Metreleptin for treating lipodystrophy

The patient experts explained the all-consuming nature of lipodystrophy. They highlighted that, other than the serious metabolic abnormalities caused by the condition, hyperphagia was a predominant debilitating feature. The company stated that this feeling of constant hunger was better described as starvation to convey the extent of its debilitating effects. The patient experts explained that eating does not relieve the hunger, so people with the condition are constantly looking for food, which results in physical, psychological and behavioural complications. For children, constant supervision is needed to ensure they do not eat inedible objects. The constant food seeking, and associated lack of concentration and fatigue, negatively affects social and professional life, and is a significant financial burden. The patient experts highlighted that, in the absence of specific treatment targeting lipodystrophy and hyperphagia, dietary advice is a mainstay of supportive treatment. They noted their frustration with this because dietary control is often impossible when overcome by a feeling of starvation. The patient experts also noted that caring for a child with lipodystrophy is a substantial burden, affecting parents and carers both emotionally and financially. The committee acknowledged that lipodystrophy is a debilitating condition, and that hyperphagia has a significant effect on quality of life, which affects patients, parents and carers. For children and young people with congenital generalised lipodystrophy, the condition may also shorten their life expectancy. The committee recognised that there is a significant unmet need for an effective treatment option.

The clinical experts explained that lipodystrophy diagnosis may be delayed because it is not immediately recognised and is a rare condition. Diagnosing generalised lipodystrophy, particularly when congenital, is easier because people typically present at between 1 to 2 years old, and develop diabetes and damage in 1 or more organs by the time they are 2 or 3 years old. However, partial lipodystrophy usually presents later, and symptoms are heterogeneous, which makes diagnosis at an early stage of the disease difficult. The clinical experts highlighted that an earlier diagnosis is important to prevent disease progression. The patient experts supported the view of the clinical experts, stating that partial lipodystrophy can progress and become severe if undiagnosed and untreated. The company explained that some people may have a mutation that is unrelated to leptin deficiency but manifests with similar symptoms, emphasising that the right clinical diagnosis is critical for effective treatment. The clinical experts confirmed that patients are routinely genotyped as part of the NHS service at Addenbrooke's Hospital. The committee was satisfied that people with lipodystrophy can be accurately identified, but noted that diagnosis in some people with milder forms of partial lipodystrophy may be delayed.

The committee discussed the clinical evidence submitted by the company:

• NIH 991265 was a pilot, dose-escalation study to determine the safety and efficacy of short-term leptin replacement (up to 8 months). After NIH 991265 ended, patients continued treatment in the extension study NIH 20010769 for long-term follow up. All but 1 patient who completed NIH 991265 moved to NIH 20010769. This meant the studies were treated as 1 study continuously enrolling patients with generalised (n=66) and partial lipodystrophy (n=41) over the course of 14 years.

• FHA101 was an open-label, single-arm, expanded-access study with 9 patients with generalised lipodystrophy and 32 patients with partial lipodystrophy enrolled over the course of 6 years.
  
  Only 1 patient in these studies was recruited from the UK, but the clinical experts confirmed that the trial populations were generalisable to patients seen in clinical practice in England. The ERG highlighted that estimates of treatment effects were based on changes from baseline in single-arm metreleptin treatment studies during the first and second committee meetings. The committee noted that there was a lack of evidence on metreleptin's relative treatment effect. In its resubmission, and in response to committee's concerns at the second meeting (see section 4.2), the company updated its systematic literature review. It used a new search strategy to include search terms for comparators and did a new indirect treatment comparison using 3 methods. The ERG noted that the new systematic literature review addressed its previous critiques on searching. The committee concluded that it was satisfied with the company's attempt to identify evidence on comparators in its updated systematic literature review.

