Methods

We identified all adults aged ≥30 years with T2DM using the cross-sectional National Health and Nutrition Examination Survey (NHANES) from 2007 to 2012. In our study, T2DM was defined as those who were aged 30 years and older when first told they had diabetes, as has been used in previous NHANES studies as a cut point for the diagnosis of T2DM. Diabetes duration was calculated by subtracting the participant’s age at survey from their age when first told they had diabetes. The sample population was then restricted to persons who had information on age, gender, blood pressure, HbA1c, high-density lipoprotein cholesterol (HDL-C), and triglycerides and were free of any known clinical CVD defined as answering no to the survey questions asking whether they were told by their doctor if they ever had CHD, heart attack, stroke, or heart failure.

The UKPDS Risk Engine includes age, sex, HbA1c, diabetes duration, smoking status, total/HDL-C ratio, systolic blood pressure, and ethnicity and calculates the 10-year probability of developing a CHD event. Diabetes duration is defined as the number of years because a person was told they first had diabetes. Smoking status is defined by whether a person never smoked, previously smoked but has since stopped (past smoker), or continues to smoke (current smoker). In terms of other risk factors, HbA1c, systolic blood pressure, total cholesterol, and HDL-C are used to calculate risk. HbA1c was measured using high-performance liquid chromatography. Systolic blood pressure measurements were taken using a mercury sphygmomanometer and then averaged up to 4 measurements. Total cholesterol was measured enzymatically after hydrolyzation and oxidation. HDL-C was measured using a direct immunoassay technique. Detailed explanations to specimen analysis and data collection are discussed in the Laboratory Procedures Manual. The SI equivalent of goal levels was calculated as follows: cholesterol (mg/dl) × 0.0259 = cholesterol (mmol/L); (HbA1c (%) − 2.15) × 10.929 = HbA1c (mmol/mol). These values were then rounded to the nearest tenth. Goal or recommended levels were defined based on adapting the American Diabetes Association guidelines for diabetes care relevant for our survey period: HbA1c <7% (53 mmol/mol), systolic blood pressure <130 mm Hg, and HDL-C ≥1.036 mmol/L (40 mg/dl) in men and ≥1.3 (1.295) mmol/L (50 mg/dl) in women. A total cholesterol level of <4.403 mmol/L (170 mg/dl) was based on the approximate equivalent of the LDL-C goal of 2.59 mmol/L (100 mg/dl). The recommended HDL-C cut points are also consistent for persons with metabolic syndrome. A subanalysis also examined the impact of a blood pressure goal of <140/80 mm Hg that has been recently recommended by European and US societies. NHANES identifies 3 specific ethnicities, non-Hispanic white, non-Hispanic black, and Mexican American, and an “other” category, whereas UKPDS identifies 3 different ethnicities, white, Afro-Caribbean, and Asian-Indian. In our study, those identified as non-Hispanic black within NHANES were classified as Afro-Caribbean in the UKPDS Risk Engine and non-Hispanic white, Mexican American, and other were classified as white.

To examine the significance of varying levels of risk factor control, 3 risk factor control scenarios were established: (1) “All to Goal” control was defined as controlling all current smokers to past smokers and setting risk factor levels to minimum goal levels (as previously mentioned) in all subjects who were not already at control; (2) “Nominal Control” was defined as controlling all current smokers to past smokers, an absolute level reduction of 1% HbA1c (7.6 mmol/mol), 10% relative reduction in systolic blood pressure, 10% relative increase in HDL-C, and a 25% relative reduction of total cholesterol in those who had a respective risk factor above the goal level; and (3) “Aggressive Control” was defined as controlling all current smokers to past smokers, an absolute level reduction of 2% HbA1c (15.2 mmol/mol), a 20% relative reduction of systolic blood pressure, a 20% relative increase in HDL-C, and a 50% relative reduction of total cholesterol, for those who did meet the goal level of their respective risk factor. Just as in “All to Goal” control, only subjects not already at goal were “treated” in the “Nominal” and “Aggressive” control scenarios. Additionally, HbA1c levels could go below the target of 7% (53 mmol/mol) but by programing lower limits were not allowed to go below 6.5% (48 mmol/mol) in any subject in these 2 scenarios.

The sample population was compiled using SAS 9.1 and exported to the UKPDS spreadsheet allowing for estimation of 10-year myocardial infarction risk. An uncontrolled 10-year probability was calculated for each individual using the initial risk profile. That probability was then multiplied by the NHANES 6-year survey weighted sample size to calculate the number of events in 10 years. Estimates were stratified by gender and by 3 age groups beginning at 30 years of age: <45, 45 to 65, and ≥65 years. The weighted sample size was calculated by multiplying diagnosed diabetes prevalence by its appropriate population total from the US Current Population Surveys. Individual risk factors were examined for attainment of goal levels. If an individual was at goal for a specific risk factor, their level was left untouched. If an individual was not at goal, a new value was calculated based off their initial level and the control scenario described previously. Both the relative risk reduction and the absolute risk reduction (ARR) were calculated as the proportion of expected events over 10 years that would be prevented by the risk factor control scenario and the difference between the original and control absolute risk, respectively. After applying the new control levels, a new 10-year risk and number of events were calculated. The number of events was subtracted from the number of events at baseline to calculate the number of projected preventable events. The percent of projected preventable events was calculated by dividing the number of controlled events by the number of uncontrolled events and then subtracting that number from one. For composite analysis, all risk factors were examined and calculated by the level of control being examined. Finally, given the potential for misclassification of rate, we conducted sensitivity analyses for race by recalculating estimates setting all participants to Afro-Caribbean.