Disease-specific reference case extension: management of overweight and obesity in adults

Population

Sections 1.1 to 1.7

Modelling needs to reflect that people with specific comorbidities will have different levels of risk of health events and capacity to benefit from treatments for overweight and obesity and will be on different care pathways.

Type 2 diabetes mellitus (T2DM) and atherosclerotic cardiovascular disease (ASCVD) were identified as key comorbidities for people living with overweight and obesity. Therefore, it was agreed that results should be stratified according to whether people have these conditions. This also reflects how treatments are recommended in NICE's guideline on type 2 diabetes in adults and NICE's technology appraisal guidance on tirzepatide for treating type 2 diabetes and semaglutide for reducing the risk of major adverse cardiovascular events in people with cardiovascular disease and overweight or obesity.

Further subgrouping of the obesity strata by body mass index (BMI) category was considered appropriate by clinical experts because:

Clinical experts and patient advocates debated using BMI compared to waist circumference for stratifying the population of interest. They highlighted the importance of waist circumference measurement and its use in waist-to-height ratio calculations, particularly for people with a BMI of less than 35 kg/m² as recommended in NICE's guideline on overweight and obesity management, but noted that it is unlikely to be useful for modelling purposes as currently there are no risk equations that use this measure to estimate risk of developing associated comorbidities.

Stratifying obesity in adults into 3 groups as outlined in the section on classifying obesity in adults in NICE's guideline on overweight and obesity management is desirable. However, that would have increased the total number of strata from 8 to 16. To keep it manageable, only stratification between living with overweight and living with obesity is 'required'. Subgrouping the obesity strata further, although desirable, might not be essential for all cost-utility analyses and so this has been rated as 'recommended'.

Ethnicity-specific BMI thresholds were included to reflect evidence reported in NICE's guideline on overweight and obesity management that certain ethnic populations (for example, South Asian, Chinese, Middle Eastern, Other Asian, Black African and African-Caribbean) have a higher cardiometabolic risk at lower BMI levels than white populations. When individual patient data is available, it was agreed that subgrouping should be done by ethnicity, taking into account the ethnic-specific BMI thresholds. The healthy weight category does not need be included as a baseline weight category, but it is relevant for later modelling.

Stratification by both BMI category and comorbidity allows the impact of interventions on cardiovascular events and T2DM to be captured.

The number and type of comorbidities in the population are important as they influence baseline risk of morbidity and mortality. Treatment may result in a greater reduction in risk for certain subgroups. Sensitivity analyses that explore the cost-effectiveness of interventions in different population subgroups defined by both type and number of comorbidities will help identify the populations that will gain the greatest benefit.

Populations without T2DM or ASCVD should be grouped using other comorbidities such as dyslipidaemia, hypertension, metabolic dysfunction-associated steatotic liver disease (MASLD), obstructive sleep apnoea and non-diabetic hyperglycaemia (in line with NICE's technology appraisal guidance on semaglutide and tirzepatide for managing overweight and obesity), and metabolic syndrome. However, it was considered these might have less power to predict health outcomes within the 6 population strata that have T2DM or ASCVD or T2DM and ASCVD, where the cost, quality of life and prognosis will already be poorer, and so subgrouping using these additional comorbidities was not recommended.

Chronic kidney disease (CKD) and chronic heart failure (CHF) have been included as subgroups for the population strata containing people with T2DM or ASCVD or T2DM and ASCVD. This was to align with modelling undertaken for the medicines update of NICE's guideline on type 2 diabetes in adults and to reflect the increased prevalence of these conditions in these populations (Dawson et al. 2023, Koye et al. 2018, Lee et al. 2024, Panchal et al. 2024).

It is acceptable to only include the strata or stratum that reflect the eligible target population for the intervention of interest. For example, if an intervention is targeting people with a BMI of 40 kg/m² or more and ASCVD, then only this stratum should be presented.

There are various causes of genomic and syndromic obesity, each of which has unique features and treatment pathways that will not be adequately covered by this reference case extension. These include Prader-Willi Syndrome, Bardet-Biedl Syndrome and Alström Syndrome. A broad range of population groups may be affected by obesity. However, as outlined in NICE's position statement, the primary purpose of disease‑specific reference case extensions is to promote consistency in health economic modelling across NICE assessments. To achieve this, reference case extensions are expected to focus on common diseases and conditions where multiple pieces of NICE guidance typically exist, and where standardisation can therefore provide the greatest value.

