The intention to treat analyses indicated that at 12 months follow-up the usual care, web-based only and regular nurse support groups had lost very similar amounts of weight (2.30 kg to 2.50 kg), whereas the website group with basic nurse support group had lost more weight (4.31 kg). However, one of the most important learning points from this feasibility trial was that practices had interpreted ‘usual care’ differently; two practices had offered patients allocated to usual care an in-practice programme of regular nurse support for weight management (even though this was not routinely offered to obese patients on their practice list), whereas three had followed our planned protocol of providing patients only with their usual medical care. This makes it hard to interpret the comparison with ‘usual care’ in the whole sample, since in three practices we were comparing the POWeR intervention with a no treatment control (as planned), whereas in the two practices that deviated from protocol we were comparing the POWeR intervention with in-person weight management support.
It is more straightforward to interpret the pattern of findings in the three practices that followed protocol. In these practices, at 6 months there was a steady increase in the amount of weight lost with added levels of support; compared with the usual care group the web-based only group lost an additional 0.39 kg on average, the website plus basic nurse support group lost an additional 1.6 kg and the website plus regular nurse support group lost an additional 2.8 kg. Interestingly, at 12 months the web-based only group and website with regular nurse support groups had not sustained these additional gains in weight loss, whereas the website with basic nurse support group had slightly increased their average weight loss to 4.64 kg (compared with 1.71 kg in the usual care group). This pattern of findings was also reflected in the proportion with substantial weight loss (i.e. ≥5% of initial bodyweight); 33.3% of those in the website with basic nurse support group achieved this compared with only 16% of those in the usual care group. Patients who adhered to the intervention (42% of those allocated to one of the website arms) had lost an average of 6.70 kg at 12 months.
These findings are of course only indicative, and require confirmation in our forthcoming fully powered trial, but they do suggest that the combination of the web-based intervention with basic nurse support could provide an effective solution to weight management support in a primary care context. The weight loss achieved in this group was better than the average weight loss achieved by web-based weight management programmes  and was very similar to the best of the six face-to-face programmes compared in a UK primary context . Moreover, our sample, although somewhat under-representing men and those from ethnic minorities, was not young or highly educated and can be considered broadly typical of the primary care population eligible to take up this intervention if implemented . However, a significant proportion of patients failed to engage successfully with our web-based intervention, and it is likely that a range of weight management options will be needed for the primary care population .
An intriguing finding was that basic nurse support (3 sessions in the first 3 months) actually resulted in better outcomes at 12 months than the more regular support (7 sessions in the first 6 months). This finding was not anticipated, and clearly requires replication as it could simply be due to chance, but it is not inexplicable. Nearly two-thirds of patients in the regular nurse support group attended for less than half the available sessions; it is probable that this group included some patients who were unwilling or unable to access the nurse support, and some may have avoided meeting the nurse if they were failing to lose weight or had lost motivation to adhere to POWeR. However one third of patients in this group did receive regular support for six months. There are indications from the literature that some people given regular support may become dependent on it and find it difficult to maintain motivation independently when it is withdrawn , which could explain why the regular nurse support group lost most weight during the period that they had support but then regained it after the support ended. The basic nurse support schedule was timed to provide support during the period when autonomous motivation declines most sharply , but may have been too limited to foster dependence, or to become onerous. Another finding of interest was that nearly one in three people chose a low carbohydrate rather than a low calorie eating plan, and outcomes for these two groups were very similar.
This feasibility trial identified some important issues relevant to optimising both the intervention and trial procedures. With regard to adherence to the intervention, patient completion of online sessions appeared comparable to most web-based weight management interventions, in which attrition is typically high , and importantly was better than the attrition rates typically observed in interventions with no health professional contact [10, 40, 41]. Since session completion rates were very similar across all intervention groups, including the web-based only group, this suggests that even the basic contact with primary care involved in nurse recruitment and weight monitoring for the trial at 6 and 12 months may have been sufficient to reduce patient attrition. Findings from a qualitative process study carried out in a sample of participants in this trial support this conclusion (unpublished observation; paper submitted), as several patients in the web-based only group commented that they were motivated by being monitored by their practice nurse. However, implementation of nurse support was lower than intended, with around half the planned support sessions delivered, and this may have diluted the intervention effect. Slightly better outcomes than ours were achieved by a web-delivered intervention when trained coaches provided intensive phone support and succeeded in delivering over 90% of planned support sessions . Support for the POWeR intervention was delivered by practice nurses with a standard clinical workload rather than by dedicated trial staff, hence these follow up rates are likely to represent real world implementation of the intervention; nevertheless, better procedures for maximising delivery of nurse support might improve outcomes in the main trial.
With regard to optimising trial procedures, the main lesson learned was that it is necessary to standardise the intervention to be offered to the control group, rather than relying on practices to continue their ‘usual care’ of recruited patients. Prior to the start of the trial we had established that practices did not routinely offer structured weight loss programmes to their obese patients, but having observed the patient demand for POWeR some practice nurses felt compelled to offer something to patients in the usual care group, and so either improved or developed programmes, drawing partly on their experiences of supporting POWeR. This finding highlights the risk of contamination of the control group in a trial - but also suggests that introducing a programme such as POWeR could raise awareness of the need to offer a range of interventions for helping obese patients. This feasibility trial also highlighted the need to introduce follow-up methods that were more acceptable to patients; while dropout from trials, like intervention attrition, is often high in web-based weight management interventions, some trials have achieved better rates of follow-up than ours. The changes we implemented to our methods of follow-up between the 6 month and 12 month follow-up (as described in the Method) did appear to improve follow-up rates, although we had already permanently lost some patients from the trial at 6 months. This suggests that by allowing patients more flexibility in how and where they are followed up it may be possible to achieve lower rates of dropout.