We aimed to recruit men who had a diagnosis of type 2 diabetes and had a BMI of at least 30 kg/m² but less than 40 kg/m² at baseline measurement. The BMI inclusion criterion was 30 kg/m² or greater, as this is the inclusion criterion for the majority of NHS tier 2 (lifestyle interventions) or tier 3 (specialist services) weight management services in England [34]. As this study was examining the change in service delivery of weight management within the NHS, we followed the criterion used within the NHS. When a patient reaches a BMI of 40 kg/m², lifestyle modification may no longer be appropriate and bariatric surgery may be recommended [35]. Men had to be aged 18 years or older, with no upper age restrictions.

Participants were required to have access to the Web (at home, the workplace, or a public location) on any device (desktop computer, laptop, tablet, or mobile phone).

We excluded patients unable to give written informed consent or access the intervention in English (resource constraints precluded adaptation of the intervention for non-English speakers) or who were identified by their general practitioners (GPs) as having a contraindication to the weight loss intervention (such as previous eating disorders or other mental health problems).

The study was accepted onto the UK National Institute for Health Research clinical research network portfolio and registered (October 26, 2012) on a clinical trial registry (ISRCTN: 48086713). We gained NHS ethical favorable opinion from National Research Ethics Service Committee East of England - Cambridge Central Proportionate Review Sub-committee on August 9, 2012 (Research Ethics Committee reference: 12/EE/0361).

We recruited participants through GP database searches; we identified participating practices through the UK Primary Care Research Network. In response to participant invitation letters, potential participants could state their intention by completing an attached reply slip and returning by reply-paid mail to the research team. Participants could also contact the research team directly via email or telephone. Those who did not want to take part in the research could return the slip and (optionally) provide a reason why. During baseline appointments in GP offices, participants provided written informed consent to the researcher (AH) prior to baseline measurements.

The researcher (AH) then randomly allocated each participant to 1 of the 2 arms using the Sealed Envelope Web-based system (Sealed Envelope Ltd). Randomization was by a one-to one allocation, to either usual care (control group) or the Web-based intervention group. We used stratification to ensure that the potential confounding variable of diabetes medication was balanced between the intervention and control arms, since this might affect outcomes. The strata were diet only, oral hypoglycemic agents, or insulin. Participants who were taking insulin and tablets were assigned to the insulin stratum. Participants were informed of allocation via postal letter by the researcher (AH). Blinding of intervention allocation was not possible for anyone involved in the pilot trial.

The control arm experienced usual care for weight loss, according to their general practice’s normal processes. This was a pragmatic trial and we did not seek to influence what was offered to the patient, with no specific arrangements to review or refer participants.

Participants randomly allocated to the intervention group were sent log-in details and encouraged to log in to the intervention website [36] before their initial face-to-face consultation with their assigned dietitian.

We asked participants randomly allocated to the intervention to state whether they were engaged in any other weight loss services. None of the intervention group were using any other services.

The website (My Dietitian) was created by PraksisCare (Odense, Denmark) based on a previous study that had identified successful weight loss via a Web-based intervention [37]. We worked together with PraksisCare to develop and adapt the My Dietitian intervention to make it relevant for use within the United Kingdom and the NHS. Table 1 describes the intervention based on the Template for Intervention Description and Replication checklist [38].

A key feature of the 12-month intervention was the Web-delivered consultations (embedded email-style messages sent within the website), which were delivered to participants by health care professionals (dietitians and exercise experts). Consultations were delivered in accordance with a scheduled protocol created prior to the start of the study (Table 1). The initial one-off face-to-face meeting with the dietitian was conducted in an hour-long appointment slot. Dietitians were expected to provide Web-based consultations on a maximum weekly basis for the first 3 months (n=12) and then monthly for the last 9 months (n=9; total planned dietitian contact n=21). Exercise experts provided Web-based consultations on a maximum monthly basis for the first 3 months (n=3) and then every 3 months for the last 9 months (n=3; total planned exercise expert contact n=6). There were thus 15 planned consultations (diet and activity) in total over the first 3 months and 27 by the end of the 12-month intervention.

Time taken to write the consultations varied across the participants based on the required advice. The content of the consultations was at the professional discretion of the dietitians and exercise experts. Every intervention participant received personalized Web-based consultations from their designated dietitian and exercise expert. This feedback was based on participant input on the website and was typically concerned with areas of improvement in relation to dietary intake and physical activity.