Populations in trials of interventions for obesity may not reflect the full eligible population in England. Therefore, baseline patient characteristics, such as age, sex and weight, for each strata should be estimated from real-world evidence representative of the NHS population in England. A sensitivity analysis should be undertaken when patient characteristics are estimated from trial populations, to assess generalisability of the trial results to a real-world population whose condition is managed in the NHS in England.

Intervention and comparators

Sections 2.1 to 2.10

To ensure consistency and relevance across cost-utility analyses, relevant comparators that are established practice in the NHS for managing weight, including those already recommended by NICE, should be included. These comparators should be relevant to the stratum being evaluated. These comparators represent the spectrum of treatment intensity and allow for meaningful comparisons of cost-effectiveness.

The inclusion of bariatric procedures as a comparator reflects NICE's guideline on overweight and obesity management and acknowledges the clinical and cost-effectiveness of surgical interventions for people with obesity. This has been included as a 'recommended' rather than 'required' statement as the strength of NICE's recommendations on assessment for bariatric procedures in its guidance vary, depending on the body mass index (BMI) of the population. There are several bariatric procedures available in the NHS. These vary in terms of their invasiveness, reversibility, effectiveness, complications and costs. As well as surgical procedures, there are also minimally invasive endoscopic bariatric interventions (NICE's HealthTech guidance on endoscopic sleeve gastroplasty for obesity). The comparator procedure should be clearly defined in the model to ensure appropriate effectiveness data and costs are applied. Justification for the choice of bariatric procedure should be provided and should reflect established NHS practice. Currently the most common bariatric procedures in NHS practice are sleeve gastrectomy and Roux-en-Y pass (National Obesity Audit's bariatric surgical procedure dashboard). The model should cover the provision of other interventions while people are waiting to receive bariatric procedures.

In line with clinical practice and NICE guidelines, if an intervention is indicated alongside a behavioural intervention, then this should be captured in the model. These interventions may be delivered in person or digitally. However, the intensity and structure of behavioural interventions can vary significantly, making it challenging to achieve consistency between models. To address this, different intensities of behavioural interventions should be explored in sensitivity analyses, with each behavioural intervention clearly described and costed. This should be done separately for the concomitant and standalone behavioural interventions.

When comparing an intervention to behavioural interventions alone, sensitivity analyses should include the cost of the minimal behavioural intervention of GP advice once a year to reflect the limited availability of services for overweight and obesity in the NHS. Such a sensitivity analysis only explores the lower cost of a minimal intervention of annual GP advice and does not explore any potential treatment effect. This is a conservative analysis that aims to provide an upper estimate of the incremental cost effectiveness ratio (ICER).

For strata that include people with type 2 diabetes mellitus (T2DM) or atherosclerotic cardiovascular disease (ASCVD), standard background treatment for those conditions should be included and costed, ensuring that models reflect realistic treatment pathways and avoid underestimating costs or overestimating incremental benefits. In the base case these background treatments should reflect those used in the NHS.

Initial treatment for T2DM included modified-release metformin and an sodium glucose transport 2 inhibitor (SGLT-2 inhibitor), except where contraindicated (see NICE's guideline on type 2 diabetes in adults). For people with ASCVD, background treatment should include lipid-lowering treatment and other relevant treatments. However, background treatment in trials may not reflect established clinical practice in the NHS, and sensitivity analysis should reflect background treatments used in the corresponding trial evidence to ensure they are aligned with the effectiveness data.

Model structure and health states

Sections 3.1 to 3.8

A state transition modelling approach, either cohort-based or individual patient-level simulation (IPS), was considered suitable by health economic modellers.

State transition models are well-suited to chronic disease modelling and have been widely used in obesity and related comorbidities (see appendix B). They offer relative transparency to more complex modelling approaches, ease of use and can capture disease progression over time. State transition models can be used for a wide range of intervention types for managing obesity, such as medicines, surgery and behavioural interventions, and allow for isolation of specific benefits when needed to support transparent decision-making.

NICE considers state transition models to be appropriate for modelling this condition. However, discrete event simulations offer some advantages and so are potentially acceptable if the model is transparently presented and can be easily interrogated.