One of the committee's concerns was the lack of evidence on metreleptin's long-term effect in people who have treatment (see section 4.2). In its resubmission, the company provided additional clinical evidence with 36‑month follow up from the metreleptin early access programme (n=31, including all patients who had treatment since the start of the programme) at Addenbrooke's Hospital. Data were collected retrospectively, with clinicians entering patient data on baseline characteristics, organ damage and complications, laboratory values and metreleptin dose. The ERG noted that the additional data showed persistence of changes with metreleptin treatment up to 36 months, with a reduction in HbA1c and triglycerides (see section 4.11). But, there were no patient-reported outcomes or measures of patient experience (including hyperphagia) presented, despite some people having treatment for 10 years or more. The company explained that when the early access programme was set up in 2005 there was no pre-specified data collection protocol, so the only data available were from a review of patient medical records. In addition, clinical experts noted that the treatment eligibility criteria currently used were not in place when the programme was set up. Because the criteria were only introduced 2 years ago, not all people with partial lipodystrophy in the programme had poor enough metabolic status at baseline to be eligible for the treatment under the current criteria. For example, people with high levels of triglycerides and a low level of HbA1c were enrolled at the time. The clinical experts also stated that the programme had a small sample size, and there were very few data on generalised lipodystrophy's response to metreleptin. This meant that no conclusion on efficacy could be made from the early access programme. The committee welcomed the company's attempt to gather additional clinical evidence to supplement the sparse evidence base. It agreed that the new evidence indicated that metreleptin treatment improves metabolic outcomes at 36‑month follow up. But, the committee did not substantially improve its understanding of the relative effectiveness of metreleptin. It concluded that evidence from the early access programme itself was not sufficient to address the gaps in the clinical evidence base.

In response to the committee's concerns about the lack of evidence on metreleptin's relative effectiveness (see section 4.2), the company did an indirect treatment comparison. This used 3 methods (inverse probability weighting, multivariate regression analysis, and naive comparison) to estimate the treatment effect of metreleptin relative to supportive care for key clinical outcomes. Data used to inform the indirect treatment comparison were from the NIH 991265/NIH 20010769 study (metreleptin arm) and the GL/PL natural history study (supportive care arm), respectively. Outcomes assessed included change from baseline in HbA1c, triglycerides and liver enzyme (alanine aminotransferase [ALT] and aspartate aminotransferase [AST]) levels, the incidence of acute pancreatitis and all-cause mortality. The results of the indirect treatment comparison showed statistically significant differences in favour of metreleptin on HbA1c, triglycerides, liver enzymes and reducing the risk of pancreatitis at 12‑month follow up. It also suggested that survival was worse with metreleptin compared with supportive care, but the difference was not statistically significant.

The committee agreed in a previous meeting that the NIH follow-up study was representative of people with lipodystrophy in the UK and appropriate to be used for treatment comparison. However, with additional evidence available from the early access programme, the ERG raised 2 concerns related to the indirect treatment comparison. These were the representativeness of the NIH follow-up study used to inform the indirect treatment comparison compared with the early access programme, and the selection of covariates used to adjust for differences between patient cohorts in the indirect treatment comparison. The ERG noted differences in the effects of metreleptin treatment (particularly for changes in triglyceride levels) between the early access programme (see section 4.8) and the NIH follow-up study (see section 4.7). For example, for people with generalised lipodystrophy, the 12‑month change in triglycerides was -3.5 mmol/litre (based only on people for whom both baseline and 12‑month data were available) in the early access programme, compared with -10.54 mmol/litre in the NIH follow up. The ERG noted that the early access programme estimate is closer to that of the GL/PL natural history study estimate of -4.43 mmol/litre. The change in HbA1c at 12 months was also lower in the early access programme (-1.5% for those with generalised lipodystrophy) than in the NIH follow-up study (-1.94% for all patients). Given these differences and the fact that people enrolled in the early access programme are from the UK, the ERG stated that the early access programme should be used in the indirect treatment comparison instead of the NIH follow-up study. The company explained that the early access programme had a much smaller sample size (n=31) and started over a decade ago. The eligibility criteria have evolved along with the growing evidence base for people with lipodystrophy. The clinical experts highlighted the limitations of the early access programme data, namely that people did not have poor enough metabolic status at baseline when they were enrolled in the programme (see section 4.8). They added that the early access programme was not a research study, it was based on 'compassionate use' rather than formal clinical criteria. One of the clinical experts involved in the early access programme explained that recent outcomes are becoming more and more aligned with those of the NIH study follow up. This is because of the application of recent clinical inclusion criteria (see section 4.8). The clinical experts stated that the indirect treatment comparison would be better informed by the NIH follow-up study. The ERG was also concerned about the limited covariates adjusted for in the indirect treatment comparison. This is because important prognostic covariates such as baseline HbA1c and triglyceride levels were not controlled for in the analysis. The company argued that these were not confounding factors because they were not related to treatment allocation. The company also explored using additional covariates in a sensitivity analysis. But, it stated that this analysis was not feasible because of small sample size, the high proportion of missing data and the large number of potential covariates. The company explained that there was more than 90% of missing data and, despite a small sample number (n=31), the 3 methods it used to adjust for cofounders (inverse probability weighting, multivariate regression analysis, and naive comparison) showed consistent results with statistical significance. The company added that its approach was driven by clinical opinion and statistical principles. The committee recognised the limitation of the company's indirect treatment comparison. However, it understood the challenges in generating evidence for ultra-rare diseases given the availability of the data. The committee concluded that, despite the limitations, it was broadly satisfied with the company's methods for the indirect treatment comparison.

