Students' height and weight were measured by using a standard protocol (18). BMI was calculated from the measured height and weight as the weight in kilograms divided by height in meters squared. For children, the Centers for Disease Control and Prevention (CDC) defines obesity as a BMI greater than or equal to the 95th percentile for age and sex, overweight as a BMI greater than or equal to the 85th percentile but less than the 95th percentile, and normal as a BMI less than the 85th percentile but greater than the 5th percentile (www.cdc.gov/growthcharts/).

Self-reported demographic information; nutrition behaviors, attitudes, and knowledge; and physical activity behaviors were assessed by the SPAN questionnaire. The questionnaires were administered on Tuesday through Friday so that responses to questions concerning the previous day reflected weekday behavior. The questionnaire was found to have acceptable reliability and validity compared with similar dietary assessment instruments for both elementary- and secondary-level students (19-21).

Daily television watching, snacking behavior, and physical activity (used as a covariate) were determined by single questions. The daily consumption of the foods most frequently advertised on television shows targeted to children (television foods) was assessed by 5 questions measuring punch, sports drinks, and other fruit-flavored drinks; sodas and soft drinks; frozen desserts; sweet rolls, doughnuts, cookies, brownies, pies, and cakes; and chocolate candy (15). A composite score for the frequency of television foods was calculated by summing the responses to these 5 questions. The questions and possible responses are listed in the Appendix.

We used Stata 9.0 (StataCorp, College Station, Texas) to analyze all data. The Stata survey procedure was used to account for the stratified, weighted, cluster-sampling procedure when conducting hypotheses testing and creating confidence intervals (CIs).

Weighted frequencies of television viewing, snack consumption, and foods advertised on television were calculated for each grade and sex. For each grade, we calculated odds ratios (ORs) with 95% CIs to demonstrate whether overweight status occurs more frequently with high levels of each of 3 independent variables (television viewing, snack consumption, or television foods). ORs were adjusted for sex, race and ethnicity, and physical activity.

To examine the relationships between television viewing, snack consumption, and television foods, we dichotomized the 3 variables as follows: daily television viewing, 0-2 hours of television vs 3 or more hours of television; daily snack consumption, 0-1 snacks vs 2 or more snacks; and daily television foods,  0-3 television foods vs 4 or more television foods.

The differences in the prevalence of overweight based on television viewing and snack consumption were analyzed by using a multiple logistic regression strategy. First, we used a typology of the television viewing and snack consumption variables to classify each child as follows: 1) 0-2 hours of television per day and 0-1 snack per day (low television/low snacks); 2) 0-2 hours of television per day and 2 or more snacks per day (low television/high snacks); 3) 3 or more hours of television per day and 0-1 snack per day (high television/low snacks); or 4) 3 or more hours of television per day and 2 or more snacks per day (high television/high snacks). We performed a separate regression analysis for each grade by using sex, race and ethnicity, and physical activity as covariates. We calculated adjusted ORs to compare the prevalence of overweight in 3 of the categories (low television/high snacks, high television/low snacks, and high television/high snacks) to the prevalence of overweight in low television/low snacks category. We used a similar multiple logistic regression strategy to analyze the differences in the prevalence of overweight based on television viewing and television foods.

Fourth-grade students who ate 2 or more snacks per day were 77% more likely to watch 3 or more hours of television per day than were those who ate 1 or fewer snacks per day. Eighth-grade students who reported eating 2 or more snacks per day were 44% more likely to watch 3 or more hours of television per day than those who ate no snacks or 1 snack. Students who ate 4 or more television foods per day were more likely to watch 3 or more hours of television per day than those who ate 3 or fewer television foods per day. Furthermore, both fourth- and eighth-grade students who ate 2 or more snacks per day were more likely than those who ate 1 or fewer snacks per day to also eat 4 or more television foods per day.

