The study population consisted of participants from the baseline cohort of the Australian Diabetes, Obesity and Lifestyle study (AusDiab); a, population-based prospective study of national diabetes mellitus incidence and associated risk factors. Non-institutionalised adults aged ≥ 25 years without physical or intellectual disabilities were invited to attend an initial household interview followed by a biomedical examination (which included an oral glucose tolerance test). Data was collected over a 21 month period between May 1999 and December 2000 in the six states and the Northern Territory of Australia from a total of 42 randomly selected urban and nonurban areas based on Census Collector Districts. The study methods, response rates, sample representativeness and main findings have been reported in detail elsewhere [16–18].
Briefly, 28,033 households were approached to participate in the study from the 42 selected clusters. Of the 19, 215 households successfully contacted, 17,129 were deemed eligible and 11, 249 agreed to participate in the initial household interview. A total of 20,347 eligible adults completed the initial household interview and 11, 247 (55.3%) attended the follow-up biomedical examination. From the 11,247 study participants, we further excluded those who were aged ≤ 35 years for whom fasting insulin data was not available (n = 1603), pregnant (n = 60), had been clinically diagnosed with diabetes (n = 475), had a history of cardiovascular disease (n = 938), reported taking anti-hypertensive or lipid-lowering medication (n = 2727), reported an implausible TV viewing time (>18 hr on a weekday or weekend day; n = 93), did not complete a food frequency questionnaire (FFQ; n = 204), reported an implausible dietary intake according to established criteria  (n = 582), had an invalid FFQ due to missing dietary data (>10% of FFQ; n = 476), or whose MetS status could not be ascertained due to missing data (n = 449). Exclusion criteria were not mutually exclusive, so participants could be excluded based on more than one criterion. There was no marked difference in the gender ratio of included and excluded participants (0.83 vs 0.80). However, those included were for the most part older, had lower BMI and had completed further education than those excluded. The final study sample consisted of 3,110 women and 2,572 men.
Assessment of metabolic syndrome components
Participants who attended the follow-up biomedical examination provided a fasting blood sample and underwent a 75-g oral glucose tolerance test at a local survey centre after an overnight fast (>8 hr). Plasma glucose levels (fasting and 2-hr post-load) were determined by a spectrophotometric-hexokinase method; fasting serum triglycerides and HDL-cholesterol were measured by enzymatic methods (Roche Modular, Roche Diagnostics, Indianapolis, USA) on an Olympus AU600 analyzer (Olympus Optical, Tokyo, Japan). Fasting serum insulin levels were determined using a human insulin-specific radioimmunoassay kit (Linco Research, St. Charles, MO). Microalbuminuria levels were determined using an immunoturidimetric method (Olympus AU600 analyzer). Height, weight, duplicate waist circumference and triplicate resting blood pressure measurements were conducted by trained personnel according to methods that have been previously published .
Diagnosis of the metabolic syndrome
Clinical criteria for the diagnosis of the MetS and its individual components was based on the 1999 WHO working definition  which includes the presence of either diabetes (fasting plasma glucose ≥7.0 mmol/L or 2-hr plasma glucose ≥ 11.1 mmol/L), impaired glucose tolerance (fasting plasma glucose <7.0 mmol/L and 2-hr plasma glucose ≥ 7.8 but <11.1 mmol/L) or insulin resistance (fasting insulin levels in the upper quartile of the non-diabetic population ) following a glucose tolerance test, plus two or more of the following; obesity (defined as a body mass index >30 kg/m2 or waist:hip ratio >0.9 for men or >0.85 for women), dyslipidemia (defined as triglycerides ≥ 1.7 mmol/L or HDL-cholesterol <0.9 mmol/L for men or <1.0 mmol/L for women), hypertension (defined as blood pressure ≥ 140/90 mm Hg) and microalbuminuria (defined as urinary albumin excretion ≥ 20 μg/min).
For comparative purposes, clinical criteria for the diagnosis of the MetS based on the harmonized definition provided by the International Diabetes Federation Task Force on Epidemiology and Prevention; Nation Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity was also analysed .
Assessment of TV viewing time
Self-reported TV viewing time was determined by asking participants to recall how much time they spent watching TV or DVDs, separately across all work days and non-work days during the preceding seven days. The recall measure only included time when TV viewing was the main activity and not when the TV was just switched on or when other activities were being undertaken concurrently. TV viewing time was then calculated using the following formula [(workdays TV viewing + non workdays TV viewing / 7] to determine hours per day spent watching TV. This measure has previously been shown to provide a reliable (intraclass correlation = 0.82 (95% CI: 0.75, 0.87) and valid (criterion validity = 0.3) estimate of TV viewing time in adults . Based on previously identified associations with individual components of MetS risk , daily TV viewing time (hr/d) was categorised into two groups: 0–2 hr/d (low) or >2 hr/d (high).
Assessment of snack food consumption
Daily consumption of snack foods was assessed using a self-administered, food frequency questionnaire (FFQ) developed by the Victorian Anti-Cancer Council of Victoria for use in Australian adults . The 74-item FFQ has previously been validated against a 7-day weighed food diary with correlation coefficients for energy-adjusted nutrient intakes in the range of 0.28 (vitamin A) to 0.78 (carbohydrate) .The present analyses are based on a subset of nine food items from the FFQ that represent common snack foods: crackers, sweet biscuits, cakes/pastries, meat pies/pastries/quiches or other, chocolate, flavoured milk, potato chips/crisps, ice cream, hot (fried) chips. Food items were selected according to the Australian Guide to Healthy Eating (AGHE) guidelines for “non-core” foods which are representative of common snack foods available today. Non-core foods are defined as foods that are not essential to provide nutrient requirements and contain too much fat, sugar and salt and are low in dietary fibre .
