker12.9 (11.6–14.2)9.9 (8.9–11.0)11.3 (10.5–12.1)
Previous CVD (stroke, angina, heart attack)7.9 (6.8–8.9)4.2 (3.5–4.9)5.9 (5.3–6.5)
Cholesterol lowering medication7.8 (6.8–8.8)6.7 (5.9–7.6)7.2 (6.5–7.9)
Blood pressure lowering medication12.2 (11.0–13.5)9.9 (8.9–11.0)12.8 (12.0–13.7)
Higher education (University/Technical and further education (TAFE) college)46.9 (44.9–48.8)37.4 (35.7–39.0)41.7 (40.4–42.9)

Incidence of diabetes

Over the 5-year follow-up, 224 (3.8%) individuals developed diabetes resulting in an annual incidence of 7.8/1000. The incidence varied from 2.9/1000per year for those identified by none of the MetS definitions, to between 17.1 and 24.6/1000per year for the four MetS definitions tested. In comparison with those not identified as at risk by any MetS definition, the OR (95%CIs) for incident diabetes for different definitions of the MetS were: IDF 5.5 (3.9–7.6), ATPIII 6.4 (4.6–9.0), EGIR 7.4 (5.2–10.4) and WHO 7.8 (5.5–11.0).

The MetS as a tool for prediction of diabetes

The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for the prediction of incident diabetes by each of the four MetS definitions, a FINDRISC score above 10, and the DPM with probability of developing diabetes over 7.5 years of 31.7% (cut-points for FINDRISC and the DPM were chosen to have comparable sensitivity to the IDF MetS definition) are presented in Table 2. For a sensitivity similar to that of the IDF and ATPIII definitions of the MetS, the FINDRISC questionnaire had lower PPV and specificity than each of the four MetS definitions, with these differences being restricted to women (P = 0.02 and P < 0.0001 respectively), while the DPM had a higher PPV and specificity, with these differences greater in men (P < 0.02 and P < 0.0001 respectively) than women (P > 0.05 and P < 0.0001 respectively).

Table 2.   Sensitivity, specificity and positive predictive value (PPV) and negative predictive value (NPV) for incident diabetes for each of four definitions of the MetS, the diabetes predicting model (DPM) and the Finnish diabetes risk prediction tool (FINDRISC) ≥10a: The ausDiab study
 WHOATPIIIEGIRIDFFINDRISC≥10DPM≥ 31.7%
  1. aThe cut-points for the FINDRISC score and DPM score were chosen to have comparable sensitivity to the ATPIII MetS definition. bPrevalence is among those without diabetes at baseline.

Sensitivity
 Total56.462.246.464.962.362.4
 Men65.261.953.568.162.266.1
 Women47.262.438.761.562.458.4
Specificity
 Total82.976.485.971.370.582.3
 Men78.071.583.165.172.780.9
 Women86.980.388.276.368.683.6
PPV
 Total11.59.411.68.26.811.9
 Men11.88.912.78.18.513.2
 Women11.110.010.28.45.510.7
NPV
 Total98.098.197.698.198.298.3
 Men98.097.797.597.997.998.2
 Women97.998.497.798.298.498.4
Prevalenceb
 Total18.625.115.330.130.619.3
 Men23.929.918.636.328.721.1
 Women14.221.212.725.032.317.8