Website pages allowed participants to record their personal daily dietary intake (Figure 1), physical activity, or weight status, which was viewed by the health care professionals. Participants were advised to enter their dietary intake (all meals, snacks, and drinks) and physical activity on a daily basis. Using the website features was important so that the health care professionals were able to provide thorough consultations to the participants. Other features were less interactive but were provided to inform or encourage participants, such as a database of recipes, relevant articles on physical activity, diet and weight loss advice, and the ability to chat online with other participants.

We recruited and trained 2 NHS dietitians and 2 exercise experts to work on the study intervention, the My Dietitian website. The health care professionals provided quality assurance checks on the content of the website. Training consisted of 2 half-day sessions covering behavior change techniques relevant to weight loss, an overview of the Web-based intervention, and practical sessions to enable the health care professionals to become familiar with and competent at using the website.

The primary outcomes were recruitment and retention in the trial, as measured by rates of eligibility, response to invitation, ineligibility, declines, consent, and retention for data collection at 3 and 12 months. Within the parallel process evaluation, we examined adherence to the intervention through collection of website use data. We examined acceptability of the intervention by conducting semistructured interviews with participants.

Secondary outcomes were comprehensiveness and feasibility of the measures proposed as primary or secondary outcomes in the future definitive RCT (anthropometric measures: body weight, height, BMI, and waist circumference). Losing 5% of one’s initial body weight is a target recommended by UK National Institute for Health and Care Excellence guidelines to improve health [35]. Another secondary outcome was parameter estimates of the proposed primary and secondary outcomes measures for the future definitive RCT to inform sample size calculations.

Reporting of the rehearsal pilot RCT follows the Consolidated Standards of Reporting Trials (CONSORT) statement extension to randomized pilot and feasibility trials guidelines [39] (see Multimedia Appendix 1 for the study’s CONSORT checklist [40]).

Data collection points were at baseline, 3 months, and 12 months and were completed in GP offices. Rates of eligibility, recruitment, randomization, retention, attrition, and adherence were logged from the initial invitation letters through intervention allocation (baseline) to follow-up (3 and 12 months), enabling an audit trail to be maintained [41].

We collected process evaluation data for adherence by tracking and monitoring website use. Website data were collected in relation to number of website log-ins for participants, self-monitoring diaries completed (food and dietary intake and exercise entries), number of consultations made by the health professionals to the participants, and Web-based messages sent to the health professionals by the participants. Any food and dietary intake that was entered on the website on a given day was classified as 1 entry, and the same also applied to exercise entries.

We examined acceptability and feasibility through semistructured interviews conducted once with participants, lasting between 15 and 60 minutes, at the end of the 12-month study.

Anthropometric measures were collected at baseline, 3 months, and 12 months. We recorded height using a Leicester height stadiometer (Marsden Weighing Group Ltd, Rotterham, England), with participants asked to stand as straight as possible with their shoes off. Body weight was measured using calibrated Shekel personal floor scales (H151-7, Class III; Shekel Scales Ltd, Lower Galilee, Israel), allowing capacity weight up to 250 kg. Each participant was required to produce 2 body weight readings to check for consistency (within 0.1 kg). If these were not consistent then a third reading was required, and the average of the 3 readings was used as the final body weight recording. Participants remained clothed but were asked to remove coats and shoes. The Shekel scales also allowed a participant’s height to be entered along with body weight to calculate a BMI recording. Waist circumference was measured, midway between the lowest rib and the iliac crest, underneath clothing, using a tape measure.

We examined the number of participants and percentages to identify rates of eligibility, response to invitation, ineligibility, declines, consent, and retention at baseline, 3 months, and 12 months.

The recorded website data were examined to identify average number of log-ins for participants, self-monitoring diaries completed (dietary intake and exercise inputs), consultations by health professionals to the participants in comparison with scheduled consultations, and diaries and messages sent to the consultants by participants. We also examined adherence to the intervention in terms of the number of users and nonusers of the website at each time point to identify adherence over time and by population group.

Interviews were recorded and transcribed. Qualitative data were imported into NVivo 10 software (QSR International) and analyzed using framework analysis, a 5-step process: familiarization, identification of a thematic framework, indexing, charting, and mapping (interpretation) [42,43].

We used descriptive statistics to characterize rates of completion for anthropometric measures, rates of implausible values, and 5-figure summaries (minimum, maximum, median, and lower and upper quartiles). We also calculated means and standard deviations to inform sample size calculations for a potential definitive trial. Quantitative data were analyzed using IBM SPSS Statistics version 21.0 software (IBM Corporation).

A total of 8 general practices agreed to perform database searches to identify potential participants, but 1 did not complete the searches due to time pressures. Practice size ranged from 1663 to 19,976 patients and varied in terms of location: town centers (n=3), housing estates (n=2), and rural villages (n=2).