The primary outcomes measured in the clinical studies included absolute change in HbA1c levels and percent change in fasting serum triglyceride levels from baseline to month 12. Metreleptin was associated with a statistically significant improvement in both primary outcomes compared with baseline in NIH 991265/20010769. In the generalised lipodystrophy population this was -2.2, percentage point (pp), p<0.001 for HbA1c levels and -32.1%, p=0.001 for triglyceride levels. In the partial lipodystrophy population this was -0.6 pp, p=0.005 for HbA1c and -20.8%, p=0.013 for triglyceride levels. Decreases in HbA1c and triglyceride levels were not statistically significant in the FHA101 study. Additional data from the early access programme (at 36‑month follow up) that became available for the resubmission also suggested that metreleptin reduced HbA1c in both people with generalised and partial lipodystrophy (-1.1% and -1.2% respectively) and fasting triglyceride levels (-57.6% and -23.9%, respectively). No statistical significance was reported. The company also presented evidence on metreleptin's treatment effect on the liver enzymes, ALT and AST. This showed treatment effect in favour of metreleptin at 12‑month follow up (see section 4.9).

The committee discussed HbA1c, triglyceride and liver enzyme levels as surrogates because changes in these outcomes were used to adjust long-term transitions between states in the modelling. As part of their resubmission, the company held a Delphi panel involving 10 international clinical experts (3 of whom were clinical experts from Addenbrooke's Hospital). The Delphi panel reached a consensus that HbA1c is a good predictor of diabetes-related complications including cardiovascular disease, kidney disease, retinopathy and neuropathy in people with lipodystrophy. The company explained that the relationship between HbA1c and long-term clinical outcomes based on 30‑year follow up in diabetes studies is established and widely accepted. The clinical experts explained that triglyceride levels were strongly linked to pancreatitis but that the association was harder to capture mathematically. The clinical experts agreed that HbA1c and triglyceride levels are used in clinical practice and are considered to be reasonably predictive of clinical outcomes. However, the relationship may not be identical to that in other disease areas such as diabetes. For example, the risk of early death is greater for people with lipodystrophy compared with people with type 2 diabetes and metabolic syndrome. They stated that, in general, people with lipodystrophy with lower HbA1c and triglyceride levels have a better prognosis than people with higher levels. They also noted that the effects of HbA1c and triglycerides were broadly correlated but there may be differences between individuals with lipodystrophy depending on other independent factors, for example, genetic factors. They added that HbA1c can predict for complications of the eye and nervous system, as in type 2 diabetes, but differently for other organ complications such as complications of the kidney, liver and heart. They also stated that many people can stop insulin completely after starting leptin treatment. The committee also considered the merit of liver enzymes as a surrogate outcome for liver complications. Clinical experts explained that these are poor predictors of liver disease. The company stated that, because it knew liver enzymes were not a good surrogate, it sought the opinion of clinical experts (through the Delphi panel) on the likely effect of metreleptin on liver disease. The committee asked about the likelihood of HbA1c and triglyceride levels falling without metreleptin. The clinical experts stated that if people had not previously been given any dietary advice, that giving it at the start of treatment could result in improved levels. But, the extent of improvement may be limited because of hyperphagia. In fact, they stated that without treatment, HbA1c levels are likely to reverse to their baseline values over 6 to 12 months, or sooner. The committee noted that evidence showed that metreleptin is effective in lowering HbA1c and triglyceride levels. Evidence also showed that metreleptin may affect liver enzymes but their correlation with the progression of liver disease is less clear. The committee concluded that using HbA1c to predict diabetes-related complications in people with lipodystrophy is acceptable but the relationship between liver enzymes and liver disease is uncertain.