From Fig. 1, it is apparent that in this population, the MetS was not superior to measurement of glucose (fasting or 2-h post load) or the DPM in identifying those who developed diabetes. At the sensitivity of each of the MetS definitions, the specificity of the DPM was significantly greater than the specificities of ATPIII (P = 0.0003), IDF (P < 0.0001), and EGIR (P = 0.004), but not WHO. The specificity of fasting glucose was significantly greater than that of ATPIII (P = 0.0009), IDF (P < 0.0001), and EGIR (P = 0.0002), but not WHO. Only the specificities of ATPIII (P = 0.0009) and WHO (P < 0.0001) MetS definitions were significantly greater than the FINDRISC score. The area under the ROC curves were 0.727, 0.759 and 0.783 for the FINDRISC score, fasting glucose and the DPM respectively. The only statistically significant difference in the area under the curves was between FINDRISC and the DPM (P = 0.0019) (FINDRISC versus fasting glucose, P = 0.16; DPM versus fasting glucose P = 0.14). When examined separately within each sex (ROC curves not shown), the WHO definition of the MetS for men, and both the ATPIII and IDF definitions of the MetS for women, had a combination of sensitivity and specificity above the respective FPG ROC curves. However, at the specificity of each of these MetS definitions, the sensitivity for incident diabetes was no more than four percentage points greater than the sensitivity of the corresponding FPG value in men or women (all P > 0.05).

image

Figure 1.  Receiver operating characteristic curve for detection of incident diabetes over 5 years in the AusDiab population, using the FINnish Diabetes RIsk SCore (FINDRISC), diabetes risk questionnaire, different definitions of the Metabolic Syndrome (MetS), impaired fasting glucose (IFG), impaired glucose tolerance (IGT), the Diabetes Prediction Model (DPM), fasting plasma glucose and postload 2-h plasma glucose.

Download figure to PowerPoint

Furthermore, in logistic regression analysis adjusted for age and sex, fasting glucose had the greatest OR (per standard deviation) for incident diabetes [fasting glucose OR 3.05 (P < 0.0001); waist circumference OR 1.95 (P < 0.0001); triglycerides OR 1.45 (P < 0.0001); HDL cholesterol OR 0.65 (P < 0.0001); and systolic blood pressure OR 1.41 (P < 0.0001)]. The four MetS definitions remained significant predictors for diabetes in models containing a single MetS definition (1.7 <  OR < 2.4; all P < 0.001), age, sex and fasting glucose, however fasting glucose was more important than the MetS in each of these models (likelihood ratio test χ2 for fasting glucose was >158 and for the MetS <29 in each of the four models tested).

The risk for diabetes among those who were not classified as having the metabolic syndrome (as defined by ATPIII) was also assessed adjusted for age and sex, with fasting glucose being the only significant predictor in this lower risk group [fasting glucose OR (per standard deviation) 2.56 (P < 0.0001)]; waist circumference OR 1.24 (P = .15); triglycerides OR 1.23 (P = 0.105); HDL cholesterol 1.16 (P = 0.2); systolic blood pressure OR 1.07 (P = 0.6).

Finally, the absolute risk for future diabetes among those identified by the MetS (ATPIII definition) but who would not otherwise be identified as being at high risk for development of subsequent diabetes because of normal glucose tolerance (5 year risk = 3.4%) and the risk among those who were glucose intolerant (IFG or IGT), but not classified as having the MetS (5 year risk = 9.0%) was calculated.

Association between the MetS and diabetes after adjustment for MetS components

Table 3 shows that in men only, after adjustment for their components (and after further adjustment for age and parental history of diabetes) only the EGIR MetS definition maintained a statistically significant association with incident diabetes [OR 2.1 (1.2-3.8)]. The relationship was maintained after further adjustment for other potential diabetes risk factors such as education level, income, smoking status, physical activity time, television viewing time, history of previous cardiovascular disease, ethnicity and fruit and vegetable consumption [OR 2.0 (1.1–3.7)].

Table 3.   Five-year risk of diabetes according to four definitions of the Metabolic Syndrome (MetS), the FINnish Diabetes RIsk SCore (FINDRISC) questionnaire and the diabetes prediction model (DPM) with and without adjustment for MetS components
MetSModel 1aModel 2bModel 3c
MaleFemaleMaleFemaleMaleFemale
  1. aModel 1 – unadjusted.

  2. bModel 2 – ORs for MetS adjusted for all components (as continuous variables) of the respective MetS definition, and for blood pressure treatment. ORs for FINDRISC and DPM adjusted for all components of the IDF MetS definition and for blood pressure treatment.

