Draft guidance consultation

Alagille syndrome is a rare condition caused by mutations in the genes encoding JAG1 or NOTCH2, which are involved in the NOTCH signalling pathway. It can affect different organs including the liver, heart, skeleton, eyes, kidneys and vascular system. It is thought to occur in 1 in 30,000 to 1 in 70,000 live births. About 85% of children with Alagille syndrome have cholestasis characterised by retained bile acids in liver cells because of impaired secretion or obstruction of bile flow. Some of these bile acids join the systemic circulation, leading to increased levels of serum bile acids (sBAs) and bilirubin, which are biomarkers of cholestasis. Cholestasis can lead to pruritus (itching) and a range of liver complications, including cirrhosis (in about 46% of people with the condition), ascites (57%) and portal hypertension (40%). Severe and debilitating pruritus affects about 74% to 88% of people with cholestasis, and commonly starts from between 6 and 14 months after birth. Other non-liver symptoms of cholestasis in Alagille syndrome include:

• hypercholesterolemia

• fat-soluble vitamin deficiency

• xanthomas (fatty deposits on the extensor surfaces)

• chronic fatigue and sleep disturbances

• neurocognitive deficits

• dental damage

• growth failure.

The patient and clinical experts explained that cholestatic pruritus is the symptom of Alagille syndrome that has the largest impact on a person with the condition's life. Unremitting and severe pruritus can have a devastating effect on all aspects of life for the person and their families. The patient experts explained that the physical distress of constant pruritus affects a person's sleep. This can often cause difficulty in concentrating, and delays in developmental and educational milestones because of tiredness, the physical need to itch and absence from school. The patient experts also explained the constant need to consider external factors that can exacerbate the pruritus such as the weather and type of clothing. Debilitating pruritus can lead to skin lesions and extensive scarring, which can have a profoundly negative impact on the person's quality of life. Also, constant scratching can be stigmatising, lead to bullying and have a social impact because of its association with having fleas and lice, and poor hygiene. This, along with the physical scarring can affect a person's self-confidence and social relationships, which can cause anxiety, and affect mental and psychological wellbeing. They explained that there can be feelings of hopelessness because of the lack of effective treatments, which can lead to suicidal ideation. The patient experts explained the wider impact of the condition on parents, carers and siblings. For example, pruritus usually starts when a person with Alagille syndrome is under 6 months, and the need for constant comfort means that parents' quality of sleep is also affected. Even though the primary carer is often the mother, the physical and mental burden of full-time caring can affect both parents. It can affect their ability to work, carry out daily activities and care for other siblings, whose mental and emotional wellbeing may also be affected. The impact on family life can be devastating, including relationship breakdown, separation and divorce, the need for psychotherapy or counselling, and an impact on finances. The committee acknowledged that cholestatic pruritus in Alagille syndrome can have an extremely negative impact on the person with the condition, and on their family and carers.

The committee recalled feedback from the clinical and patient experts that highlighted the multidimensional and subjective nature of chronic pruritus (see section 3.2). It acknowledged the difficulty and complexity in assessing the severity and impact of pruritus, especially in a paediatric population. The committee noted that the eligibility criteria of the maralixibat clinical trials included a clinical outcome measure for the severity of pruritus. This was the Itch Reported Outcome (ItchRO), Patient (Pt) or Observer (Obs); see section 3.6). It noted that some of the trials also reported outcomes on pruritus severity using the Clinician Scratch Scale (CSS; see section 3.8). The ItchRO scores pruritus severity on a 5‑point Likert scale from 0 to 4 (where 0 represents no itch and 4 represents the most severe itch). Similarly, the CSS use a 5‑point Likert scale (where 0 represents no visible skin damage and 4 represents cutaneous mutilation, haemorrhage and scarring). The clinical expert explained that these scales are rarely used in clinical practice. Rather, healthcare professionals discuss the impact of the pruritus on the quality of life of the child and family, particularly on sleep quality and the ability to function during the day. They emphasised that the impact on quality of life was the main factor guiding management. The committee noted that the maralixibat trials had all reported health-related quality-of-life data using the Paediatric Quality of Life Scale (PedsQL). It acknowledged the impact that chronic pruritus can have on secondary conditions such as sleep disorders, anxiety, depression and quality of life. It concluded that the clinical assessment of pruritus severity should consider the intensity and the impact on health-related quality of life using validated tools.