The clinical studies did not include an objective measure to capture the effect of metreleptin on hyperphagia even though, as the patient experts explained, it is a defining characteristic of lipodystrophy with important physical and emotional consequences. The clinical experts agreed that treatment targeting hyperphagia is critical because eating less means some complications of lipodystrophy will improve. The clinical experts also explained that hyperphagia is caused by a deficiency in the hormone leptin (see section 2.1). They also noted that it was hard to predict the impact of metreleptin effect on hyperphagia and quality of life, because there is no simple relationship between eating less and quality of life. The committee acknowledged the expert comments and agreed that any improvements in hyperphagia would be important, but difficult to capture when considering the clinical benefits of metreleptin. The ERG noted that the NIH study measured food intake (kcal) per day in a subset of patients. This showed that, while intake decreased from baseline initially (4‑month assessment), decreases were not statistically significant at the end of the year. Results from the NIH follow-up study showed that 99% of people who had metreleptin reported improvements in hyperphagia. The ERG highlighted that improvements were assessed in a review of medical notes and, although results suggested metreleptin improved hyperphagia, these judgements were not made using an objective measure. The patient experts stated that, since starting metreleptin, they had experienced a feeling of fullness after eating and that this had dramatically altered their lives. One patient expert also noted that, before their daughter started metreleptin, they had to lock cupboards because her intense hunger led her to search for food. But, since starting the treatment her hunger had reduced, and food seeking had stopped. The clinical experts agreed that metreleptin reduces hunger and food seeking. They also noted that children with congenital leptin deficiency could continue benefiting from metreleptin when having treatment for over 20 years, including avoiding obesity and related comorbidities. The committee queried whether there was any continued effect of metreleptin if patients stop the treatment. Clinical experts stated that symptoms of hyperphagia will return after a few days once metreleptin is stopped. The committee noted that, in the studies, a substantial number of patients stopped metreleptin treatment. The clinical experts explained that this could be because metreleptin only treats leptin deficiency and does not correct problems associated with loss of adipose tissue. The company explained that some patients who were seen as stopping treatment had remained on metreleptin but moved to local treatment centres where treatment was provided through compassionate use programmes. It stated that it intended to capture patient-reported outcomes, including hunger scores, as part of its post-authorisation commitments. The committee agreed that any additional information (qualitative or quantitative) about the experience of hyperphagia in people with lipodystrophy, and the effect metreleptin has on it, could substantially improve its understanding of the potential clinical benefits. It concluded that capturing the effect of metreleptin on hyperphagia could be challenging but was important to assess the nature and magnitude of any clinical benefits of metreleptin while patients are on treatment, and also after stopping treatment.

The committee queried whether all people diagnosed with lipodystrophy and covered by the marketing authorisation would be expected to have treatment with metreleptin. The clinical experts stated that most people with generalised lipodystrophy have hyperphagia and organ dysfunction, so would be expected to benefit from leptin treatment. The committee was also aware that everyone diagnosed with generalised lipodystrophy had leptin treatment as part of the early access programme at Addenbrooke's Hospital.

The marketing authorisation for partial lipodystrophy (for people aged 12 and over) was not restrictive. But, the committee was aware that several different subsets of people with partial lipodystrophy had been presented in the company evidence:

• NIH studies 991265/20010769 included leptin level criteria in their definition of people with partial lipodystrophy (used for the indirect treatment comparison):

  • NIH 2001769: below 12.0 nanogram/ml in women, below 8.0 nanogram/ml in men, and below 6 nanogram/ml in children between 6 months and 5 years

  • NIH 991265: 8.0 nanogram/ml and below in women and 6.0 nanogram/ml and below in men

  • HbA1c 6.5% or more or triglycerides 5.65 mmol/litre or more, or both.