  3. cModel 3 – includes all variables from model 2, plus age and parental history of diabetes.

WHO6.6 (4.4–9.9)5.9 (4.0–8.8)1.1 (0.6–2.0)1.0 (0.5–1.8)1.1 (0.6–2.1)0.9 (0.5–1.7)
ATPIII4.1 (2.8–6.0)6.8 (4.5–10.0)0.8 (0.5–1.4)1.5 (0.8–2.8)0.8 (0.4–1.4)1.4 (0.7–2.7)
EGIR5.6 (3.8–8.3)4.7 (3.1–7.1)2.1 (1.2–3.7)0.9 (0.5–1.8)2.1 (1.2–3.8)1.0 (0.6–2.0)
IDF4.0 (2.7–6.0)5.1 (3.5–7.6)0.9 (0.5–1.5)0.8 (0.5–1.5)0.8 (0.5–1.5)0.8 (0.4–1.5)
FINDRISC
 Low1.01.01.01.01.01.0
 Slightly elevated3.2 (1.8–6.0)3.6 (1.9–7.0)1.9 (0.9–3.8)1.6 (0.8–3.4)0.8 (4.2–0.0)1.2 (0.5–2.8)
 Moderate9.0 (4.7–17.2)4.9 (2.3–10.4)3.9 (1.7–8.8)1.7 (0.7–4.3)4.2 (1.5–11.6)0.9 (0.3–2.8)
 High10.9 (5.1–23.6)9.8 (4.5–21.4)3.9 (1.5–10.1)2.8 (1.0–7.6)4.4 (1.3–15.8)1.1 (0.3–4.4)
 Very highn/a59.4 (17.3–203.5)n/a12.5 (2.8–55.7)n/a4.0 (0.6–25.4)
 Questionnaire (OR per unit score)1.2 (1.2–1.3)1.2 (1.2–1.3)1.1 (1.0–1.2)1.1 (1.0–1.2)1.1 (1.0–1.2)1.1 (1.0–1.2)
Diabetes predicting model
 OR per 10% increase in probability1.7 (1.6–1.9)1.6 (1.5–1.7)1.5 (1.1–1.9)1.4 (1.1–1.8)1.3 (0.9–1.9)1.2 (0.9–1.6)

Conclusions

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References

Used in isolation as a clinical predictor of incident diabetes over 5 years, our results demonstrate that each of the four tested MetS definitions were not superior to measurement of blood glucose alone or a published diabetes risk prediction model. Those with prediabetes (IFG or IGT) but without the MetS had an age-standardized absolute risk for incident diabetes almost triple that of those with the MetS but without prediabetes, further emphasizing the utility of glucose testing alone in predicting diabetes.

In a previous comparison of the MetS and the DPM as diabetes predictors, the ATPIII definition of the MetS was inferior using ROC curve analysis [5]. That study was used to create the DPM, and the results of our study confirm the DPM’s utility and superiority in an independent Caucasian population, for both sexes.

On the basis of our overall findings, of the tools tested, a single FPG measurement could be argued to be the best and most practical predictor of incident diabetes. This does not suggest that the diabetic risk associated with the metabolic syndrome is irrelevant beyond the effect of fasting glucose, but rather that in their published forms, this risk is not captured. While these findings may suggest that the value of the MetS as a clinical tool for diabetes prediction needs to be questioned, prediction of incident diabetes was not the sole justification for creation of the latest MetS definitions. Valuable additional roles for the MetS include placing an emphasis on the interaction between, and importance of, multiple modifiable risk factors; as an explanation to patients for why they are developing multiple conditions simultaneously; and ensuring that tackling the components of the MetS (using the proven treatment of lifestyle modification) will prevent or delay onset of not only diabetes, but other associated conditions such as CVD, and improve general health and wellbeing. Whether the diagnosis of the MetS actually helps improve the uptake of lifestyle modifications is an unanswered question. Being the only risk prediction tool tested not to rely on measured biological parameters, the ROC curve for the FINDRISC score suggests that this is a potentially valuable screening tool in this population.