IPS models allow for tracking of individual patient histories, which is particularly important in obesity because events such as treatment discontinuation, bariatric procedures or cardiovascular events can occur at varying times and influence future risks. IPS models can better capture dependencies between events, which are difficult to represent in memoryless cohort models. If a cohort model is used, tunnel states should be added to address these limitations.

Obesity is associated with a wide range of health conditions, some causally linked and some potentially contributing to obesity itself, such as polycystic ovary syndrome. Certain comorbidities, such as type 2 diabetes mellitus (T2DM) and atherosclerotic cardiovascular disease (ASCVD), especially stroke and myocardial infarction (MI), can significantly alter treatment pathways and long-term outcomes. Therefore, economic models should include a carefully selected set of comorbidities and health events that reflect the burden of obesity. The inclusion of comorbidities and health events should be guided by the following criteria: the need for a strong, evidence-based association with obesity, that they have a substantial impact on costs or quality of life, and their inclusion in a high-quality core outcome set (COS). The association should be identified by measuring the change in the outcome relative to a reduction in weight or body mass index (BMI) – see, for example, Khunti et al. 2023.

When there are several core outcomes sets (COSs), the quality of the COS development process can be assessed using the Core Outcome Set-STAndards for Development (COS-STAD) criteria (Kirkham et al. 2017). A recent report in the COMET database (Gorst et al. 2025) concluded that the COS developed by Coulman et al. (2016) was methodologically sound, meeting 10 of the 12 COS-STAD criteria, and with a good number of participants with lived experience throughout the process. Therefore, all of the outcomes in that COS are included in this reference case extension.

A minimum set of health states to be included in economic model have been provided. Only events or outcomes that have been directly measured or use validated risk equations (see rationale and supporting information section on treatment effects for details on using validated risk equations) should be included. These should reflect the most common and impactful obesity-related comorbidities.

Chronic kidney disease (CKD) was included to ensure consistency with the modelling undertaken for the medicines update of NICE's guideline on type 2 diabetes in adults. The link between obesity and CKD is well established, and there is some evidence that incretin agonists may improve renal outcomes via both direct effects (including glucose lowering in those with diabetes and other tissue effects) and indirect effects (via weight loss). However, the magnitude and mechanism of effect remain uncertain, and ongoing trials such as REMODEL (Cherney et al. 2025) or SURMOUNT-MMO (Lam et al. 2025) may clarify this.

Transitions in the model should reflect realistic disease trajectories to ensure clinical credibility and avoid overestimating intervention benefits. They are essential because they reflect the natural history of obesity and its complications. The included transitions are supported by robust evidence and are key drivers of long-term costs and health outcomes.

Other events, while not needing to be captured as health states in the economic model, have significant cost and utility implications and so should be included as health events in the model. These were identified by reviewing previous obesity health economic models (as outlined in appendix B) and after discussions with clinical experts and patient advocates. The following health events should be included for the reasons given:

The inclusion of these events, as well as discontinuation of medication, should help ensure that models capture the full burden of obesity and the real-world consequences of treatment. They are all 'recommended' for inclusion, except for treatment and procedure-related adverse effects, which is 'required' because, unlike the other events, it is relevant for all assessments.

Some health events are recommended for inclusion in sensitivity analyses if there is emerging or partial evidence of a treatment effect. These include:

These events were not included in the minimum health state set because of insufficient direct evidence of treatment effect, uncertainty in causal pathways and potential for modelling complexity or bias. This flexible approach ensures that models remain evidence-based, while allowing opportunity to incorporate new data as it becomes available.

Several additional obesity-related health outcomes such as fertility, depression, chronic lower back pain and inflammatory conditions were discussed at workshops involving clinical experts, patient advocates, health economic modellers, industry stakeholders and commissioners. They were not included in the minimum health states or events because there was insufficient direct evidence of treatment effect and concerns with the potential for modelling complexity and double counting of impact on quality of life. It was agreed that these additional health outcomes could be captured qualitatively. However, if new direct evidence of treatment effect emerges to warrant their inclusion, then they can be considered for inclusion within future versions of this reference case extension.

It was noted that for populations with MASLD at baseline, consideration of additional health states may be necessary. Some guidance is available specifically for metabolic dysfunction-associated steatohepatitis (MASH) – see NICE's health technology assessment innovation laboratory report on evaluating MASH treatments.