There are no NICE guidelines or NICE technology appraisals guidance on managing cholestatic pruritus in Alagille syndrome. The current clinical management without maralixibat involves using off-label treatments, including:

• bile acid binding resins (such as colestyramine)

• ursodeoxycholic acid (a synthetic bile acid)

• rifampicin that makes bile acids less pruritic and more excretable by kidneys

• opioid antagonists (such as naltrexone).

For this evaluation, the company positioned maralixibat as an add-on treatment to standard care for the treatment of cholestatic pruritus in people with Alagille syndrome 2 months and over. The clinical expert explained that, in clinical practice, it is likely that maralixibat would be offered after people have had standard care. The committee agreed with the company's positioning of maralixibat as an add-on treatment to standard care. It concluded that the company's choice of comparator was appropriate.

The key clinical-effectiveness evidence came from 3 phase 2 multicentre double-blind randomised controlled trials with open-label extensions. These compared maralixibat plus standard care (maralixibat from now) with placebo plus standard care (placebo from now) in children aged 1 year to 18 years. They were all diagnosed with Alagille syndrome and moderate to severe pruritus defined as an ItchRO(Obs) score of at least 2 for 2 consecutive weeks. The trials were:

• ICONIC: this was considered by the company to be the pivotal trial, and included 31 people from the UK, Europe and Australia. It comprised an 18‑week open-label single-arm phase in which people had up to 380 micrograms/kg/day of maralixibat. This is the recommended dose in the summary of product characteristics for maralixibat. This was followed by a randomised withdrawal phase in which people were randomly allocated to either continue with maralixibat or to have placebo for 4 weeks. People were followed up for up to 204 weeks. The primary endpoint was the mean change from baseline in fasting sBA after 4 weeks of treatment in people whose condition had responded to treatment. Response was defined as at least a 50% decrease in sBA from baseline.

• ITCH: this included 37 people from Canada and the USA. People were randomised to have 13 weeks (5 weeks of dose escalation and 8 weeks on a stable dose) of maralixibat up to 280 micrograms/kg/day or placebo. They were followed up for up to 220 weeksin the extension study, IMAGINE 2. For ITCH, the primary endpoint was the mean change from baseline at 13 weeks in the ItchRO(Obs) score.

• IMAGO: this included 20 people from the UK. People were randomised to 13 weeks (5 weeks of dose escalation and 8 weeks on a stable dose) of maralixibat up to 280 micrograms/kg/day or placebo. They were followed up for 156 weeks or more in the extension study, IMAGINE. For IMAGO, the primary endpoint was the mean change from baseline at 13 weeks in fasting sBA levels.

The EAG highlighted the differences in characteristics between the maralixibat trials and its marketing authorisation with respect to age, definition of pruritus and maralixibat doses. It highlighted that the company had not provided clinical evidence for adults, and had not investigated the differential impact of using licensed and unlicensed doses in the maralixibat trials. It also noted that the placebo arms excluded colestyramine, which the EAG thought is commonly used in clinical practice. The clinical expert explained that colestyramine was excluded from the maralixibat trials because of the potential confounding effect it would have on the outcomes. The committee noted the clinical expert's differing view to the EAG's, that colestyramine is rarely used in the NHS (see section 3.4). It was unclear whether maralixibat would be used as an add-on treatment to colestyramine (see section 3.5). The committee noted that if it is, this would impact the generalisability of the trials to NHS practice. The committee acknowledged the differences in the baseline characteristics of the people included in the maralixibat trials and its broader marketing authorisation. But it acknowledged the rarity of Alagille syndrome. It concluded that the trial populations are likely representative of people in the NHS who would have maralixibat.