• A subgroup of people with partial lipodystrophy from the NIH studies with baseline HbA1c 6.5% or more, or triglycerides 5.65 mmol/litre or more, or both (used for model inputs for the economic analysis, including baseline levels of metabolic surrogates and changes from baseline of HbA1c). The company considered that this subgroup of people with partial lipodystrophy represents a more severe group compared with the overall population of people with partial lipodystrophy in the NIH studies. This is because they are more at risk of organ damage and this best reflects the licensed indication.

• The early access programme included people with partial lipodystrophy with baseline leptin below 12 nanogram/ml and HbA1c 6.5% or more, or triglycerides 5.65 mmol/litre or more, or both (used by the company for baseline patient distribution in the model). The early access programme also included people with partial lipodystrophy who did not meet these criteria (see section 4.8).

• After the technical engagement, the company suggested further restricting the group of people with partial lipodystrophy from their original definition (HbA1c 6.5% or more, or triglycerides 5.65 mmol/litre or more, or both), to baseline HbA1c above 58 mmol/mol (7.5%), or fasting triglycerides above 5.00 mmol/litre, or both. But, they accepted the HbA1c could be lower under exceptional circumstances such as in cases of extreme hyperphagia, or because of severe side effects from other glucose-lowering medications or other serious complications of inadequate metabolic control such as progressive liver disease. The company explained these restrictive criteria were developed based on the clinical evidence from the NIH studies 991265/20010769, UK clinical expert experience and consensus from European lipodystrophy centres.
  
  The ERG pointed out that there was no clinical or economic evidence for this further restricted subgroup of partial lipodystrophy proposed by the company. It was concerned that the effect on the population size eligible for treatment was unknown given the lack of information on the number of people who would actually fulfil these criteria, including the exceptions. It noted that the narrower population would cause issues around application of the various data sources to UK clinical practice. The clinical experts explained that response to metreleptin is greater in people with lipodystrophy and with more severe metabolic status (that is, HbA1c and triglyceride levels) at baseline. They further explained that the criteria of 'HbA1c above 7.5%' was based on the recommended value for a person with diabetes to target in clinical practice. The criteria of 'fasting triglycerides above 5.00 mmol/litre' was based on the value above which people with diabetes are at risk of pancreatitis. The committee understood the rationale for choosing the further restricted criteria for the subgroup of people with partial lipodystrophy. It acknowledged the ERG's concerns, but agreed that partial lipodystrophy with more severe metabolic status (that is, HbA1c above 58 mmol/mol (7.5%), or fasting triglycerides above 5.00 mmol/litre, or both) may benefit more from metreleptin. Because the subgroup of people with partial lipodystrophy included in the economic model had less severe metabolic status (HbA1c 6.5% or more, or triglycerides 5.65 mmol/litre or more, or both), the estimated cost effectiveness for metreleptin was likely to be too high. The committee concluded it would take this into account in its decision making.

The committee noted that the proportion of patients in the main clinical trials who had a treatment‑emergent adverse event (TEAE) was high:

• In the NIH studies, around 89% of patients with generalised lipodystrophy and 85% with partial lipodystrophy had an event. This was severe in 44% and 39% of patients, respectively, and caused 8% and 2%, respectively, to stop treatment.

• In the FHA101 study, around 78% of patients with generalised lipodystrophy and 84% with partial lipodystrophy had a TEAE. This was severe in 67% and 28% of patients, respectively, and caused 11% and 9%, respectively, to stop treatment.
  
