MESA was initiated in July 2000 to investigate subclinical cardiovascular disease in a population-based sample of 6,814 men and women aged 45–84 years (mean age, 63 years) (46
). The cohort was selected from 6 study sites: Baltimore City and County, Maryland; Chicago, Illinois; Forsyth County, North Carolina; New York, New York; Los Angeles County, California; and St. Paul, Minnesota. Sampling and recruitment have been previously described (46
). Among those screened and deemed eligible, the participation rate was 59.8%. Analyses were restricted to the 6,191 MESA subjects who agreed to participate in the MESA Neighborhood Study, an ancillary study that collected information on neighborhood characteristics, including reported opportunities to purchase fast food. A total of 558 persons were excluded because of missing data, leaving 5,633 participants for analyses. Analyses using geographic information systems measures were restricted to participants in Maryland, North Carolina, and New York, where locational data on fast-food outlets were collected (n
= 2,447). The study was approved by institutional review boards at each site, and all participants gave written, informed consent. Analyses presented in this paper are based on self-administered survey data obtained from the baseline visit (July 2000–September 2002).
Dietary intake was assessed by a self-administered, modified-Block-style, 120-item food frequency questionnaire adapted from the Insulin Resistance Atherosclerosis Study instrument (45
). Dietary quality was characterized by 2 global dietary measures derived from the food frequency questionnaire, the AHEI (43
), and an empirically derived dietary pattern (45
AHEI is a summary index of dietary indicators that have been associated with a lower chronic disease risk (43
). AHEI was derived following prior work (43
), except where noted. Higher scores indicate higher intake of fruits and vegetables, nuts and soy protein, white versus red meat, cereal fiber, and polyunsaturated versus saturated fat. Higher scores also reflect moderate alcohol consumption, multivitamin use, and lower intake of trans
fatty acids. Previous studies used fiber from all grain sources and long-term (5-year) multivitamin use (43
), data that were not available in MESA. Cereal fiber and vitamin use at least once per month were substituted for these items in this study. Participants whose AHEI scores ranked in the top quintile of the distribution for the sample (range, 53–81) were classified as having a high-quality diet or a healthy diet. In other cohorts, scoring in the top quintile of the population distribution (i.e., AHEI scores of 47–86) versus in the bottom quintile was associated with a 28%–39% reduced risk of cardiovascular disease (43
A Western-type dietary pattern, the “fats and processed meats” dietary pattern (FPM), was also used to measure diet quality. Nettleton et al. (45
) developed the FPM by using a principal components analysis of MESA food frequency data. Higher scores indicate higher intake of fats and oils, high-fat and processed meats, fried potatoes, salty snacks, and desserts. Participants scoring in the bottom FPM quintile were classified as having a healthy diet. Persons with lower values for the FPM pattern had lower mean levels of biochemical markers related to early atherosclerosis in this cohort (45
Two MESA survey questions assessed fast-food consumption. To assess overall fast-food consumption, participants were asked to indicate “in an average week how often do you eat [eat in or take out] a meal from a fast-food place such as McDonald's, KFC, Taco Bell, or take out pizza places?” Five response options were provided: never, less than once per week, 1–2 times a week, 3–4 times a week, or 5 or more times a week. Because small numbers of participants consumed fast food more than 3 times a week, the top 3 response categories were collapsed into the single category, more than once a week.
To assess average fast-food consumption near home, participants who reported consuming fast food were also asked how frequently they “eat a meal from a fast-food place which was located within 1 mile (or 20-minute walk) from their home.” Response options were identical to those above. Participants who reported never eating fast food (i.e., from the first survey question) or never consuming fast food within 1 mile from home (i.e., from the second survey question) were classified as “never eating fast food within 1 mile of home.” All other responses were coded as “eating fast food within 1 mile of home.”
The following 3 measures of neighborhood exposure to fast food were investigated: self-report, informant report, and geographic information systems–derived densities of fast-food outlets. To estimate self-reported neighborhood exposure to fast food, each MESA participant was asked to consider his or her neighborhood as the area within a 20-minute walk or 1 mile around the home and indicate the extent to which they agreed (1
strongly agree to 5
strongly disagree) with the statement, “There are many opportunities to purchase fast food in my neighborhood.” Responses were reverse recoded so that a higher score indicated greater neighborhood exposure to fast food.
Informant-based measures of neighborhood exposure to fast food were created by aggregating survey responses of neighboring MESA participants. For each participant, we averaged the responses of all neighboring participants, referred to as informants, within 1 euclidean mile of their home. The median number of informants was 47 (25th and 75th percentiles, 16 and 147). By pooling the responses of multiple informants within a given geographic area, this approach may reduce noise resulting from individual subjectivities and improve the validity of the measure. Informant report measures may also avoid same-source bias that may arise if, for example, respondents who eat more fast food are more likely to report fast food near their home.
For each participant, we estimated the density of fast-food restaurants within a 1-mile window of the residence for the 3 study sites for which data on location of fast-food outlets were available (Maryland, North Carolina, and New York). The density of fast-food restaurants available per square mile was calculated by using the Spatial Analyst extension of ArcGIS v.9.2 software (ESRI, Inc., Redlands, California). Information on restaurants was purchased in November 2003 from InfoUSA Inc. (Omaha, Nebraska). Restaurants were identified on the basis of 33 nationally recognized chain names (47
). Densities were calculated by using kernel estimation so that restaurants located closer to the residence are given more weight than those located further away, with the weight approaching 0 at the boundary of the window (48
). The weights follow a bivariate normal (Gaussian) distribution (48
Three sets of statistical analyses using logistic regression were conducted to examine associations among neighborhood exposure to fast food, fast-food consumption, and diet quality. All models were adjusted for study site, participant age (years), sex, race/ethnicity (Hispanic, non-Hispanic white, Chinese, non-Hispanic black), education (less than high school, high school graduate, some college, college graduate, graduate degree), and annual per capita household income (not reported, $0–$9,999, $10,000–$19,999, $20,000–$29,999, ≥$30,000). Categories for annual per capita household income were calculated by dividing interval midpoints of 13 household income categories (in US dollars) by the reported number of persons within the household. Participants who did not report income (n
190) were included in models as a separate category. Heterogeneity of associations by study site was also examined by using interaction terms and stratified analyses, where appropriate.
In the first set of analyses relating diet to fast-food consumption, the odds of having a healthy diet (top quintile of AHEI or bottom quintile of FPM, in separate models) was modeled by the number of times fast food was consumed in an average week (never, <1, ≥1). The second set of analyses explored whether being exposed to more fast foods in neighborhoods was associated with consuming fast food near home. In this analysis, the odds of consuming fast food near home (1 = ate fast food near home, 0 = never ate fast food near home) was modeled as a function of neighborhood exposure to fast food. The third set of analyses examined the relation between neighborhood exposure to fast food and the odds of having a healthy diet (top quintile of AHEI or bottom quintile of FPM (separate models for each)). Neighborhood exposure to fast food was modeled in standard deviation units and was assessed by self-reports, informant reports, and densities of fast-food outlets in separate models.