In its economic model, the company defined treatment response using the surrogate biomarker sBA in its base case rather than using clinical outcomes. Treatment response was defined as a 50% decrease in sBA levels from baseline at 12 weeks. The committee noted that the primary endpoint for clinical effectiveness in ICONIC and IMAGO was the absolute mean change from baseline in sBA levels (see section 3.7). It was not a dichotomous assessment of treatment response using the 50% cut off of sBA levels. The clinical expert explained that sBA is not a validated surrogate endpoint but is an important biomarker for treatment response in cholestatic pruritus. They confirmed that pruritus does not occur when sBA levels are within the normal or near-normal range. They explained that cholestatic pruritus stems from retained bile acids, and that a reduction in sBA level suggests a physiological reaction in response to treatment. The clinical expert thought that a 50% decrease in sBA from baseline indicates a large physiological change. They added that a smaller percentage decrease in sBA levels would also indicate a response to treatment. Also, in clinical practice, treatment response is usually assessed based on descriptive feedback of symptoms and their impact from the person with Alagille syndrome, or their family and carers. But the clinical expert explained that sBA levels may be used to confirm that a physiological response had occurred. The committee noted that the maralixibat trials also reported treatment response based on:

• a 70% threshold for sBA levels

• the clinical outcomes of ItchRO(Obs) using thresholds of -1, -1.5 and -2 change from baseline, and

• CSS using thresholds of -1 and -2 change from baseline.

In its pivotal trial, ICONIC, the company reported statistically significant reductions in sBA levels and ItchRO(Obs) scores after 4 weeks of treatment with maralixibat compared with placebo. When using different definitions of treatment response in ITCH (see section 3.8), there was a treatment response in more people having 13 weeks of maralixibat (below the recommended dose; see section 3.6) than having placebo. The company consider the absolute figures to be confidential, so they cannot be reported. The EAG thought that the maralixibat trials were exploratory, rather than confirmatory because:

• they were limited in terms of the short randomised periods (4 or 13 weeks)

• there was no wash-out period in ICONIC following initial maralixibat treatment in both arms

• of the lack of long-term comparative efficacy data because of the open-label single-arm extension follow-up phases.

In its submission, the company made the case that maralixibat plus standard care compared with standard care alone prolongs life and improves health-related quality of life. It assumed that, because treatment reduces the level of bile acids retained in the liver, it:

• slows down the progression of Alagille syndrome to severe stages of liver disease (cirrhosis, portal hypertension and ascites)

• delays the need for surgery (surgical biliary diversion or liver transplant).

In its base case, the company used sBA data from the initial 12‑week single-arm open-label prerandomised phase of ICONIC to inform the treatment response of the maralixibat arm. It assumed that, in the standard care arm, there would be no response to treatment in anyone. The EAG highlighted that data from the randomised phase of ITCH showed that Alagille syndrome would respond to treatment in some people in the placebo plus standard care arm using the 50% sBA response threshold applied in the company's base case. The clinical expert noted the small North American sample size and the uncertainty in the data from ITCH. But they thought that there would be some people in NHS clinical practice whose condition would likely respond to standard care. The committee acknowledged the methodological limitations, including the small population size in ICONIC and ITCH because of:

• the rarity of the condition (see section 3.6)

• the potential for a placebo effect

• the variability of the condition and regression to the mean.

The company used overall-survival Kaplan–Meier data from the treatment-naive population of the GALA study to extrapolate long-term mortality risk. It did this using the log-logistic distribution in the non-response health states. It applied the mortality risk from age 2 months in line with maralixibat's marketing authorisation. The company explained that it had applied general population life expectancy such that the model predicted lower survival in people with Alagille syndrome than in the general population. The EAG thought that all the extrapolations were unrealistic because of the immature survival data. It preferred using the exponential distribution. This was because the median survival was 77 years compared with 217 years using the company's preferred log-logistic distribution. The EAG also preferred applying the mortality risk from age 2 years, which represented the lower quartile of the GALA population age. The clinical expert confirmed that the mortality risk is greater in people with Alagille syndrome at all ages than in the age-matched general population. They explained that, typically, people with Alagille syndrome have a higher risk of dying when they are younger but the risk decreases as the person grows older. The committee acknowledged the unrealistically high median survival of 217 years using the company's preferred log-logistic distribution. It noted that background mortality for the general population was applied such that the long-term extrapolations made little difference to which distribution was chosen. It thought that the company's model best reflected the early mortality of the condition. It concluded that the company's modelling of overall survival for the non-response health states was acceptable for decision making.

