Methods

NOMAS is a population-based cohort study designed to determine the incidence of and risk factors for stroke and other vascular outcomes in a multiethnic urban community. Details of the NOMAS design have been described previously. Briefly, eligible participants (1) had never been diagnosed with ischemic stroke; (b) were >40 years old; and (3) resided in Northern Manhattan for ≥3 months, in a household with a telephone. Participants were identified by random-digit dialing, and interviews were conducted by trained bilingual research assistants. The telephone response rate was 91%. Participants were recruited from the telephone sample to have an in-person baseline interview and assessment. The enrollment response rate was 75%, the overall participation rate was 69%, and a total of 3,298 participants were enrolled. Participants with an MI before baseline (n = 237) were excluded. Food frequency data sufficient to calculate a DT score were available for 2,568 participants, and of these participants, measurement of LV mass was available for 1,937. This study was approved by the Institutional Review Boards of Columbia University Medical Center and the University of Miami, and participants provided written informed consent.

Data were collected through interviews with trained research assistants in English or Spanish. Physical and neurologic examinations were conducted by study neurologists. Race or ethnicity was based on self-identification through a series of questions modeled after the US census and conforming to standard definitions outlined by Directive 15. Standardized questions were adapted from the Behavioral Risk Factor Surveillance System by the Centers for Disease Control regarding hypertension, diabetes, smoking, and cardiac conditions. Blood pressure was measured with mercury sphygmomanometers and appropriately sized cuffs. Hypertension was defined by a blood pressure (BP) ≥140/90 mm Hg (on the basis of the average of 2 measurements during 1 sitting), the patient’s self-reported hypertension, or use of antihypertensive medications. Diabetes mellitus was defined by the patient’s self-reported diabetes, use of insulin or oral antidiabetic medication, or fasting glucose ≥126 mg/dl. Body mass index (BMI) was calculated in kg/m ² . Smoking was categorized as never smoking, former smoking, and current (within the past year) smoking. Physical activity was defined as the frequency and duration of 14 recreational activities during the 2-week period before the interview, as described previously.

At baseline, participants were administered a comprehensive in-person diet assessment using a modified Block National Cancer Institute food frequency questionnaire (FFQ) by trained research assistants in English or Spanish depending on participants’ primary language. This FFQ listed 207 foods and was intended to represent typical food consumption over the previous year. The available food responses were modified to include specific dietary items frequently consumed by Hispanic populations.

Construction of the DT score has been described previously. Briefly, we regressed energy intake (kilocalories) and calculated the derived residuals of daily gram intake for the following food categories: dairy, meat, fruits and nuts, vegetables (excluding potatoes), legumes, grains and cereals, and fish. Participants scored 1 point for each beneficial component (fruits and nuts, vegetables, legumes, grains and cereals, and fish) whose consumption was at or above the gender-specific median; 1 point for each component associated with increased risk of CVD (meat and dairy products) whose consumption was below the median; and 1 point for a ratio of monounsaturated fats to saturated fats above the median and for mild-to-moderate alcohol consumption (defined as >0 drinks/week but ≤2 drinks/day over the previous year). The DT score was the sum of the scores in the food categories (range 0 to 9) with a greater score indicating stricter adherence. The DT score, the primary exposure of interest, was analyzed categorically (the top quartile score 6 to 9 vs the bottom 3 quartiles score 0 to 5) and as a continuous variable (per 1-point increment).

Transthoracic echocardiography was performed, and measurements were taken by standard 2-dimensional protocols according to the guidelines of the American Society of Echocardiography. Left ventricular end-diastolic diameter (LVDD), left ventricular end-systolic diameter, interventricular septum (IVS), and posterior wall thickness (PWT) at end-diastole were measured. LV mass was calculated using the corrected American Society of Echocardiography method: <SPAN role=presentation tabIndex=0 id=MathJax-Element-1-Frame class=MathJax style="POSITION: relative" data-mathml='0.8×(1.04×[{IVS+LVDD+PWT}3−LVDD3]+0.6)’>0.8×(1.04×[{IVS+LVDD+PWT}3−LVDD3]+0.6)0.8×(1.04×[{IVS+LVDD+PWT}3−LVDD3]+0.6)
0.8 × ( 1.04 × [ { IVS + LVDD + PWT } 3 − LVDD 3 ] + 0.6 )
. For this analysis, LV mass was examined in 2 ways, consistent with previous reports: (1) as a raw uncorrected value and (2) indexed to body size by dividing raw LV mass by height to the allometric power of 2.7 (LV mass). Both LV mass indexes were examined as continuous variables.

The cross-sectional associations between the DT score (assessed continuously and as a dichotomous variable) and the LV mass indexes were examined using multivariable-adjusted linear regression models after confirming the normal distributions of the LV mass variables. We used a sequence of 4 models: (1) unadjusted; (2) adjusted for age, gender, race or ethnicity, and education level; (3) adjusted for the variables in model 2 and physical activity, average total daily kilocalorie consumption, and smoking (never, past, or current); and (4) adjusting for potential mediators and confounders, including the covariates in model 3 and diabetes, systolic BP, diastolic BP, and use of antihypertensive medications. We also conducted an additional analysis that included BMI in addition to the variables in model 4. In a supplementary analysis, we simultaneously entered the DT score components into model 3 as predictors of the LV mass variables. Because self-reported kilocalories <500 or >4,000 might indicate inaccurate reporting of dietary information, we conducted sensitivity analyses excluding these participants. Finally, we examined the potential for effect modification by age, gender, race or ethnicity, and hypertension by including interaction terms with the continuous DT score in model 4 for each of the LV mass–dependent variables of interest.