Eighth-grade students who watched television 1 to 2 hours per day were more likely to be overweight than were those who watched less than 1 hour per day (Table 3). We found no significant relationships between television viewing and overweight for fourth-grade students. Children in both grades who ate 2 snacks per day and 3 or more snacks per day were less likely to be overweight than were those who ate none. Fourth-grade students who ate 1 snack per day were 44% less likely to be overweight than were those who ate no snacks. Eighth-grade students who ate 2 or 3, and 4 or more television foods per day were less likely to be overweight than those who ate no television foods or 1 television food per day. We saw no difference in the likelihood of overweight by television foods for fourth-grade students (Table 3).

Fourth-grade students who watched little television and ate many snacks were 45% less likely to be overweight than were those who watched little television and ate few snacks (Table 4). Eighth-grade students who watched little television and ate many snacks per day and those who watched many hours of television and ate many snacks per day were also less likely to be overweight than were those who watched little television and ate few snacks.

Eighth-grade students who watched little television and ate many television foods per day and eighth-grade students who watched many hours of television and ate many television foods per day were less likely to be overweight than were members of the referent group (Table 4).

Television viewing and snack consumption were positively associated for students in the fourth and eighth grades, which supports the results of previous studies on the relationship between television viewing and snacking (4,5,11). Additionally, the relationship between television viewing and television foods was straightforward; the more television foods eaten, the more television watched. Similar results have been found by many researchers and attest to the power of food advertisements on television (12,13,17,25-27). Although SPAN had questions on television viewing and snacking behaviors, there was no question linking snacking behaviors to watching television; thus, we cannot know if the snacks or television foods were eaten while watching television or eaten directly because of commercials.

Generally, children who reported eating more snacks per day were less likely to be overweight than those who consumed fewer snacks per day regardless of hours of television viewing. The pattern for television viewing and television foods was similar. These results were surprising because of the unhealthy nature of the television foods. Although these analyses were adjusted for sex, race and ethnicity, and physical activity, they were not adjusted for other factors that might be associated with the development of overweight, such as socioeconomic status, other sedentary activities, total caloric intake, or overall quality of diet. The SPAN survey was designed to examine weekday habits; including weekend and weekday habits may yield different results, particularly because the number of food advertisements aimed at children is high on weekend days, especially Saturday morning (28). Although our analyses were adjusted for race and ethnicity, the results of this study may have been influenced by the large proportion of Hispanic SPAN participants, in a way that other, less multiethnic studies are not. Finally, the SPAN survey had a truncated number of options for each of the behaviors (eg, 3 or more snacks per day). A different relationship might exist if we had been able to distinguish between higher frequencies (5 snacks) and those in the middle (3 snacks).

The major strength of this study is that it was based on a large, diverse sample that is representative of fourth- and eighth-grade children in Texas public schools. The SPAN questionnaires were tested for reliability and validity before being used (19-21), and BMI was calculated by using objectively measured height and weight.

This study is limited by the use of cross-sectional data and self-reported data for television viewing, snacking behaviors, and race/ethnicity. Because the study was cross-sectional, we were able to look only at relationships that existed when the questionnaire was administered. Children self-reported their television viewing, snacking behaviors, and ethnicity, which has the potential to create inaccuracies or skew the data. Overweight children are more likely than children of normal weight to underreport their food intake (29,30). A selective underreporting of food or dieting by restricting certain foods, such snacks and television foods, by overweight children in the SPAN study could have contributed to the results. Additionally, SPAN assessed the frequency of consumption of snacks, not portion sizes, and hours of television watched, not type of programming (eg, movies, educational, or commercial television), which limits our ability to interpret the meaning of eating 2 snacks or watching 2 hours of television without having more information about either behavior.

Snacking, eating foods commonly advertised on television, and watching television were all frequently reported by students participating in the SPAN survey. The results suggest that the relationships between weekday snacking behaviors and television viewing in a multiethnic population are complicated, and relationships to overweight status may be related more to the number of snacks than to the amount of television watched. To determine the exact relationship, additional research into the food type and serving size of snacks, timing of snack consumption with television watching, the amount and type of television watched, and racial/ethnic differences in snacking and television viewing, especially among Hispanic children, is warranted.