The company modelled a greater probability of surviving in people whose condition had responded to maralixibat. It did this using data for a surrogate outcome for overall survival, the composite endpoint of event-free survival used in the GALA cohort comparison study (see section 3.9). It used the hazard ratio of event-free survival data (adjusted HR 0.305; see section 3.9) to adjust the overall survival of the standard care arm to derive the overall survival of people having maralixibat whose condition had responded to treatment. The EAG thought that there was high uncertainty about the hazard ratio for event-free survival because of issues it had previously outlined (see section 3.9). It also highlighted the uncertainty in the crude estimates of death rates in the aggregated maralixibat and GALA standard care cohorts. So, in its base case, the EAG preferred to assume that there was no difference in mortality risk for people whose condition had responded or not to maralixibat (HR 1). The company's submission stated that the long-term reduction in hazard for event-free survival was used to estimate reductions in individual components of event-free survival related to maralixibat treatment. The EAG clarified that event-free survival data from the GALA cohort comparison study was only used as a proxy for overall survival as outlined above. The committee recalled that the GALA cohort comparison study included everyone who had maralixibat regardless of response (see section 3.6). This may have led to an underestimate of the potential benefit of maralixibat. The company explained that it did not have sBA data for the GALA cohort, so could not dichotomise outcomes based on sBA response thresholds. The committee thought that it would have preferred to have seen the GALA cohort comparison study analysed by sBA response thresholds. But it acknowledged the limitations of the available data. The committee had concerns about the lack of clarity about how treatment response related to event-free survival and mortality in the model. The committee agreed it wanted to ensure that the model was producing credible estimates. So, it would have preferred to have seen time-to-event data reported for each event (liver transplants, surgical biliary diversion, liver decompensation and death) in the endpoint event-free survival from the GALA study for standard care compared with the predictions from the model.

The summary of product characteristics for maralixibat states that 'alternative treatment should be considered in patients for whom no treatment benefit can be established following 3 months of continuous daily treatment with maralixibat'. The committee recalled that, in the company's base case, treatment response was defined as a 50% decrease from baseline in sBA levels at 12 weeks (see section 3.8). It recalled the clinical expert testimony that treatment response is normally assessed clinically based on feedback on quality of life from people and their carers. Then, sBA levels are used to confirm a physiological response (see section 3.8). The EAG highlighted that fewer people having maralixibat met the company's treatment sBA response threshold at week 18 than week 12 in the initial prerandomised phase of ICONIC. The clinical experts explained that sBA levels may fluctuate but that any reduction in sBA level would show a physiological response to treatment (see section 3.8). So, it is unlikely that a defined sBA level would be used to indicate when to stop maralixibat in clinical practice. The committee noted that the company had not included treatment effect waning in its model. Instead, the company had modelled the probability of stopping treatment with maralixibat. It assumed this was equivalent to the proportion of people who had stopped maralixibat because of adverse events in the initial 18‑week prerandomised phase of ICONIC. The clinical expert explained that many people with Alagille syndrome and their families also withdraw consent to continue treatment because of inadequate benefit. The committee thought that clarification was needed on the stopping rule and its implementation in clinical practice.