  The company commented that stopping happened not only because of adverse events, but also because the studies included some patients for whom metreleptin was not effective because their condition was not related to leptin. Also, some patients stopped treatment because they were pregnant. The company added that TEAEs happened over a long period of time, more than 3 years in the NIH studies and over 2 years in the FHA101 study, and that several reported TEAEs (such as decreased appetite and weight, and hypoglycaemia) were consistent with metreleptin's mechanism of action. The clinical experts suggested that adherence is likely to improve in practice as the benefits of metreleptin become clear. The company and clinical experts also noted that episodes of pancreatitis improved with metreleptin compared with baseline. A patient expert highlighted that she had gone from having frequent events of pancreatitis to no events since starting metreleptin treatment. The ERG noted that, in its original submission, the company only included data for pancreatitis as an adverse event happening after metreleptin withdrawal: 4 patients with generalised lipodystrophy and 2 patients with partial lipodystrophy had treatment-emergent pancreatitis across studies (1 patient died, 5 recovered). However, the ERG also noted that the NIH data indicated that patients had improvements in pancreatitis on metreleptin. The company noted that TEAEs were not reported in the data collected from the early access programme. The ERG highlighted that this contradicted the company's comments from the first evaluation committee meeting that additional adverse event information could be provided. The committee agreed that additional real-world data on adverse events would be informative. But, based on what it had heard from the experts, it concluded that the tolerability profile of metreleptin was likely to be acceptable.

In the NIH 991265/20010769 and FHA101 studies, 88% of people developed antibodies to metreleptin, and 4% developed neutralising antibodies to metreleptin. This raised concern that developing neutralising antibodies could affect metabolic control and immune function. The company explained that neutralising antibodies were only reported in 4% of people. Therefore, it did not anticipate this affecting a significant proportion of people or affecting outcomes such as HbA1c and triglyceride levels in the long term. Clinical experts stated that neutralising antibodies are rare, and although relevant to some people, do not seem to be a frequent problem in clinical practice. The same issues are seen in other metabolic diseases, but it does not affect the efficacy of the drug. They noted that neutralising antibodies should still be monitored. The committee concluded that neutralising antibodies did not seem to be a significant issue, but this should still be monitored in the future.

The company developed a de novo individual patient-level model addressing previous committee concerns (see section 4.2). The model consists of 6 Markov sub-models. These simulate the progression of disease on distinct organ systems affected by lipodystrophy, capturing the key lipodystrophy-related complications that affect health-related quality of life, costs and mortality over the lifetime of people with lipodystrophy. The 6 organ sub-models are: pancreas, liver disease, cardiovascular disease, kidney, neuropathy (nerves) and retinopathy (eyes). In each cycle (which lasted for 1 year), a person is simultaneously in a single discrete health state in each of the 6 independent organ sub-models. A person can die during each cycle, in which case they are removed from all sub-models into the death state.

• Pancreas sub-model starts in the 'no pancreatitis' state when a person is at risk of developing pancreatitis.

• Liver sub-model is based on NICE's guideline on non-alcoholic fatty liver disease (NAFLD). People transition through health states, from having no or asymptomatic fibrosis, to advanced fibrosis, to compensated and then decompensated cirrhosis.

• Cardiovascular sub-model: people start in the 'no cardiovascular disease' state, when they are at risk of experiencing cardiovascular complications (stroke, angina, congestive heart failure, and myocardial infarction).

• Kidney sub-model is based on the structure of the Sheffield diabetes model. People start in the 'no chronic kidney disease' state, when they can transition through microalbuminuria and macroalbuminuria to the end-stage renal disease (ESRD) health state. From ESRD, people can transition to receiving a kidney transplant (tunnel state), moving to the post-transplant state in the following cycle.

• Neuropathy sub-model: people start in the 'no peripheral neuropathy' state and can progress to peripheral neuropathy, peripheral arterial disease with amputation, and can move to the post‑amputation state in the following cycle.

• Retinopathy sub-model: people start in the 'no retinopathy' state and can progress to blindness either directly, or by progressing through various retinal diseases such as background retinopathy, proliferative retinopathy and macular oedema.

• People are at risk of death in all states.
  
