The aim of this study was to assess the cost-effectiveness of e-learning devices (e-LDs) compared with conventional care (CC) in people with obesity. An economic evaluation based on discrete event simulation techniques was used to synthesise data from the published literature together with results from a systematic review and meta-analysis of RCT evidence on the impact of treatment on BMI.
The model results suggest that under most circumstances, e-LDs are unlikely to be cost-effective at any reasonable willingness-to-pay threshold. Mainly because of the fixed costs of installing the devices coupled with a negligible impact on BMI. Only when the fixed cost of the e-LD was removed, or substantially lowered, did e-learning devices appear to be cost-effective.
The fixed costs associated with the e-LDs were difficult to assess. Largely because most of the published clinical evaluations did not report resource use/cost data, they are heterogeneous in nature and it was unclear whether any were commercially available (and therefore had an associated user fee). For these reasons, the fixed cost of the e-LDs, was taken directly from the trial by McConnon - an internet-based intervention. The associated costs comprised of the development and running costs of the website over the 12 month study period. An economic evaluation was performed alongside the RCT, and it too concluded that the e-LD was unlikely to be cost-effective largely because of the fixed cost. However, it is unclear whether all e-LDs have an associated fixed cost. Second, assuming they do, fixed costs are likely to decline as the number of users increases meaning they could be substantially lower than the mean value assumed in this analysis. Third, even if the fixed cost was zero, it is unclear whether commercial programme developers would charge a fee for using the programme, how much it would be and how frequently it would be charged.
The choice of comparator programme(s) is an important design decision for any economic evaluation. Here the choice was ‘conventional care’ as delivered by health care professionals, as the control arms in the systematic review tended to use a mix of interventions. For example, most of the trials that focussed on dietary change were not particularly explicit as to what advice was given (eg. choice of diet) or who provided it (eg. a physician or nurse). This is important because it is possible that different approaches may be more or less cost-effective compared with each other, and (implicitly) averaging them as is the case here, could be misleading from an incremental perspective. Of equal importance, CC, however delivered, is not the only method of reducing weight. For example, it is possible to promote exercise, to use drug treatments such as orlistat or to use combinations of these approaches. Although the analysis containing orlistat was only crude in so much that it was not based on formal indirect treatment comparisons, it does illustrate the point that even if e-LDs were considered cost-effective compared with CC, they might not be compared with other treatments.
A number of economic evaluations of web-based interventions to promote weight loss have previously been published
[24, 45–47]. McConnon et al.
 concluded that it cost about £40,000 per additional QALY if an internet programme replaced usual care, while the two other studies reported ICERs nearer US$5,000 to $7,000 per life-year gained. While these results are different to ours, there are a number of possible explanations. First, our study focused purely on e-learning devices. The other studies were not so restrictive in terms of the intervention specifics. Indeed the interventions evaluated by Krukowski
 and Hersey
 included an element of interaction with other patients, individual/group coaching sessions and telephone support, they were not e-learning devices. Second, our estimate of intervention effect was based on the results from a systematic review of 43 RCTs whereas the studies by McConnon and Krukowski were based on the results from single trials. Third, the study by Krukowski assumed in the base case that weight loss could not be regained, an assumption we consider to be unrealistic. Lastly, our study assessed the cost-effectiveness in people who were already considered to be obese. This is was an explicit criterion in the studies by McConnon and Hersey, but not in the evaluation by Krukowski. Indeed, the latter was said to have been performed in individuals who were highly educated, but few other details are provided. Therefore, while we have concluded that in our opinion e-LDs to promote dietary change and weight loss are unlikely to be cost-effective in obese populations, this conclusion should not necessarily be generalised to all web-based weight management interventions or to all population groups
There are undoubtedly a number of other limitations with the evidence used in the model. First, the QRISK2 risk equation is designed to assess the probability of developing primary CVD-events. Thus, the model takes does not take into account the possibility that individuals who survive one CVD event are more likely to experience another. However, it is unlikely that this would have a major bearing on the results given the negligible treatment effects.
The CVD and T2D risk equations take into account risk factors such as systolic blood pressure and cholesterol levels in addition to BMI. While the trials rarely reported changes in these risk factors, a more sophisticated modelling approach could take into account their likely correlations.
Potentially counter intuitive results were produced in a number of scenarios. For example, when the costs of T2D and CVD were increased, the ICER associated with e-learning also increased. This is because people treated with e-learning devices live longer on average with these conditions, even though they are less likely to develop them in the first instance. The net result is an increase in the incremental cost and the associated ICER. Such seemingly counter intuitive results were also reported in the NICE Obesity Guideline, along with a similar explanation
The EVPI analysis suggested that the value of further research was arguably large, even with a 2 year horizon; e-based technologies are likely to have relatively short life-cycles. This is because despite the evidence of a relatively small clinical effect on BMI, the e-LDs are relatively cheap and the number of obese people is high compared with many other conditions