Participants reported how frequently they consumed the nine snack food items during the previous month, using the following seven response categories: never or not in the last month, several times per month, once a week, a few times a week, on most days, once per day and several times per day. Daily snack food consumption (number of serves per day) was determined using an established method of converting the frequency of consumption during the previous month for each food item into a daily equivalent score . To control for variation in the serving size of the nine food items, a gender-specific portion size calibrator was derived for each participant based on their individual responses to photos of scaled portions of four foods in the FFQ. Daily equivalent scores for each of the nine food items were then multiplied by the portion size calibrator and the resultant scores combined to calculate daily consumption of snack foods (serves/d).
Consumption of snack foods was categorised into two groups based on self-reported daily intake; 0–3 serves of snack foods per day (low) and >3 serves of snack foods per day (high). These pragmatic cuff-off values for the two categories were partly informed by the current AGHE gender and age-group specific recommendations for the consumption of non-core foods . According to the AGHE, the recommended daily consumption of non-core foods is up to 2.5 serves for women and up to 3 serves for men if aged 19–60 years.
Assessment of demographic and behavioral attributes
Demographic (age, gender, parental history of diabetes, education level, employment status) and behavioral attributes (leisure-time physical activity, smoking status, daily alcohol intake, diet quality) of participants were assessed using interviewer-administered questionnaires.
Leisure-time physical activity was assessed using the Active Australia Survey Questionnaire  which has been shown to provide a reliable (intraclass correlation = 0.59 (95% CI: 0.52, 0.65) and valid (criterion validity = 0.3) estimate of leisure-time physical activity amongst Australian adults [26, 27]; total leisure-time physical activity (min/wk) was calculated by methods previously described .
Self-reported FFQ data was used to calculate daily alcohol intake (KJ/d) and to derive a measure of diet quality (the Dietary Guideline Index) based on adherence to Australian dietary guidelines for the consumption of five “core” food groups (vegetables, fruits, cereals, meat and alternatives, dairy) and “non-core” foods (i.e. energy-dense, nutrient poor foods) (Diet Guideline Index –scale 1–130 with 130 representing high diet quality) [29, 30].
The Ethics Committee of the Baker IDI Heart and Diabetes Institute (formerly known as the International Diabetes Institute) approved the study, and written informed consent was obtained from all participants.
Statistical analyses were conducted using STATA Statistical Software Package, Release 11.0 (STAT, College Station, TX, USA) survey commands for analysing complex survey data. Sample weights based on the 1998 estimated residential Australian population aged >35 years were used to account for clustering and stratification in the survey design and for non-response. Data were stratified according to gender, due to a statistically significant interaction effect between gender and MetS risk (results not shown). Unadjusted means with 95% confidence intervals were calculated for continuous confounders (age, leisure-time physical activity, total alcohol intake and diet quality) and percentages with 95% confidence intervals for categorical confounders (education level, smoking status, employment status, parental history of diabetes). Regression analyses were used to determine gender-specific differences in potential confounders across joint categories of TV viewing time and snack food consumption. Bivariate correlations (Spearman’s r) assessed the gender-specific relationships between exposure variables, snack food consumption (serves/d) and TV viewing time (hr/d). Continuous measures of TV viewing time and snack food consumption were presented as geometric means with 95% confidence intervals. Odds ratios (with 95% confidence intervals) for the MetS and for individual components of the MetS according to separate categories of snack food consumption and TV viewing time were estimate using gender-specific, forced entry logistic regression models. Model A adjusted for age (years) only. Model B additionally adjusted for education level (completed university or higher education/no further education), smoking status (current smoker/non-smoker), employment status (employed/unemployed), parental history of diabetes (yes/no), leisure-time physical activity (min/wk) and daily alcohol intake (KJ/d). To examine the independent associations of TV viewing time (hr/d) and snack food consumption (serves/d), with risk for the MetS and its components, Model B was adjusted by each variable accordingly (Model C). To further control for the confounding effect of diet, Model C was additionally adjusted for diet quality (Model D). For individual components of the MetS (excluding the measure of obesity), a fifth model (Model E) additionally adjusted for waist circumference.
To test the joint association of TV viewing time and snack food consumption with risk for the MetS and its individual components, gender-specific, forced-entry logistic regression models that combined categories of TV viewing time and snack food consumption were created. The categories included: 0–2 hr/d of TV viewing and 0–3 serves/d of snack foods (reference category, low TV/low snack food), 0–2 hr/d of TV viewing and >3 serves/d of snack foods(low TV/high snack food), >2 hr/d of TV viewing and 0–3 serves/d of snack foods (high TV/low snack food ), >2 hr/d of TV viewing and >3 serves/d of snack foods (high TV/ high snack food). Models adjusted for age (Model A) and potential confounders including educational level, smoking status, employment status, parental history of diabetes, leisure-time physical activity and daily alcohol intake (Model B). Model C additionally adjusted for diet quality. For individual components of the MetS (excluding the measure of obesity), a forth model (Model D) additionally adjusted for waist circumference. For all statistical tests, a p-value ≤ 0.05 was considered significant. As there were no missing values for any exposure, outcome or confounder measures, the sample size is the same for all analyses.