  The ERG noted that the company's sub-model structure is an improvement from previous submissions (see section 4.2). This is because it is better structured to account for the potential progression of complications related to lipodystrophy over time. However, it relies on the assumption that people with diabetes or elevated triglyceride levels, because of lipodystrophy, will follow a similar course to people with similar metabolic abnormalities but different aetiologies. It stated that this was an area of considerable uncertainty. The company explained that its choice to follow a diabetes framework for its model structure was in response to the committee's previous suggestion (see section 4.2). The model is based on the structure of the diabetes-related complications seen in the Sheffield diabetes model, and the established model structure from NICE's guideline on NAFLD to reflect liver disease progression. It chose the Sheffield diabetes model, among others, because it considered it the most robust, and it was previously used in a multiple technology assessment for diabetes. The committee was satisfied with the company's attempt to capture the disease progression using sub-models for each relevant organ. It was aware that there are uncertainties as to what extent people with lipodystrophy will follow a similar course to people with diabetes and fatty liver disease. However, given the scarcity of evidence in lipodystrophy, the committee concluded that the model structure is appropriate for decision making.

The company used published literature for baseline transition probabilities for 4 of the sub-models that are diabetes-related complications: cardiovascular, kidney, neuropathy and retinopathy. Baseline transition probabilities were then adjusted according to the absolute change in HbA1c level from baseline to 12 months from NIH 991265/20010769, to generate probabilities for people having metreleptin. For the liver sub-model, the baseline transition probability for liver disease was derived from the model in NICE's guideline on NAFLD, and then adjusted using reduction in risk estimated from the Delphi panel (company's base case) or changes in liver enzymes from the indirect treatment comparison (company's scenario analysis). For the pancreas sub-model, the baseline transition probability for pancreatitis was directly sourced from the GL/PL natural history study. It was then adjusted using odds ratios estimated from the indirect treatment comparison (see section 4.9).

For the 4 diabetes-related sub-models, the ERG noted that the company did not adjust the transition probabilities based on triglycerides, because of a lack of data in the literature. The company explained that it meant that the transition probabilities were underestimated in cases when hypertriglyceridaemia contributes to the risk of a complication, such as cardiovascular disease. The committee noted that using only HbA1c to adjust the transition probabilities in the 4 diabetes-related sub-models could either under or overestimate metreleptin's treatment effect on those organ complications. But, the exact magnitude of that is unlikely to be quantified because of limited evidence. It agreed that, qualitatively, HbA1c has value as a surrogate for clinical outcomes in the cardiovascular, kidney, neuropathy and retinopathy sub-models. It was generally satisfied with the approach taken by the company to model transition probabilities. The ERG preferred using the change in liver enzymes (trial-based data) estimated from the indirect treatment comparison (rather than the risk reduction estimate directly obtained from the Delphi panel) to adjust the baseline transition probability for the liver in their base case. The committee recalled that liver enzymes are a poor predictor for liver complications (see section 4.12). It agreed that the Delphi panel, based on clinical experts' judgement, would be an appropriate source to inform the liver transition probabilities in this case. It was also aware that the choice between liver enzymes and the Delphi panel estimates only had a minor impact on the cost‑effectiveness estimate. The committee concluded that it was generally satisfied with the company's approach to modelling transition probabilities.

The committee was aware that the clinical trials of metreleptin did not collect any quality-of-life data. In its resubmission, the company did a systematic literature review to identify sources of utility values from the literature. The utility decrement for pancreatitis was based on the discrete choice experiment (DCE) from the original submission. The DCE was done for a large sample of the general population, to estimate utility decrements associated with key lipodystrophy attributes. The utility decrements for the other organs were taken from published sources and those previously used and accepted in NICE's technology appraisal guidance on type 2 diabetes and fatty liver disease. The committee agreed with using utility decrements from other conditions than lipodystrophy, given the scarcity of data for this condition. Also, to capture specific symptoms not already accounted for in the sub-models (that is, hyperphagia, polycystic ovary syndrome, inability to work and impaired physical appearance), the company's model applied a utility differential (compared with supportive care) of 0.12, based on the rescaled DCE from the original company submission (see section 4.2). The committee recalled it had had concerns relating to the validity of the utility values estimated from the DCE (see section 4.2). The company acknowledged the limitations but stated that no alternative value could be sourced from the literature. The patient expert explained that people with lipodystrophy with organ damage may also have fertility issues, pain and fatigue, so those should be taken into account in the model. The clinical expert added that, even when including the utility differential of 0.12, there could be other symptoms or factors not accounted for. The committee concluded it was generally satisfied with the approach to include a utility differential to account for symptoms not captured by estimates of utility related to organ damage.