Figure 2 shows the CONSORT diagram. GP database searches achieved the recruitment target after a period of 5 months, with a total of 61 men providing consent and being randomly allocated. The first patient was recruited during November 2012 and March 2013, with the last 12-month follow-up appointment held in March 2014. Of the participants, 66% (40/61) completed 3-month follow-up measurements. By 12 months, retention rates had 32/61 (52%) of the men remaining.

Men with diabetes (n=968) were identified from database searches (Figure 2). The response rate, expressed as those who stated an interest in joining the study as a percentage of those contacted by their practice, was 85 of 968 (8.8%). Only a small proportion of these were found to be ineligible at the baseline appointment due to changes in BMI (5/968, 0.5%).

Of those invited to the study, 187 of 968 (19.3%) of the men with diabetes explicitly declined to the invitation letter, and 115 of 187 (61.5%) did not give a reason for nonparticipation. The most common reasons for declining participation were no Web access (29/187, 15.5%), work commitments (12/187, 6.4%), poor health (10/187, 5.4%), and age (8/187, 4.3%). Those who declined due to lack of Web access did so in response to the invitation letter. Although this meant they were not eligible for participation, as these patients declined the invitation letter before any contact with the research team, they were never seen at the baseline assessment and therefore were not formally excluded. The vast majority of invited participants did not respond to the letter (696/968, 71.9%).

Data completion rates and retention rates decreased over the study time period (baseline to 3 months to 12 months). For the control group, 12 of 28 (43%) participants had dropped out by 3 months, and 16 of 28 (57%) had dropped out by the end of the study (12 months). For the intervention group, attrition was lower, with 9 of 33 (27%) participants dropping out by 3 months and 13 of 33 (39%) dropping out by 12 months. The main reasons for participants leaving the study were being too busy or having family or work commitments, for both the intervention- and the control-arm participants (Figure 2).

Table 2 shows baseline demographic and anthropometric descriptive statistics.

Change variables relate to change in measurements from baseline to 3 months and baseline to 12 months (Table 3). All changes, at both 3 and 12 months, were decreases (ie, positive changes). Of the remaining men in the intervention, 8 lost 5% of their initial body weight at 12 months in comparison with 4 of the control group. In terms of mean weight loss, intervention participants in our study lost on average 5.4 kg (Multimedia Appendix 2).

We examined adherence to the intervention in terms of users versus nonusers among those allocated to use the website (intervention group) (Table 4). “Users: relates to participants who entered information onto the website. “Nonusers” were those participants who never logged on to the website or only logged on to register on the website and did not enter any inputs.

At 3 months, the 9 intervention men who dropped out of the study had all been nonusers of the website from the outset (Figure 2).

We analyzed website use for those remaining in the intervention group (Table 5). We collected data from 28 men, as 5 intervention participants never registered on the website. Data were examined in relation to their website use between baseline (0 months) and the 3-month data collection point. Data were then examined for website use between 3 months and the end of the study (12 months): 9 intervention participants left the study at the 3-month data collection point and therefore we included 19 men in the analysis. We divided website activity to examine participant’s use in relation to food intake and exercise levels, as well as interactions with allocated dietitians and exercise experts (Table 5). As outlined in Table 1 (intervention description), the proposed number of consultations within the intervention protocol over the first 3 months was 15, with the total number of consultations at the end of the 12-month intervention stated as 27. Table 5 shows that 13 of the 15 intended health professional (dietitian, exercise expert) consultations were delivered at 3 months, whereas at 12 months, 22 out of the 27 were delivered. At the end of the intervention, the number of dietitian-delivered consultations was 1 less than the number scheduled to occur (20 out of 21); however, exercise experts averaged up to 4 fewer than scheduled (2 out of 6). Consultations were delivered fewer times than expected due to nonresponsive participants and time pressures on the health care professionals due to competing demands (eg, full-time jobs). Food-related messages from participants to the health professionals were sent more often than exercise messages. The food intake entries were used more frequently than the exercise entries, with both demonstrating high levels of variability between participants (Table 5). No log-in data were obtained at 3 months, as the website hosts did not record this information, and therefore Table 5 shows only 12-month results. Participants did not use the social support features, interactive chat room, and discussion forums at all. We identified no unintended effects were identified from the pilot trial.

We conducted 13 semistructured interviews, which enabled us to explore participants’ views in relation to the acceptability of the Web-based weight loss intervention. We classified 7 of the interviewees as active users, participants who logged on to the website and entered information until the end of the study, while 6 of the interviewees were nonusers and therefore were no longer using the website by the end of the study. Interviewees’ ages ranged from 41 to 78 years.

