Introduction
Overweight and obesity is a risk factor for several negative health outcomes including cardiovascular disease (CVD), diabetes and cancer [
1]. Behavioural weight management programmes have been associated with significant weight loss [
2] and can even result in remission from type 2 diabetes [
3] but there is evidence that, on average, individuals regain weight loss by 5 years post-treatment [
4]. Furthermore based on a large observational study, only 21% of individuals are successful at maintaining weight loss, defined as losing at least 10% of their body weight and maintaining this weight loss for at least one year [
5]. While moderate reductions in weight have positive benefits for individuals who are overweight or obese and for those who have type 2 diabetes even if weight loss is regained [
6‐
8], weight loss maintenance is required to maintain full improvements in risk reduction. For example, individuals who lost 8–20% of their initial body weight and maintained this for 4 years (regained less than 3% of initial body weight) in a randomised control trial of a behavioural intervention achieved sustained improvements in blood glucose (HbA
1c), systolic blood pressure (SBP) and cholesterol, all biomarkers linked with health outcomes [
9]. Thus, there is a need to develop cost-effective weight loss maintenance interventions in order to prolong the positive impact of weight loss on health outcomes [
10].
Conducting pre-trial health economic modelling is recommended to estimate the likelihood of cost-effectiveness, inform decision about whether a trial is justified, and identify potential improvements to the intervention (9). Using an estimated intervention effect based on previous research, a maximum cost-per-person (justifiable cost) can be estimated at which the intervention would remain cost-effective given a certain incremental cost-effectiveness ratio (ICER). This can be compared to expected costs to ensure that an intervention is not predicted to incur a cost at which it is unlikely to be cost-effective. Pre-trial modelling has been conducted previously; for example Asaria et al. (2016) used a health economic model to estimate the annual costs at which interventions with varying impacts on cardiovascular risk would be cost-effective for individuals with different risk profiles [
11] and pre-trial modelling was used to inform the design of a fall-prevention intervention and trial [
12]. However, these studies were either based on hypothetical, rather than intervention-specific, risk changes (10) or based on the results from a pilot trial (11) and so is not a method that can be use before a pilot trial has taken place. The aim of this analysis was to use a health economic model to determine the justifiable cost of a behavioural weight loss maintenance intervention compared to no intervention in two populations; i) individuals with a Body Mass Index (BMI) of 28 kg/m
2 or above without diabetes and ii) individuals with a diagnosis of type 2 diabetes prescribed a single non-insulin diabetes medication.
Discussion
At an ICER of £20,000, the maximum justifiable cost was estimated to be £105 for individuals with a high BMI, £159 for individuals with a high BMI and a high HbA
1c (high risk of diabetes) and £88 for individuals with a diagnosis of type 2 diabetes on a single non-insulin medication. The finding that the maximum justifiable cost is lower on average for those with a diagnosis of diabetes than for those with a high BMI may seem counterintuitive given that those with a high BMI and at high risk of diabetes had the highest maximum justifiable cost. This is likely to be because, for individuals without type 2 diabetes, this intervention may be able to avert or delay a diagnosis of diabetes, which is associated with a reduction in the immediate costs associated with this diagnosis. This is particularly important for those with a high HbA
1c as the intervention averts or delays a potentially imminent diagnosis. Conversely, simulated individuals that have diabetes already have a higher associated cost than those without and less potential incremental gains; simulated individuals will have lower utility at the start and during the intervention period than patients with no diabetes and so the potential QALY gains are lower for patients with diabetes, and they cannot be ‘undiagnosed’ in the model. Although there is some evidence that remission from diabetes can be achieved [
3] which contradicts the model assumption that type 2 diabetes is irreversible, it is not yet clear that this remission is maintained. Overall, this indicates that the benefits of intervening in high-risk individuals (and therefore preventing or delaying diabetes) are higher than the benefits of intervening in people who already have diabetes.
In sensitivity analysis, duration of effect and the initial weight loss had the greatest impact on justifiable cost. The time it takes for participants to return to their original trajectory, if they do at all, is hard to determine due to short-term follow-up within trials [
4] and therefore a range of values should be considered when calculating a justifiable cost. There was a large difference between the scenario in which both the control and weight loss maintenance intervention had a duration of 4 years (£89) and the scenario in which the duration of the effect was 4 years for the control group and 6 years for the intervention group (£204) indicating the importance of the differential duration of effect between the control and intervention. The limited data on duration of weight management interventions indicates that intervention effect has diminished by an average of 5 years [
4] but there is little research available on the impact of a weight maintenance intervention in the long-term and this will vary depending on the characteristics of the intervention and the control group. Researchers should consider plausible durations of effect for the control and intervention groups based on the characteristics of the planned intervention (e.g., mode of delivery or duration). The outcomes of sensitivity analysis also indicated that a weight maintenance intervention is more likely to be cost-effective for individuals with a larger initial weight loss. Previous evidence does suggest that greater initial weight-loss is associated with weight maintenance [
57] supporting these findings.
