Methods

NHANES is a periodic survey of a representative sample of the civilian noninstitutionalized United States population. The purpose is to determine estimates of disease prevalence and the overall health status of the United States population. All participants gave written informed consent at the time of study enrollment. Participant characteristics, electrocardiography methodology, and ascertainment of mortality in NHANES III have been previously published. Briefly, NHANES III baseline data were collected during an in-home interview and a subsequent visit to a mobile examination center between 1988 and 1994. Data collected during the in-home interview included demographic and medication information. Blood samples were obtained at mobile centers, and basic laboratory values were recorded for each participant (total cholesterol, high-density lipoprotein cholesterol, and plasma glucose). The present analysis included 6,298 participants who were free of baseline CVD and who had available baseline demographic, laboratory, medication, and mortality data. Exclusion of participants with CVD was determined by a self-reported history of heart attack and/or stroke, electrocardiographic evidence of myocardial infarction, or major ST-T depression by Minnesota Electrocardiogram Classification.

Participant characteristics recorded during NHANES III were used in this analysis. Age, sex, race/ethnicity, and smoking history were self-reported. Medication history, including the use of antihypertensive and lipid-lowering medications, also was self-reported. Smoking was defined as current or former smoker. Blood pressure measurements were obtained, and values used were average readings from 3 in-home measurements and 3 mobile center measurements. Body mass index was computed as the weight in kilograms divided by the square of the height in meters. Diabetes was defined as a fasting plasma glucose level ≥126 mg/dl, glycosylated hemoglobin A1c values ≥6.5, or a history of diabetic medication use.

Standard 12-lead electrocardiograms were recorded using a Marquette MAC 12 system (Marquette Medical Systems, Milwaukee, Wisconsin) by trained technicians during each participants’ visit to a mobile examination center. Computerized automated analysis of the electrocardiographic data was performed with visual inspection of outlier values by a trained technician in a central electrocardiography core laboratory. The calculation of CIIS and methodology have been previously described. Briefly, the score is defined by a set of 11 discrete features in combination with 4 features measured in continuum and provides a simple scoring scheme suitable for both visual and computer classification of the conventional 12-lead electrocardiogram. By design, CIIS values were multiplied by a factor of 10 in NHANES III to avoid using decimal points. We divided the reported CIIS values by 10. Subclinical myocardial injury was defined as CIIS values ≥10, representing the limit for abnormal CIIS.

Mortality data for NHANES III participants were available through December 31, 2006, and methods for mortality ascertainment have been described. A probabilistic matching algorithm based on 12 identifiers was used to link participants with death information captured in the National Death Index. Matching identifiers included social security number, sex, and date of birth. Follow-up was defined as the interval between the NHANES III examination and either of the following, depending on which came first: date of death, date of censoring, or December 31, 2006. The end points of CVD and all-cause mortalities were examined and analyzed using data from the NHANES III Linked Mortality File. International Classification of Diseases, Tenth Revision , codes were used to identify each end point. CVD mortality was defined by codes 100 to 178. When censoring at the time of death did not include CVD, participants were grouped under all-cause mortality. Participants who were unable to be matched with a death record were considered to be alive throughout the entire follow-up period.

Continuous variables were reported as mean ± standard error, whereas categorical variables were reported as frequency and percentage. Statistical significance for continuous variables was tested using the t test procedure and the Rao-Scott chi-square method for categorical variables. Unadjusted CVD and all-cause mortality rates (per 1,000 person-years) were calculated. Kaplan-Meier estimates were used to compute unadjusted survival estimates for CVD and all-cause mortalities, and the differences in estimates were compared using the log-rank procedure. Cox proportional hazards regression was used to generate hazard ratios (HR) and 95% confidence intervals (95% CI) for the association between subclinical myocardial injury and CVD and all-cause mortalities. Additionally, CIIS was analyzed as a continuous variable per 5-unit increase. Multivariate models were constructed with incremental adjustments as follows: model 1 adjusted for age, sex, and race/ethnicity; model 2 adjusted for model 1 covariates plus smoking status, systolic blood pressure, diabetes, body mass index, total cholesterol, high-density lipoprotein cholesterol, antihypertensive medications, and lipid-lowering medications. We tested for interactions between our main effect variable and age (dichotomized at 65 years), sex, and race/ethnicity (white vs nonwhite). We also constructed a restricted cubic spline model to examine the graphical relation between each end point and CIIS and incorporated knots at the fifth, fiftieth, and ninety-fifth percentiles. The proportional hazards assumption was not violated in our analysis. Statistical significance was defined as p ≤0.05. Data were analyzed using SAS, version 9.3 (SAS Institute Inc., Cary, North Carolina). All analyses accounted for the complex sampling design of NHANES by including recommended sample weights.