The company recalled that the committee had concerns about how it had modelled carer utility in its previous submission (see section 4.2). In its resubmission, the model also included a utility decrement to account for the burden on carers of -0.0986. This decrement was estimated as the difference between the mean value for carers in the Lipodystrophy Caregiver Burden Survey (done by the company to explore carers experience in the UK) and the general population norm, obtained from the EQ‑5D. The committee was satisfied with the company's approach to model carer utility.

The company assumed that each patient had 2 carers. This was the median and the rounded value of the mean of 1.67 carers, based on the Lipodystrophy Caregiver Burden Survey. The company argued that the median is more likely to be representative of the number of carers in UK clinical practice and was validated by UK patient experts. The ERG used the mean of 1.67 in their base-case model as is usual in health economic modelling. The committee agreed with using 1.67 carers from the ERG because not all people with lipodystrophy will have 2 carers. It concluded that it was satisfied with the ERG's approach.

The company's model structure for metreleptin included a stopping rule for people with partial lipodystrophy. At 9 months after the start of metreleptin treatment, a specialist service review will see whether treatment should be stopped if the following metabolic criteria have not been met: an HbA1c reduction of at least 0.75% from baseline, or a fasting triglyceride reduction of at least 50% from baseline. A clinical expert stated that there was a consensus in European lipodystrophy centres to stop metreleptin at 6 to 9 months if the person did not take the treatment properly or did not engage with appointments, or there was no HbA1c reduction of at least 0.5% from baseline or a fall in fasting triglycerides of at least 50% from baseline. The clinical expert can apply their own judgement and agree to continue to treat the lipodystrophy even if the criteria are not met. The committee noted that the company's stopping rule did not match the one from the clinical experts, but the rule proposed did incorporate significant clinical discretion.

The committee understood that NICE's interim process and methods of the highly specialised technologies programme (2017) specifies that a most plausible incremental cost-effectiveness ratio (ICER) of below £100,000 per QALY gained for a highly specialised technology is normally considered an effective use of NHS resources. For a most plausible ICER above £100,000 per QALY gained, judgements about the acceptability of the highly specialised technology as an effective use of NHS resources must take account of the magnitude of the incremental therapeutic improvement. This is seen through the number of additional QALYs gained and by applying a 'QALY weight'. It understood that a weight between 1 and 3 can be applied when the QALY gain is between 10 and 30 QALYs. The committee considered that although there was uncertainty, the undiscounted QALY gains for the scenarios incorporating its preferred assumptions did not meet the criteria for applying a QALY weight.

The company's base case showed that metreleptin was associated with an ICER of £60,611 per QALY gained for the overall population. The ICER for generalised lipodystrophy was £46,000 per QALY gained, and the ICER for partial lipodystrophy was £81,584 per QALY gained. The total costs and QALYs are considered by the company to be commercial in confidence and so cannot be reported here.

The ERG's base case showed that metreleptin was associated with an ICER of £110,460 per QALY gained for the overall population. The ICER for generalised lipodystrophy was £92,593 per QALY gained, and the ICER for partial lipodystrophy was £130,334 per QALY gained. The total costs and QALYs are considered by the company to be commercial in confidence and so cannot be reported here.

Considering both the company and ERG's scenario analyses, the committee's preferred assumptions were:

• using the estimate directly obtained from the Delphi data to adjust transition probabilities in the liver model (rather than liver enzymes, see section 4.19)

• reversing HbA1c to baseline level after stopping treatment (excluding 0.15% drift, see section 4.23)

• maintaining liver benefit for 1 year when metreleptin is stopped (see section 4.24)

• removing assumed lifetime maintenance of 50% of quality-of-life treatment differential and carer utility gain after stopping metreleptin (see section 4.25 and section 4.22)

• correcting number of carers to 1.67 (rather than 2 in company base case, see section 4.22).
  
  The committee's preferred assumptions were associated with an ICER of £108,267 per QALY gained (for the overall population). The ICER for generalised lipodystrophy was £87,545 per QALY gained, and the ICER for partial lipodystrophy was £133,606 per QALY gained.