Laaksonen et al. concluded that in middle-aged Finnish men the WHO definition was superior to the ATPIII definition for prediction of diabetes, and our results show a similar trend, although this did not reach statistical significance [1]. When men and women are combined, however, our results are consistent with American data suggesting that the ATPIII definition is at least as good as that of the WHO, with this result largely because of the considerable advantage of ATPIII among women (difference in sensitivity, P < 0.0001) [3, 4]. This difference emphasizes the importance of separate analyses conducted among men and women and different ethnic groups. The often ignored sex-specific differences in the prevalence of both IFG (usually more common among men) and IGT (usually more common among women) as well as in waist circumference, mean that sex-specific differences in the prevalence and characteristics of those identified by the MetS, and hence its ability to predict incident diabetes will be ubiquitous [25].

From our results, we have demonstrated that the ability of current MetS definitions to predict future diabetes can largely be attributed to the individual risk of each of its components. An individual exception to this was the MetS as defined by EGIR in men, which had an independent statistical association with incident diabetes after adjustment for its components (and other risk factors), which could suggest either a significant interaction between the components and/or the influence of factors not included in the definition including the possibility of an underlying abnormality not directly measured. The independence from its components of the EGIR MetS definition may suggest that the glucose/insulin component may be a more important (or at least better measured) driver of diabetes than the comparatively crudely measured obesity component, at least over the 5-year follow-up used in this study.

Other potential methods for demonstrating that the MetS predicts diabetes independently of its components include factor analysis and assessment of statistical interactions. Factor analysis can identify clustering of MetS components, but not whether the identified clustering is associated with clinical outcomes independently of the individual variables utilized, and statistical interactions for a dichotomous outcome are functionally limited to two-way interactions [26]. While the majority of the MetS definitions tested proved not to be independent from their components in this analysis, it is true that both a lack of an independent association does not rule out an underlying phenotype (particularly as the current definitions are somewhat arbitrary and based entirely on dichotomous cut-points) and that a statistically independent association may be explained not by an interaction between the components but by nonlinearity of the continuous variables modeled (e.g. threshold effects or skewed distributions), sub-optimally measured component variables or a reduction in measurement error from the pooling of multiple risk factors [27].

The response to the baseline and follow-up AusDiab studies of 55% and 58%, respectively, means that the sample may not be fully representative of adult Australians, despite being population-based. Further caveats are that the ability to identify incident cases of diabetes may be different with follow-up longer than 5 years and in other ethnic and cultural groups; and that the cut-points for the top quartile of fasting insulin and bottom quartile of HOMA-S were derived from the AusDiab population aged over 35 years, and therefore may be different than if they had been calculated using the full sample aged 25 years or more.

While the MetS is an effective predictor of incident diabetes, these results suggest that in this population, the MetS was no better at predicting diabetes than a single blood glucose measurement. In addition, it appears not to be independent of its components for the commonly used clinical MetS definitions, although in men, the EGIR definition remained significant after adjustment for its component variables. Given these results, debate surrounding the public health utility of a MetS diagnosis, at least for identification of incident diabetes, is required. Meanwhile, those classified as having the MetS should be strongly encouraged to develop strategies for lifestyle modification with its proven benefits of diabetes risk reduction and related positive outcomes [28, 29].

