Work package 2.3

Modelling of long-term impacts on the incidence of type 2 diabetes and related complications, and health resource use

This WP used health economic modelling to estimate a justifiable cost for a weight loss maintenance intervention based on previous evidence of effectiveness. It also used findings from WP2.2 to estimate the likely cost-effectiveness of the SWiM intervention and the return on investment for a definitive RCT of cost-effectiveness.

Rationale

Pre-trial health economic modelling is recommended to estimate the likelihood that an intervention will be cost-effective before investing in a costly trial. Using an estimated intervention effect based on previous research,77 we used the SPHR Diabetes Model to estimate a maximum cost per person (justifiable cost) at which a weight loss maintenance intervention is likely to be cost-effective. Used as part of intervention development (WP2.1), this gave us a notional budget to ensure that we did not develop an intervention that was unlikely to be cost-effective.

Pre-trial modelling can also be used to design efficient trials that offer good value for money in terms of reducing uncertainty about the cost-effectiveness of an intervention. We used the SPHR Diabetes model57,58 (WP1.3) to estimate the cost-effectiveness of the SWiM intervention (WP2.1) based on data from the feasibility study (WP2.2) and an expert elicitation exercise. We conducted scenario analyses to inform the decision to proceed to a definitive RCT and the design of that trial.

Objectives

1. To estimate a justifiable cost for a weight loss maintenance programme conditional on a proposed end point.
2. To model the cost-effectiveness of our new weight loss maintenance programme and impact on incidence of T2D and related complications.
3. To inform the decision to proceed to a full-scale definitive trial and the design of that trial.
4. To use value of information methods to estimate the decision uncertainty associated with the intervention’s cost-effectiveness and determine the value of a definitive trial, based on ability to reduce decision uncertainty.

Using health economic modelling to inform a justifiable cost for the SWiM intervention

This work is reported in detail in Bates et al.77

Method

To estimate a likely effect size for our weight loss maintenance intervention, we meta-analysed the effect of previous studies identified from a recent systematic review.40 We then used the SPHR microsimulation model to estimate the long-term QALY gains and healthcare costs associated with this intervention effect in individuals with (1) a BMI of 28 kg/m² or above without diabetes and (2) a diagnosis of T2D and prescribed one non-insulin diabetes medication. Sensitivity analysis was conducted around the rate of regain, duration of effect and initial weight loss.

Key findings

The estimated intervention effect was a regain of 1.33 kg and 4.38 kg in years 1 and 2, respectively, compared to 2.84 kg and 5.6 kg in the control group, following a 9 kg initial weight loss. The justifiable cost for an ICER of £20,000 per QALY was £104.64 for an individual with a BMI ≥ 28 kg/m² and £88.14 for the population with T2D. Within sensitivity analysis, this varied from £36.42 to £203.77 for the former and between £29.98 and £173.05 for the latter.

An expected value of sample information analysis to assess the value for money of different randomised controlled trial designs, from a policy-maker prospective

This work is reported in full in Breeze et al.78

Method

In WP2.2, we estimated the mean difference in weight between the SWiM intervention and usual care after 6 months. However, this was a small feasibility study with high uncertainty and the long-term effectiveness beyond 6 months was unclear. We used the Sheffield ELicitation Framework to gather expert opinions (N = 4) on the distribution of uncertainty in the intervention’s effects on weight loss maintenance compared with usual care. Experts reviewed an evidence dossier based on the findings from the SWiM feasibility study and published data on similar interventions. We then guided the experts through an exercise to estimate probability distributions for the weight difference at 12 and 24 months. The effect of SWiM on HbA1c was assumed to be conditional on weight loss, with statistical analyses from WRAP (WP1.1) and GLoW (WP1.2) informing associations based on diabetes status. The cost of the SWiM intervention was estimated at £221, with no cost assigned to usual care.

We used the SPHR Diabetes microsimulation model to simulate 24 different trial designs, varying:

• sample size (500 or 2000 participants)
• follow-up duration (12 or 24 months)
• inclusion criteria (tier 2 behavioural weight management or specialist weight management or NHS Digital Weight Management or NHS DPP or NHS Pathway to Remission or all programmes combined).

For each trial design, we used expected value of sample information (EVSI) analyses to estimate the benefits of the trial: the reduction in the costs of an inefficient decision for future populations over 10 years. Trial designs are considered to offer value for money if the trial benefits exceed the trial costs.

Key findings

The expected ICERs are less than the threshold of £20,000 per QALY for SWiM compared with usual care across all weight management populations, but varies from £597 to £4940 per QALY gained (Figure 5). There is uncertainty around these estimates and the probability that the intervention is cost-effective ranges from 65% to 80%.

FIGURE 5 Estimated expected Net benefit of Sampling by trial design. Reproduced from Breeze et al.78 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure above includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

For three inclusion criteria options (groups recently completing ‘diabetes remission’, ‘digital weight management’ or ‘specialist weight management’), the cost of the proposed trials was estimated to exceed the estimated trial benefit over 10 years. For the other 3 inclusion criteria options (groups recently completed ‘behavioural weight management’, ‘diabetes prevention programme’, or ‘all programmes combined’), 12 trial designs produced greater benefits than costs.

The optimal trial design option would include ‘all weight loss programmes’, with 2-year follow-up and sample size n = 2000. Investment in a trial of this nature would give the greatest return on investment for the NHS because of the large number of people who could potentially benefit from the reduction in decision uncertainty.

Strengths and limitations

This EVSI analysis provides investigators and research funders with guidance on the trial design for a future evaluation of SWiM that offers best value to society. Expert elicitation exercises are very time-consuming and require substantial engagement from experts, with each elicited value requiring approximately 1–1.5 hours. We were only able to recruit four participants for the expert elicitation exercise because of participant availability and we conducted a 4-hour workshop to limit participant burden. A longer and larger workshop could have incorporated additional factors that could impact the probability distribution estimate (e.g. we could not elicit differences in intervention effects by baseline characteristics) and would improve confidence in these findings.

We needed to use approximation methods to estimate the EVSI in this study because of the complicated microsimulation model. Approximation methods for estimating the EVSI have been validated in previous studies and substantially reduce the computational burden of value of information analysis. This allowed us to evaluate 24 different trials designs.

Conclusions

There is good probability that the SWiM intervention would be cost-effective compared to standard care following a range of behavioural weight loss programmes. There is a strong likelihood that a trial of cost-effectiveness would result in net gains because of reductions in decision uncertainty. A large-scale trial with 2-year follow-up, encompassing participants recruited from various weight management programmes, was the most cost-effective trial design to evaluate the benefits of the SWiM intervention.