Procedure-related adverse events that are important to patients include (Coulman et al. 2016):

Clinical parameters

Measuring and valuing health effects

Sections 5.1 to 5.5

Models submitted in NICE's technology appraisal guidance on liraglutide, semaglutide and tirzepatide for managing overweight and obesity used baseline utility based on age, body mass index (BMI) and sex, with utility decrements applied to specific events.

A study by Luah et al. in 2024 estimated the association between BMI and EQ-5D-5L among the general population in England using data from 2017 and 2018 health surveys. It distinguished utility values by sex and BMI level. It also derived the coefficients for comorbidities, including for diabetes, heart and circulatory disease, respiratory disease, musculoskeletal disease, cancer and mental health disorders. This may be a suitable study to inform the baseline utility values and utility decrements associated with comorbidities in a consistent manner.

EQ-5D data from patients in relevant clinical trials should be used to calibrate the quality-of-life treatment effect in the short-term (at trial follow-up), as direct evidence of quality-of-life improvement is best when it is precise and of high quality.

Treatment-related adverse effects and complications from bariatric procedures should be estimated by weighting by incidence and effect duration, using an evidence-based estimate of duration.

The most common treatment-related adverse effects reported for incretin agonists were gastrointestinal. Clinical experts at the workshop noted this was a key reason for dose titration of medicines. Such adverse effects occur early and are unlikely to be ongoing because those experiencing problems after minimising the dose will stop taking the medicine.

In the case of bariatric procedures, the model should reflect the fact that quality of life will be reduced while patients are recovering from their procedure. In the case of minimally-invasive procedures, this time is likely to be less than for surgical procedures but will depend on complication rates.

Utility decrements for treatment-related adverse effects were applied additively in base cases, and a multiplicative approach has been explored in scenario analyses in previously published technology appraisal guidance (see appendix B for further details). Health economic modellers noted that the most appropriate approach to applying health state utility estimates depends on whether or not the health effects are seen to be independent.

If the analysis by Luah et al. (2024) is used to estimate health state utility values, an additive approach is most appropriate as the utility decrement values in the study were estimated with a linear regression.

If other sources are identified for health state utilities and the utility decrements associated with type 2 diabetes mellitus (T2DM), obesity and any other comorbidities show no significant interactions, then an additive approach also seems reasonable, otherwise a multiplicative approach as discussed in DSU Technical Support Document 12 should be used. This document also outlines methods for adjusting utility values when calibration is required, such as for quality-of-life effect in the short-term (at trial follow-up), and for avoiding double counting improvements from the reduced incidence of T2DM progression or adverse effects.

Obesity is associated with mental illness. This can worsen when lost weight is regained, and so strategies that produce sustainable weight loss will be most effective at improving rates of mental illness associated with obesity. This was an important health outcome highlighted by clinical experts and patient advocates. Impact on mental health should generally be captured by the EQ-5D if it is used in trials or if weight-related EQ-5D estimates are used in the model, although it may not capture elements around the social stigma of obesity. For strategies where there is weight regain, the subsequent loss of utility will not be fully captured by simple weight-related utilities. Therefore, we recommend sensitivity analysis for weight regain.

Cost and healthcare resource use

Sections 6.1 to 6.9

Interventions for managing weight require assessment for eligibility and can be associated with complications, for example, adverse effects for medicines or re-operation or re-intervention for procedures. The costs associated with both assessment and managing complications should therefore be captured in the model.

For medicines, supply chain costs may include dispensing, refrigeration, waste disposal and delivery costs. For bariatric procedures, capital investment costs, including infrastructure costs, should be captured (see NICE technology appraisal and highly specialised technologies guidance: the manual and NICE HealthTech programme manual). Costing should consider device reuse parameters, organisational impact (such as training requirements or investments in infrastructure) and learning-curve effects, when relevant, consistent with Drummond et al. (2018).