Weight maintenance interventions that cost more than the maximum justifiable cost estimated are unlikely to be cost-effective based on the estimated intervention effect. While there is evidence that weight maintenance interventions are able to result in an additional 3.2 kg maintenance of weight loss over 18 months [
10], there is less evidence regarding the cost. In a weight loss maintenance trial for participants that had lost at least 5% of their body weight, intervention costs were between £16 and £49 depending on the amount of face-to-face contact but it was concluded that neither intervention was likely to be cost-effective in routine practice [
58]. Further evidence is required to determine the feasibility of developing an effective intervention within the justifiable costs estimated.
The method used in this analysis highlights the role that health economic modelling can have in the design and development of a new weight loss maintenance intervention. Although this type of modelling is recommended in intervention design guidance, there is little published research detailing the methods used to do this. While previous studies have used the results from a pilot trial [
12], the method presented here provides an estimate of justifiable cost without a pilot trial based on a range of previous studies; this can inform the design of the trial before a pilot trial. In addition, while pre-trial modelling has been used to identify the cost of an intervention that achieves a certain risk reduction [
11], the estimated impacts were not specific to a planned intervention which may limit application to certain interventions. The maximum justifiable cost provides an estimated upper bound over which the intervention would not be cost-effective, which can be compared to the predicted cost of the planned interventions. This could help to avoid an intervention which is unlikely to be cost-effective proceeding to the trial stage. Subgroup and sensitivity analysis can also inform decisions about whom the intervention should be targeted at and what factors are most likely to impact on cost-effectiveness. Although the current study is specific to a weight management intervention in the UK the methods can be applied to behavioural interventions in other health areas and countries. The increased number of public health economic models being developed [
59] will facilitate this type of modelling. However, as with many public health interventions, there is likely to be a large amount of heterogeneity in effect within the patient groups and therefore there may be limited application when using the mean effect only. Additional research into the different factors that impact on the intervention effect would be informative for this type of pre-trial modelling.
There were some limitations of this analysis. Firstly, due to limited research on the impact of weight loss maintenance intervention and, in particular, the impact of weight loss maintenance interventions for people with type 2 diabetes, the same weight loss and regain was applied for each person and in both populations, despite some evidence of heterogeneity in weight trajectories [
4,
57], and some differences between the baseline populations on risk factors such as systolic blood pressure. In addition, the estimate of weight regain at 24 months was based on only two intervention arms and so caution should be exercised in interpreting this result. Given the potential impact of differing weight trajectories, we conducted a range of sensitivity analysis to estimate the impact of alternate trajectories [
60]. Second, individual participant data was not used for the baseline population for individuals with diabetes due to limitations on availability of data. This may limit how representative this baseline population is of a population with diabetes. Furthermore, the population was selected because they were on a single diabetes medication, but this does not rule out having been on more than one medication in the past and so the population may have been more heterogenous than the potential target population for the intervention. However, the population was generated based on many variables and based on a large dataset that is representative to the UK. Third, remission from diabetes is currently not a scenario in the model. There is some evidence that remission from diabetes (an HbA1c of below 6% and no requirement for antidiabetic medication) can be achieved by following a low-calorie diet for 3–5 months, with stepped re-introduction to food and ongoing weight loss maintenance support [
3]. Given that those eligible for a weight loss maintenance intervention have already been successful in weight loss, in this study approximately 9 kg, there is a possibility that some individuals would go into remission. This means that the model may underestimate the positive impact of the intervention for those with diabetes as the cost-reduction associated with potential diabetes remission wasn’t captured. However, it is not yet clear that this remission is maintained and it’s likely that these patients will be required to attend regular screenings due to their previous diagnosis and so associated costs will still apply. Ongoing research will provide more information about the long-term impact of diabetes remission on costs and QALYs [
61]. Finally, as the healthcare perspective was used, the costs incurred by patients as a result of a change in lifestyle are not considered. These costs may differentially impact subgroups, and this is not accounted for in the analysis.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.