Acknowledgements

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References

We are enormously grateful to A Allman, B Atkins, S Bennett, A Bonney, S Chadban, M de Courten, M Dalton, D Dunstan, M D’ Embden, T Dwyer, H Jahangir, D Jolley, I Kemp, P Magnus, J Mathews, D McCarty, A Meehan, N Meinig, S Murray, K O’Dea, P Phillips, P Popplewell, C Reid, A Stewart, R Tapp, H Taylor, T Whalen and F Wilson for their invaluable contribution to the set-up and field activities of AusDiab. A Cameron is supported by Postgraduate Research Scholarship PP04M1794 from the National Heart Foundation of Australia. The work was supported by a grant from the National Health and Medical Research Council of Australia (grant number 233200). We are most grateful to the following for their support of the study: The Commonwealth Department of Health and Aged Care, Abbott Australasia Pty Ltd, Alphapharm Pty Ltd, Aventis Pharmaceutical, AstraZeneca, Aventis Pharmaceutical, Bristol-Myers Squibb Pharmaceuticals, Eli Lilly (Aust) Pty Ltd, GlaxoSmithKline, Janssen-Cilag (Aust) Pty Ltd, Merck Lipha s.a., Merck Sharp & Dohme (Aust), Novartis Pharmaceutical (Aust) Pty Ltd, Novo Nordisk Pharmaceutical Pty Ltd, Pharmacia and Upjohn Pty Ltd, Pfizer Pty Ltd, Roche Diagnostics, Sanofi Synthelabo (Aust) Pty Ltd, Servier Laboratories (Aust) Pty Ltd, BioRad Laboratories Pty Ltd, HITECH Pathology Pty Ltd, the Australian Kidney Foundation, Diabetes Australia, Diabetes Australia (Northern Territory), Queensland Health, South Australian Department of Human Services, Tasmanian Department of Health and Human Services, Territory Health Services and Victorian Department of Human Services and Health Department of Western Australia.

