The English Longitudinal Study of Ageing (ELSA) is an on-going cohort study containing a nationally representative sample of the English population living in households [18]. The cohort consists of men and women born on or before 29 February 1952. For the purpose of the present analyses data collected during wave 4 (2008 – 2010) were used as the baseline, as this was the first occasion on which information on television viewing was gathered. Clinical information was gathered by nurses in participants’ homes at wave 4 and at follow-up (wave 6; 2012/13). Participants gave full written informed consent to participate in the study and ethical approval was obtained from the London Multicentre Research Ethics Committee.
Exposure variables: baseline television viewing and physical activity
Television viewing time
Participants were asked “how many hours of television do you watch on an ordinary day or evening, that is, Monday to Friday” and “How many hours of television do you normally watch in total over the weekend, that is, Saturday and Sunday.” Average daily time spent watching television was calculated as [(weekday television time x 5) + (Weekend television time)]/7. Next, average daily television was categorised into four categories (<2 hours/day, ≥2 < 4 hrs/d, ≥4 < 6 hrs/d, ≥ 6 hrs/d). This categorisation of average daily television has been used in a previous study [19].
Physical activity
Participants were asked how often they took part in vigorous, moderate, and low intensity physical activity. Response options were: more than once a week, once a week, one to three times a month, and hardly ever/ never. Based on response options participants were then categorised into one of three groups (inactive/ moderate at least 1/wk/ vigorous at least 1/wk). For more information on this physical activity measure see Hamer et al. [20]. The physical activity and television viewing measures have been shown to have excellent convergent validity in grading a plethora of psychosocial, physical and biochemical risk factors [20-22].
Outcome: incidence of obesity
Research nurses measured participants’ body weight using Tanita electronic scales, participants were measured without shoes and in light clothing, and height was measured using a Stadiometer with the Frankfort plane in the horizontal position. BMI was calculated using the standard formulae [weight (kilograms)/height (meters) squared]. Research nurses recorded waist circumference twice mid-way between the iliac crest and lower rib using measuring tape. An average of the first two measurements was used provided these differed by no more than 3 cm; otherwise a third reading was taken and the two closest results utilised. Central obesity was defined as >102 cm in men and >88 cm in women [23].
Covariates
Age, sex and long standing illness (yes/no) were self-reported. Trained interviewers asked questions on smoking (current, previous, or non-smoker), alcohol intake (daily, at least once a week, monthly, rarely, never), and depressive symptoms (using the eight-item Centre of Epidemiological Studies Depression Scale [24]). Disability was assessed based on participants’ responses to interviewers’ questions on perceived difficulties in six basic activities, such as difficulty dressing, and seven instrumental activities of daily living, such as preparing a hot meal [25]. Participants with difficulties in one or more activities were considered to have some degree of disability. Use of prescribed medication (including medication for diabetes, high blood pressure, cholesterol, blood thinning) was self-reported. These covariates were included in the analyses because they were all hypothesised to be independently associated with both exposures (television viewing time and physical activity) and outcomes (obesity and central obesity).
Statistical analyses
Characteristics of the study population at baseline were described as means (continuous variables) and percentages (categorical variables). Obese participants (BMI ≥ 30) at baseline were removed from the analysis. We calculated odds ratios (OR) and 95% confidence intervals (CI) for the risk of obesity in relation to television viewing categories (<2 hrs/d, ≥2 < 4 hrs/d, ≥4 < 6 hrs/d, ≥6 hrs/d) using multiple logistic regression. The models were adjusted for age, sex, physical activity, smoking, alcohol, depressive symptoms, long standing illness, disability, cardiovascular (CV) medications. Similar models were run to examine the association with incident central adiposity as the outcome after removing participants with central adiposity at baseline. We tested for statistical interactions with respect to sex although none were found. Thus men and women were pooled together and the analyses were adjusted for sex. All analyses were conducted using SPSS version 21.