The main themes identified from the interviews were (1) practicality, (2) interaction with the health care professional, and (3) future development of the Web-based intervention.

We calculated sample size, as issue 1 in Table 6 reports [44], using PS: Power and Sample Size Calculation version 3.1.2 [45] to identify the sample size needed for a main trial, while we based our other calculations on response and retention rates from this pilot study. Sample size was calculated based on the main trial being a superiority trial and for a target difference of 5% between the control and the intervention arms in terms of percentage weight change from participants’ initial baseline weight to follow-up at 12 months (eg, 5% loss in the intervention group versus 0% in the control group, or 6% vs 1%), with 90% power, 5% significance level, one-to-one allocation, and analysis with independent-samples t test; a standard deviation of weight loss change of 5.6% was assumed.

The CONSORT guidelines recommend that pilot trials provide prespecified criteria to judge whether to proceed with a future trial [39]. A previous study identified 14 methodological issues for feasibility research to consider when making the decision to proceed with a future trial [44]. Therefore, Table 6 outlines this study’s findings, evidence, and suggested improvements for a main trial. Findings detailed in Table 6, along with the overall findings of the study, have enabled the recommendation that a main trial of the intervention should not proceed without the modifications and improvements identified. The main challenges that arose within this study were the low response rates to express interest in joining the study and low retention rates. Table 6 outlines possible solutions to these problems. A potential solution to the low response rates to join the study is the use of incentives for participation to try to increase uptake and recruitment. However, we achieved the recruitment target of 60 participants; therefore, although a low number responded to the GP invitation letters, we were also able to reach a large audience. In relation to the challenge of retaining participants in the study, the interview findings identified that improvements to the website would have encouraged participants to use the website more frequently and would have upheld their interest to a greater degree. Participants viewed the website as acceptable. The conversion to consent, recruitment, and randomization protocols were all found to be effective.

We achieved the recruitment target. Participant interviews identified the Web-based intervention as an acceptable method of delivery for weight loss; however, improvements to the website were suggested in relation to ease of use and to maintain adherence. Participants’ Web-based messages to health professionals tended to be directed to the dietitians rather than the exercise experts. However, it was also evident that the dietitians achieved more of their scheduled number of professional-initiated consultations than the exercise experts. Data completion rates at each time point were sufficient to inform sample size calculations for a future definitive trial.

To address problems identified in pilot studies, solutions should relate to study context, trial design, the intervention, or all 3 of these, and whether these could be effective or feasible within trial or real-world settings [46]. This study has examined these feasibility issues noted in Table 6.

In agreement with this study, findings from previous research [47] identified Web-based weight loss interventions as acceptable and feasible, including Brandt et al [37], the Danish study that originated the My Dietitian website.

We used general practices in this study, as this is how patients would be referred to exercise experts or dietitians for weight loss within the NHS. Within this pilot trial, only 9% of those invited expressed an interest in the study, but we met the recruitment target of 60 participants. A previous study that contacted patients via GP mailouts achieved a 6.5% response rate [48]. A suggestion to improve research study recruitment strategies, such as GP mailouts, is the use of opt-out techniques. Although these are disliked by ethics committees, previous research suggested it can increase response rates by 12% and should be used in low-risk groups, as opt-in techniques can result in a biased sample [49].

Previous research has shown great variability in Web-based weight loss trial recruitment levels, ranging from 6% to 83%, and in terms of the recruitment techniques implemented. Previous methods for recruitment range from a wider audience approach, such as advertisement techniques [50-53], to a targeted approach through GP mailouts or referrals [47,54], with varied success. Many of the studies used several methods of recruitment rather than one single approach [50,51,55,56]. One difficultly when comparing against previous research is the reporting of response and recruitment numbers. Published work can often report on the number screened for eligibility assessment and not the actual numbers who viewed the invitation, with the reach of some recruitment strategies, such as advertisements, unknown [37,47].

Attrition levels from previous studies range from 17.4% to 51.4% for Web-based intervention arms and 15.2% to 35.5% for control arms [47,53,57] at 12 months. This study experienced higher rates of attrition for the control arm but rates similar to those in previous research for the intervention arm. However, the control group had higher attrition than in previous research. Control groups have been discussed by Morgan et al [58], with the suggestion that a minimal intervention is necessary due to some form of intervention being more acceptable to participants than no intervention, which therefore prevents attrition, with attrition rates identified as 29% by 12 months. Tate et al [47] had similar attrition rates for both included groups—an Internet-only group and an e-counselling group—with an overall attrition of 16%. Their study used incentives for appointment attendance, which could be a potential improvement for our study. Our study used usual care, which we discovered to be near nonexistent in terms of specific weight loss treatment, with only 1 participant being referred to exercise classes. For this study, it was deemed important, and was achieved, to identify what usual care constituted for this population within the NHS. However, an improvement may be the use of a minimal intervention as the control group. Further investigation into different modes of delivery would also be beneficial to identify whether delivering an intervention in person or via the Internet would affect the overall findings.