In its base case, the company assumed that 47% of people whose condition does not respond to treatment would have a liver transplant by age 4 years. The company explained that there is a range of estimates for the proportion of people who would be expected to have a liver transplant in the literature. It used a proportion in the model from the higher end of the range, which it said was a conservative estimate. The clinical expert thought that the figures used in the company's model were higher than what they had seen. They suggested that about 25% of people would likely progress to having a liver transplant by age 4 years and 40% by age 11 years. The clinical expert highlighted that there is a shortage of donor organs and that children may die while on the waiting list for a liver transplant. The committee concluded that a transition probability of 25% of people whose condition does not respond to treatment would have a liver transplant by age 4 years should be used in the economic model.

To determine patient and carer health-state utilities in Alagille syndrome, the company applied utilities from its vignette study that used:

• EQ‑5D‑5L

• time trade-off

• visual analogue scale valuation methods.

In its base case, the company derived carer disutility using its vignette study. Carer disutility was applied in the treatment response health state and all the non-response health states. The company considers the utilities to be confidential, so they cannot be reported. Carer disutility was applied to 1.7 carers per person with Alagille syndrome in all health states until age 18 years. The company also commissioned an online survey of the impact of caring for 105 carers of people with Alagille syndrome and cholestasis. It then compared the data with 95 UK-matched parents from the general population. The committee noted that the mean EQ‑5D score for carers (0.82) was similar to the UK-matched sample (0.79). The EAG thought that the utility values lacked external validity. This was because the disutility for the treatment non-response health state was 6 times that of the disutility in the treatment response health state. In its base case, the EAG preferred to assume that there was no carer disutility. But it provided a scenario analysis using assumptions on carer disutility from NICE's highly specialised technology guidance on odevixibat. This was a disutility of 0.05 for treatment response health state and 0.1 for treatment non-response health states applied to 1 carer per person with Alagille syndrome until age 18 years. The committee thought that the disutility for carers used by the company was much higher than disutilities used in other NICE technology appraisals and NICE highly specialised technology evaluations. It had concerns about the methodology used in the vignette study (see section 3.16), and was unclear about how the carer disutility was applied in the model. The committee recalled the patient expert feedback on the impact of caring for a person with Alagille syndrome and cholestatic pruritus (see section 3.2), and that both parents are affected. It concluded that it was appropriate to model carer disutility. But it thought that the likely estimates were between the company's base case and EAG's scenario analysis using assumptions from NICE's highly specialised technology guidance on odevixibat.

In its submission, the company highlighted that people with Alagille syndrome are expected to be shorter and weigh less than the general population. This is because of difficulty digesting fats and absorbing fat-soluble vitamins that leads to malnutrition. In its base case, the company used the fifth percentile bodyweight band from the general population, in line with data observed from ICONIC. Normal distributions based on bodyweight and standard deviation in each cycle were used to estimate the average cohort doses for maralixibat at each timepoint. The EAG thought that the impact of effective treatment in boosting growth should be considered. So, it provided a scenario analysis using the bodyweight band of the twenty-fifth percentile of the general population after 2 years in the model. The company also provided a scenario analysis using the bodyweight band of the seventy-fifth percentile of the general population from the start of the model. The committee thought that it had not seen evidence of the natural growth history of people with Alagille syndrome and cholestatic pruritus, and the impact of effective treatment. It concluded that this was an area of uncertainty and that it would consider a range of bodyweight bands in its decision making.

The committee considered the severity of the condition (the future health lost by people living with the condition and having standard care in the NHS). The committee may apply a greater weight to quality-adjusted life years (QALYs; a severity modifier) if technologies are indicated for conditions with a high degree of severity. The company provided absolute and proportional QALY shortfall estimates in line with NICE's health technology evaluations manual. The company consider the QALY shortfall estimates to be confidential, so they cannot be reported. Using the company's utilities, the committee thought that the severity weight of 1.2 applied to the QALYs was appropriate. But the committee noted that the estimates of the QALY shortfall depended on whether the utilities were derived from the company-preferred vignette study or the EAG-preferred Kamath et al. (2015) mapping study. It concluded that it would consider both cases in its decision making.