As noted by the committee for NICE's technology appraisal guidance on tirzepatide for managing overweight and obesity, there is currently uncertainty about the resource use needed for behavioural interventions (related to encouraging a healthy diet and physical activity), both as comparators and concomitant treatments with other interventions. NHS England's interim commissioning guidance on implementation of NICE's technology appraisal guidance on tirzepatide states that wraparound care for tirzepatide should incorporate appropriate nutritional and dietetic advice, physical activity guidance and behavioural change components, over a minimum timeframe of 9 months from the point of prescribing. The importance of exploring the resource use for these behavioural interventions in sensitivity analyses was highlighted during the workshops. If real-world evidence is available about implementing tirzepatide (for example, data collected as part of NHS England's interim commissioning guidance on implementation of NICE's technology appraisal guidance on tirzepatide) in terms of the resource use required for concomitant behavioural interventions, then this should be used.

For the cost of comorbidities and acute events, adjustments should be made to ensure that these costs are not double counted. For example, the cost of treating obesity should be subtracted from the cost of treating type 2 diabetes mellitus (T2DM) or atherosclerotic cardiovascular disease (ASCVD), where relevant. When multiple comorbidities develop sequentially, ideally costs should be measured incrementally from the same data set. However, as a minimum, acute event care costs should be greater than costs associated with chronic states. Clinical experts emphasised the importance of capturing the impact of weight management on mental health and therefore there is a requirement to include the cost of mental healthcare services, as the impact on mental health itself should already be captured by the EQ-5D.

The cost of treating comorbidities often increases over time as more additional comorbidities develop. With T2DM, for example, average resource use is likely to underestimate costs for advanced disease and overestimate costs for incident cases. Workshop experts and the committee for NICE's technology appraisal guidance on tirzepatide for managing overweight and obesity advise that attempts should be made to distinguish the costs of early and later disease.

For people with T2DM, management and complication costs from cohort studies are recommended. For example, the committee for NICE's technology appraisal guidance on tirzepatide for managing overweight and obesity preferred using UK Prospective Diabetes Study (UKDS) costs, which include inpatient and non-inpatient healthcare costs as a function of T2DM-related complications, age and sex, rather than the company's initial approach of using diabetes-related NHS reference costs and assuming 1 hospital attendance or stay per patient per year. The cost of medicines is not included in the UKPDS costing and therefore would need to be costed separately.

Equality and other considerations

Sections 7.1 to 7.3

Obesity is more prevalent in the most socioeconomically deprived quintiles of the population (see NICE's health inequalities briefing report for its guideline on overweight and obesity management and its health inequalities report for its medicines update of its guideline on type 2 diabetes in adults).

Treatments for obesity can potentially help reduce the differences in health outcomes between different socio-economic groups if uptake is particularly supported in those groups.

Distributional cost-effectiveness analysis (DCEA) can show the potential impact of a weight management intervention on health inequalities, which could support the case for recommending the intervention (see the section on distributional cost-effectiveness analysis methods in NICE's health technology evaluations manual). However, the number of adults with obesity in England is very large and so committees should be aware that the opportunity cost associated with recommending an intervention with a slightly higher incremental cost effectiveness ratio (ICER) could be substantial. DCEA can also be used to indicate when additional recommendations or targeted implementation support might improve take-up of treatments in deprived areas and strengthen the equity impact.

There is some evidence that take-up of, and adherence to, weight management treatments might be lower in more deprived groups. This should be captured in DCEAs – see NICE's inequality analysis of health outcomes of different diets in achieving and maintaining weight loss for NICE's guideline on overweight and obesity management (January 2025).

The methods for DCEA for aspects of health inequality other than socioeconomic deprivation are less well developed. Therefore, these should be considered outside of the model.

For pragmatic reasons, models inevitably do not cover every effect. A few elements that are unlikely to be captured in economic models were identified as important by clinical experts, patient advocates, health economic modellers, industry stakeholders and commissioners. These have been highlighted for qualitative consideration outside of the model.

Health inequalities for people living with disability and for specific ethnic groups have been highlighted based on the NICE's health inequalities briefing report for its guideline on overweight and obesity management and the advice of workshop experts.

Weight management is a key consideration in provision of certain treatments. Two treatments where this is the case were identified in the expert workshops: fertility treatment (Boddeti et al. 2025, Pandey 2010) and transplant surgery (Ghanem et al. 2024). Others include dialysis, atrial fibrillation ablation and surgery more generally. It should be noted that if weight loss enables other treatments to go ahead, then it also incurs the associated costs of these treatments, which can be considerable, for dialysis, in particular.