In terms of mean weight loss, intervention participants in our study lost on average 5.4 kg, which is greater than in previous studies, which had losses of 4.4 kg [47], 4.6 kg [54], and 5.3 kg [58], although lower than Brandt and colleagues’ study, in which mean weight loss was 7 kg [37].

In addition to the pilot RCT, it was possible to conduct a process evaluation alongside the trial, which enabled us to investigate participants’ views and to track website use. We attributed the nondelivery of over half of the exercise consultations to the change in job status of the exercise experts, as well as nonresponsive participants. However, these reasons emerged from the process evaluation interviews. The health care professionals were not required to provide reasons for nondelivered consultation, and this was not recorded in the website use data. Therefore, we do know how many consultations were not delivered due to nonresponsive participants and how many were due to the health care professionals. A potential improvement going forward would be to require the health care professionals to enter reasons for missed consultations. As the health care professionals did not meet the proposed number of consultations, we suggest that large-scale trials should employ health care professionals as research study staff. If this is not possible, at least providing health care professionals with dedicated time to work on the study would be beneficial to the fidelity of intervention delivery rather than having their involvement with the study being in addition to their other full-time employment.

Participants were sent details of their username and operational directions in a postal letter in advance of the face-to-face meeting. In hindsight, the face-to-face meeting could have been a potential opportunity to explain the functions and resources within the website and explore how to effectively interact with it. This identifies a potential improvement and refinement of the intervention procedures for future research.

The intervention is reliant on feedback from a health care professional and, unless provided through the NHS, this would not be possible in a standard weight loss website. However, this study demonstrates how a Web-based weight loss intervention may be used for a high-risk population. The study sample captured a wide age range of 41 to 79 years and contained a range of employment status (unemployed, employed, retired). A limitation of the study sample is the lack of ethnic diversity: the sample was all white British men. South Asian and Black African groups are known to be twice as likely to develop type 2 diabetes and therefore a future study should aim to recruit ethnic minority groups to identify whether a Web-based intervention is acceptable to people of different ethnicities.

It is important to acknowledge the conflict between conducting a rigorous RCT and the need to keep up with the fast progression of technologies. Evaluation research faces the reality of falling behind commercial companies with the ability to regularly update their websites or apps. Large companies may have the advantage of greater financial stability and flexibility of funding and resources in contrast to academic research, where budgets can be extremely constricted and individual costs and resources tend to be outlined in advance of receiving funding. However, RCT methodology remains the most robust way of determining the effectiveness of an intervention [59]. One way to keep up-to-date with technology and maybe another potential improvement for the study is the use of a mobile phone app, in replacement of or in addition to a computer-based website. Mobile phones have now overtaken laptops or desktops as the devices used to access the Internet [60]. Although the study website was accessible via a mobile phone, the creation of an app may make the format easier to access on a mobile phone and potentially improve engagement and adherence. Over 12 months, each participant had 3 visits by the researcher at either their home or general practice, with data collection typically ranging between 20 minutes and 1 hour per visit. This level of face-to-face assessment with participants could be feasible in a main trial. However, the use of electronic scales to measure and transmit weight status to the research team, as implemented within the NULevel study [61], may be a more efficient and feasible method of data collection in a definitive trial.

Research is lacking with regard to implementing a Web-based weight loss intervention within the NHS. Given the high number of obese patients and NHS resources being increasingly stretched, services are struggling to provide sufficient referrals. Therefore, alternative modes of delivery, with the potential to reduce health professional input and time per patient while still enabling individual and tailored care, need to be investigated to identify if they can be effective and thus benefit the NHS. Although not powered to assess changes in outcomes, the descriptive statistics show positive indications of increased weight loss (in kilograms and as a percentage), reduced waist circumference, and decreased BMI for the intervention group from 3 to 12 months, in comparison with the control group. This research provides preliminary findings that recruitment of men with type 2 diabetes is possible within a Web-based intervention. Suggested improvements to the website were valuably gained from the parallel process evaluation and could be incorporated to potentially improve adherence and retention in future research.