The committee thought that neither the company's nor EAG's base cases included all its preferred assumptions. It thought that there were many areas of uncertainty (see section 3.21). It also noted that it had only preferred the company's modelling of overall survival for the non-response health states which used the log-logistic distribution applied from age 2 months (see section 3.12). In addition to the areas of uncertainty highlighted in section 3.21, the committee said that it would like clarification on:

• the criteria that would be used to start maralixibat in NHS clinical practice, for example, using ItchRO(Obs) as in the maralixibat trials or a health-related quality-of-life scale or both (see section 3.3)

• using bile acid binding resins such as colestyramine in the NHS (see section 3.4 and section 3.7)

• the likely use of maralixibat as an add-on treatment to standard care in NHS clinical practice (see section 3.5)

• the evidence of the correlation between sBA levels and the clinical outcomes used in the maralixibat trials, including assessment of health-related quality of life (see section 3.8)

• the evidence of the level of change considered to be clinically meaningful in the clinical outcomes (see section 3.8)

• the details of the stopping rule and how it will be implemented in NHS clinical practice (see section 3.14)

• the methodology used in the company's vignette studies for the health-state utility values and carer disutility (see section 3.16)

• detailed information about the methods of validation and the methods used in the time trade-off study including full descriptions of the health states (see section 3.16 and section 3.17)

• the approach of applying carer disutility in the model (see section 3.17).

NICE's manual on health technology evaluations notes that, above a most plausible incremental cost-effectiveness ratio (ICER) of £20,000 per QALY gained, judgements about the acceptability of a technology as an effective use of NHS resources will take into account the degree of certainty around the ICER. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. But it will also take into account other aspects including uncaptured health benefits. The committee noted the high level of uncertainty, specifically related to the:

• definition of severity of cholestatic pruritus and eligible population in clinical practice (see section 3.3)

• positioning of maralixibat as an add-on treatment to standard care after standard care options had been exhausted (see section 3.5)

• limited evidence on the longer-term comparative effectiveness of maralixibat and lack of data in adults (see section 3.6 and section 3.9)

• treatments used in standard care, in particular, colestyramine (see section 3.7)

• assessment of treatment response (see section 3.8)

• assumptions of the treatment response in maralixibat and standard care in the model (see section 3.9)

• modelling of overall survival in people whose condition responds to maralixibat (see section 3.13)

• the stopping rule for maralixibat and its implementation in clinical practice (see section 3.14)

• the transition probability used for moving from the treatment non-response health state to the liver transplant health state (see section 3.15)

• methods used to derive the health-state utility values, and their plausibility and impact on the severity modifier (see section 3.16 and section 3.19)

• methods used to derive carer disutility, associated assumptions and implementation in the model, and plausible values (see section 3.17)

• assumption of bodyweight of population throughout the time horizon (see section 3.18).

The committee considered the cost effectiveness of maralixibat plus standard care compared with standard care alone. It noted that none of the company's and EAG's base cases and scenarios provided its preferred ICER (see section 3.20). The committee thought that the range of ICERs addressing the areas of uncertainty listed in section 3.21 would likely be above the range that NICE normally considers an acceptable use of NHS resources. The company consider the ICERs to be confidential, so they cannot be reported here.

The company suggested that most carers are mothers who may experience a greater carer burden compared with the other parent. The committee recalled feedback from the patient experts that the primary carer is often the mother, but that both parents are affected by the condition (see section 3.2). The committee concluded that this issue could not be addressed within a technology appraisal.

The committee considered whether maralixibat was innovative. It noted that maralixibat is the first UK licensed treatment option for cholestatic pruritus in Alagille syndrome, a rare multi-organ condition. Maralixibat represents a 'step-change' in a treatment pathway addressing a high unmet need for people with no licensed options available on the NHS. The committee noted that it had identified additional benefits of maralixibat not captured in the economic modelling. But it had already considered this in its decision making (see section 3.21).

The committee concluded that the range of cost-effectiveness estimates for maralixibat plus standard care compared with standard care alone are likely to be above the range that NICE normally considers an acceptable use of NHS resources. So, maralixibat is not recommended for treating cholestatic pruritus in Alagille syndrome.