Predicting Diabetes

Summarized by Robert W. Griffith, MD
June 28, 2007

Summary

A simple model of clinical factors is quite adequate for predicting type 2 diabetes. It includes the fasting blood sugar level, some lipid levels, BMI or waist measurement, and blood pressure.

Introduction

Type 2 diabetes is becoming increasingly common in US adults - it increased from 7.3% of the population in 1993 to 7.9% in 2000, and it certainly hasn't decreased since then. There have been attempts in the past to develop a predictive model for the disease to help guide interventions (e.g. lifestyle changes) and formulate health policies. The risk factors for diabetes, though well-known, are numerous, and cannot readily be included in algorithms. A group of eminent scientists have developed a relatively simple predictive tool, using subjects from the Framingham Offspring Study. It's been published in the Archives of Internal Medicine, and is summarized here.

What was done

The subjects studied were 3140 men and women who were examined as part of the Framingham Offspring Study in the mid-1990s. (People with diabetes - comprising 7.9% - had been excluded earlier.) The examination included a fasting blood sugar level and a 2-hour oral glucose tolerance test. Blood lipid levels, height, waist and weight measurements, and blood pressure levels were used to diagnose subjects with the metabolic syndrome.¹ Sophisticated insulin tests were included: the insulin resistance index, the beta-cell index, and the Gutt insulin sensitivity index.

The average age of the subjects was 54, and 54% of them were women. They were followed for an average of 7 years, when they were examined again. They were considered to have developed new diabetes during the follow-up period if they: (1) started on oral anti-diabetic or insulin medication; or (2) had a fasting glucose level of 126 mg/dL or higher at either of the two follow-up exams.

Different models of the risk of developing diabetes were developed using the available data, starting with the simplest parameters, and proceeding to the more complex; the driving principal was 'parsimony'.

The simple clinical model

The subjects were allocated to categories according to different ages, gender, parental history of diabetes, and body mass index (BMI). Age categories were under 50, 50 to 65, and over 65 years. BMI categories were below 25, 25 to 30, and over 30. The analyses showed that higher categories of age and BMI and a parental history of diabetes were significantly related to the development of diabetes in the next 7 years or so.

Enhanced clinical models

Additional categories were added to the original simple clinical model:

• blood pressure - over 130/85 mm Hg or on antihypertensive meds
• HDL cholesterol - below 40mg/dL in men, below 50 mg/dL in women
• Triglyceride level - 150 mg/dL or above
• Waist circumference - 102 cm or above in men, 88 cm or above in women
• Fasting glucose level - between 100 and126 mg/dL

Then 3 models were compared, using different measures of obesity: BMI only, waist circumference only, and BMI plus waist circumference. The results showed there was a statistical association between the occurrence of diabetes and parental history of the disease, raised blood pressure, a low HDL-C level, a raised triglyceride level, and an 'impaired' fasting glucose level (i.e. between 100 and 126 mg/dL). In the comparison between the 3 models, both the BMI and the waist circumference alone predicted diabetes development; including both the BMI and girth together did not improve the level of prediction.

The complex clinical models

Each of these included the parameters in the simple clinical model plus additional factors:

• 2-hour glucose level - between 140 and 200 mg/dL
• fasting insulin level
• C-reactive protein (CRP) level
• Three insulin sensitivity indices - Gutt sensitivity, HOMA insulin resistance, beta-cell index

Three models were created: all three included the 'simple' and 'enhanced' variables, plus some of the above factors. One included the 2-hour glucose, fasting insulin, and CRP, another the Gutt insulin sensitivity index, and the third the HOMA insulin resistance and the beta-cell index.

These models did not show any further improvement in identifying those who were to go on to develop diabetes, i.e. they were no better than the simple clinical model.

A prediction tool

The authors of the study developed a point system for the factors that played the greatest role in establishing the likelihood of developing type-2 diabetes within the subsequent 8 years; this is shown in the two tables:

Item	Points
Fasting blood glucose 100-126 mg/dL	10
BMI 25.0 - 29.9	2
BMI 30 or above	5
HDL-cholesterol over 40 mg/dL (men) or 50 mg/dL (women)	5
Parental history of diabetes	3
Triglyceride level 150 mg/dL or above	3
Blood pressure 130/85 mm Hg or above	2
TOTAL

Point Total	8-Year Risk of Type 2 Diabetes
10 or less	3% or less
11, 12	4%
13	5%
14	6%
15	7%
16	9%
17	11%
18	13%
19	15%
20	18%
21	21%
22	25%
23	29%
24	33%
25 or more	Over 35%

Conclusions

In evaluating the usefulness of the prediction tool that resulted from this study, it must be remembered that the population used in its creation was 99% white and non-Hispanic. This is important, as two studies have shown that people of Hispanic descent are more likely than other Americans to develop type 2 diabetes.

The metabolic syndrome factors are all represented in the final list used in the prediction tool - although waist circumference is used instead of BMI for overweight/obesity. However, as shown in the enhanced models, the use of either measure gives about the same result.

It's not likely that physicians will determine the exact %age risk for their individual patients. The important message from this study is the need for patients to actively address the factors in the above listings, initially by appropriate lifestyle changes (exercise, nutrition), and later, if necessary, using medication.

Source

• Prediction of incident diabetes mellitus in middle-aged adults. PWF. Wilson, JB. Meigs, L. Sullivan,  et al., Arch Intern Med, 2006, vol. 167, pp. 1068--1074

Footnotes
1. The metabolic syndrome requires 3 of the following risk factors to be present: (a) Waist size over 40 inches (102 cm) in men, or 35 inches (88 cm) in women. (b) Serum triglyceride level over 150 mg/dL (1.7 mmol/L). (c) Serum HDL ('good') cholesterol below 40 mg/dL (1.0 mmol/L) in men, or 50 mg/dL (1.29 mmol/L) in women. (d) Blood pressure over 130/85 mm Hg (either number), or being on blood pressure medication. (e) Fasting blood sugar over 110 mg/dL (6.1 mmol/L). Take the self-test, at the link below.

Related Links
Do You Have the Metabolic Syndrome?
When Should Diabetics Start a Heart Protection Program?
Blood Sugar Predicts Cardiac Mortality

Please take a moment to give us your comments. For questions about Health matters you may check our "Questions & Answers" Portal and Service.