Participants were recruited from GP practices in Leeds, UK, following ethical approval from Leeds (West) Research Ethics Committee in August 2002. Posters and flyers were placed in patient waiting areas, advertising the study and asking interested potential participants to call the study centre or to inform their GP or practice nurse of their interest. Potential participants were screened by telephone for eligibility. Eligibility criteria included: individuals with a body mass index (BMI) of 30 or more, aged 18–65 years (due to body composition changes over the age of 65 years), able to access the Internet at least once per week and able to read and write in English (for the purposes of accessing the website and completing questionnaires).
On confirmation of informed written participant and GP consent, a baseline appointment was scheduled for each eligible participant, where height and weight were measured by the researcher and a baseline questionnaire completed, before each participant was randomly allocated into either the Internet group (n = 111) or the usual care group (n = 110). Weight (in light clothing) was measured to the nearest 0.1 kg using Marsden portable weighing scales. Height (without shoes) was measured using a portable height measure to the nearest 0.1 cm. Recruitment into the trial took place from May to November 2003. Participants were followed-up six months and 12 months after randomisation, when height and weight measurements were repeated by the researcher, along with a follow-up questionnaire administered by post. Baseline and follow-up appointments were conducted at the participants' GP practice.
All non-responders were followed up following a strict protocol by post and telephone. Three attempts were made to contact each participant not responding to the initial mailing at follow-up by telephone, following this a second questionnaire was mailed to participants asking them to complete the questionnaire even if they did not wish to attend a follow up visit.
Current research evidence supports a lifestyle approach to treating obesity, offering a combination of dietary advice, physical activity advice and behaviour therapy [14–17]. Based on these guidelines and clinical evidence, the intervention website was developed to reflect these factors. The website provided advice, tools and information to support behaviour change in terms of dietary and physical activity patterns. It was designed to enable patients to manage their own care and to vary the frequency of use according to their own needs. The website also offered personalised advice to participants, which, in the context of this trial, involved targeting the information provided to an individual, based on their responses to a series of online questions regarding eating and activity habits and current weight status. This enabled specific motivational statements to be generated to participants whenever they visited the website. Motivational statements were generated based on participants self report of progress in terms of reaching their personal behaviour change goals. In addition, details of progress in terms of self-reported weight loss were stored on the website, accessible only to the individual concerned. Automatic generic e-mails were generated if participants did not visit the website regularly to encourage them to visit more often. The website and questionnaires were piloted in a sample of overweight University staff. The results of this pilot were used to inform the final version of the website.
The trial aimed to compare the additional effect of the website against usual care available within the UK. Participants randomised into the Internet group were given a demonstration of the website and its services, along with a username and password to access the website and were asked to log on to the website at least once a week over the trial period. Participants randomised to the usual care group were advised to continue with their usual approach to weight loss and were given a small amount of printed information at baseline, reflecting the type of information available within primary care.
The primary outcome was the ability of the Internet package to promote change in weight and BMI over six and 12 months compared with usual care. Secondary outcomes were the ability of the Internet package to promote change in reported lifestyle behaviours compared with usual care, along with differences in quality of life. Lifestyle and dietary habits were assessed with a questionnaire previously used in the UK Women's Cohort Study . This questionnaire obtained information on methods of cooking, portion size and frequency of consumption of various foods and participants' 'dieting' practices. Physical activity level was assessed using the Baecke physical activity questionnaire which measures work, leisure and sports activity providing a comprehensive evaluation of habitual physical activity . Quality of life was assessed using the EuroQol questionnaire, a short, self administered questionnaire which was also used in the cost effectiveness analysis . A brief series of questions were used to assess participants' confidence in their ability to make positive lifestyle changes on a scale of one to seven (where one indicates not at all confident and seven indicates very confident). These measures were combined in one questionnaire, which participants completed at baseline, six months and 12 months. Additional questions were added at six and 12 months for the purpose of the cost-effectiveness analysis of the programme. Participants in the Internet group also completed an additional section on their use and views of the website at six and 12 months.
A sample size of 180 participants was required to detect a difference of 5 kg weight loss (approximately 5% of body weight) or less than 2.5 kg/m2 in BMI between the two groups with 80% power, assuming a two-sample t-test, 5% significance levels, a standard deviation for weight of 12 kg and for BMI of 5.5 kg/m2. An additional 22% of participants were recruited to take account of any loss to follow-up, giving a recruitment target figure of 220.
A computer-generated randomisation procedure was employed, using the software package 'minim' . Participants were allocated to groups by the programme according to the minimisation criteria, i.e. balanced for gender (male/female), age group (18–34, 35–49, 50+) and BMI category (30–33.9, 34–37.9, 38+). Due to the pragmatic nature of the trial and the intervention being evaluated, it was not possible to blind either the participants or researchers to the group assignment.
Analyses were performed using Statistical Package for the Social Sciences (SPSS for Windows, version 11.5; SPSS, Chicago, IL). Independent sample t-tests (or non-parametric tests where appropriate) and chi-squared tests were used to investigate differences in baseline characteristics and response rates between the two groups. Analysis of covariance (ANCOVA) on weight at 12 months was used to investigate the difference between the two groups. This adjusted for any imbalance in age, sex, baseline weight, baseline physical activity score or baseline confidence score introduced by losses to follow-up. Changes in secondary measures were investigated using ANCOVA adjusting for possible baseline imbalances as before. Primary analyses were conducted based on all available data. Analyses using LOCF and BOCF were performed to assess the robustness of the primary analysis for the effect of losses to follow up and missing data.
The cost-effectiveness analysis was undertaken in Stata SE 8.2. Quality adjusted life years (QALYs), calculated using utilities collected by the EuroQoL (EQ-5D), were compared with total costs estimated using a societal perspective. Costs were estimated from a variety of sources, including the Personal Social Services Research Unit (PSSRU) for visits to the GP and practice nurse . Adjusted (for characteristics of patients that differed between the group) estimates of mean cost and QALYs were obtained by a multi-level difference-in-difference econometric model . The incremental cost-effectiveness of the two alternative methods of weight loss support based on mean differential costs and QALYs was then established. Finally the probability that Internet-based support is cost-effective according to a range of alternative threshold values, which the health care system may be willing to pay for an additional QALY, was calculated . This is then plotted as a cost-effectiveness acceptability curves (CEAC) . The CEAC shows the proportion of simulations (produced by the econometric model) in which web-based support (and conversely traditional weight loss) is the more cost-effective across a range of alternative threshold values, which the health care system may be willing to pay for an additional QALY.