References

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Research design and methods
  5. Results
  6. Conclusions
  7. Conflict of interest
  8. Acknowledgements
  9. References
  • 1
    Laaksonen DE, Lakka HM, Niskanen LK, Kaplan GA, Salonen JT, Lakka TA. Metabolic syndrome and development of diabetes mellitus: application and validation of recently suggested definitions of the metabolic syndrome in a prospective cohort study. Am J Epidemiol 2002; 156: 10707.
  • 2
    Sattar N, Gaw A, Scherbakova O et al. Metabolic syndrome with and without C-reactive protein as a predictor of coronary heart disease and diabetes in the West of Scotland Coronary Prevention Study. Circulation 2003; 108: 4149.
  • 3
    Hanley AJ, Karter AJ, Williams K et al. Prediction of type 2 diabetes mellitus with alternative definitions of the metabolic syndrome: the insulin resistance atherosclerosis study. Circulation 2005; 112: 371321.
  • 4
    Lorenzo C, Okoloise M, Williams K, Stern MP, Haffner SM. The metabolic syndrome as predictor of type 2 diabetes: the San Antonio heart study. Diabetes Care 2003; 26: 31539.
  • 5
    Stern MP, Williams K, Gonzalez-Villalpando C, Hunt KJ, Haffner SM. Does the metabolic syndrome improve identification of individuals at risk of type 2 diabetes and/or cardiovascular disease? Diabetes Care 2004; 27: 267681.
  • 6
    Wilson PW, D’Agostino RB, Parise H, Sullivan L, Meigs JB. Metabolic syndrome as a precursor of cardiovascular disease and type 2 diabetes mellitus. Circulation 2005; 112: 306672.
  • 7
    Wannamethee SG, Shaper AG, Lennon L, Morris RW. Metabolic syndrome vs Framingham Risk Score for prediction of coronary heart disease, stroke, and type 2 diabetes mellitus. Arch Intern Med 2005; 165: 264450.
  • 8
    Schmidt MI, Duncan BB, Bang H et al. Identifying individuals at high risk for diabetes: the atherosclerosis risk in communities study. Diabetes Care 2005; 28: 20138.
  • 9
    Lorenzo C, Williams K, Hunt KJ, Haffner SM. The National Cholesterol Education Program-Adult Treatment Panel III, International Diabetes Federation, and World Health Organization Definitions of the Metabolic Syndrome as Predictors of Incident Cardiovascular Disease and Diabetes. Diabetes Care 2007; 30: 813.
  • 10
    Dunstan DW, Zimmet PZ, Welborn TA et al. The Australian Diabetes, Obesity and Lifestyle Study (AusDiab) – methods and response rates. Diabetes Res Clin Pract 2002; 57: 11929.
  • 11
    Barr L, Magliano D, Zimmet P, et al. AusDiab 2005. The Australian Diabetes, Obesity and Lifestyle Study. Tracking the Accelerating Epidemic: Its Causes and Outcomes. Melbourne: International Diabetes Institute, 2006.
  • 12
    World Health Organisation. Defintion, Diagnosis and Classification of Diabetes Mellitus and its Complications. Part 1: Diagnosis and Classification of Diabetes Mellitus. Geneva: Department of Non-communicable Diseases Surveillance, World Health Organization, 1999.
  • 13
    Cameron AJ, Welborn TA, Zimmet PZ et al. Overweight and obesity in Australia: the 1999-2000 Australian Diabetes, Obesity and Lifestyle Study (AusDiab). Med J Aust 2003; 178: 42732.
  • 14
    Briganti EM, Shaw JE, Chadban SJ et al. Untreated hypertension among Australian adults: the 1999-2000 Australian Diabetes, Obesity and Lifestyle Study (AusDiab). Med J Aust 2003; 179: 1359.
  • 15
    World Health Organization. Definition, Diagnosis and Classification of Diabetes Mellitus and its Complications; Part 1: Diagnosis and Classification of Diabetes Mellitus. Geneva: Department of Noncommunicable Disease Surveillance, 1999.
  • 16
    Alberti KG, Zimmet P, Shaw J. Metabolic syndrome – a new world-wide definition. A Consensus Statement from the International Diabetes Federation. Diabet Med 2006; 23: 46980.
  • 17
    Balkau B, Charles MA, Drivsholm T et al. Frequency of the WHO metabolic syndrome in European cohorts, and an alternative definition of an insulin resistance syndrome. Diabetes Metab 2002; 28: 36476.
  • 18
    NCEP/ATP I. Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA 2001; 285: 248697.
  • 19
    Levy J, Matthews D, Hermans M. Correct Homeostatic Model Assessment (HOMA) Evaluation Uses the Computer Program. Diabetes Care 1998; 21: 21912.
  • 20
    Lindstrom J, Tuomilehto J. The diabetes risk score: a practical tool to predict type 2 diabetes risk. Diabetes Care 2003; 26: 72531.
  • 21
    Australian Institute of Health and Welfare (AIHW). The Active Australia Survey. A guide and manual for implementation, analysis and reporting. http://www.aihw.gov.au/publications/cvd/aas/aas.pdf , 2003.
  • 22
    Brown WJ, Trost SG, Bauman A, Mummery K, Owen N. Test-retest reliability of four physical activity measures used in population surveys. Journal of science and medicine in sport / Sports Medicine Australia 2004; 7: 20515.
  • 23
    Stern MP, Williams K, Haffner SM. Identification of persons at high risk for type 2 diabetes mellitus: do we need the oral glucose tolerance test? Ann Intern Med 2002; 136: 57581.
  • 24
    McNeely MJ, Boyko EJ, Leonetti DL, Kahn SE, Fujimoto WY. Comparison of a clinical model, the oral glucose tolerance test, and fasting glucose for prediction of type 2 diabetes risk in Japanese Americans. Diabetes Care 2003; 26: 75863.
  • 25
    Williams JW, Zimmet PZ, Shaw JE et al. Gender differences in the prevalence of impaired fasting glycaemia and impaired glucose tolerance in Mauritius. Does sex matter?. Diabet Med 2003; 20: 91520.
  • 26
    Norton EC, Wang H, Ai C. Computing interaction effects and standard errors in logit and probit models. Stata J 2004; 4: 15467.
  • 27
    Brotman DJ, Walker E, Lauer MS, O’Brien RG. In search of fewer independent risk factors. Arch Intern Med 2005; 165: 13845.
  • 28
    Tuomilehto J, Lindstrom J, Eriksson J et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med 2001; 344: 134350.
  • 29
    Knowler WC, Barrett-Connor E, Fowler SE et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002; 346: 393403.