Naldemedine for treating opioid-induced constipation

3.1 Opioid receptors are present in the gastrointestinal tract. When opioids bind to these receptors they can disrupt normal gastrointestinal function, usually resulting in opioid-induced constipation. Treatment for opioid-induced constipation could be a single treatment with a peripherally acting mu-opioid receptor antagonist (PAMORA) such as oral naloxegol or subcutaneous methylnaltrexone. But it commonly involves a combination of a PAMORA and a conventional laxative. Naldemedine is an alternative oral PAMORA taken as a single daily dose. The clinical expert explained that opioid-induced constipation is very common in people with non-cancer and cancer pain, and continues regardless of the type of opioid used. The expert estimated that over 80% of patients with cancer pain will have opioid-induced constipation, while the prevalence is likely to be lower in patients with non-cancer pain. The clinical expert also highlighted that in clinical practice, many patients taking a PAMORA have mixed aetiology constipation and so need a combination treatment to target the different causes of constipation. For some patients the burden of opioid-induced constipation on quality of life is greater than the pain that needs an opioid. This often means patients stop opioid treatment. The clinical expert said that a key benefit of a PAMORA is that patients can have a normal stool, while those taking conventional laxatives often experience a continual back and forth of being constipated and then having diarrhoea. This is a huge burden for both patients and carers in terms of continually managing bowel function. The committee concluded that people with opioid-induced constipation would welcome a new treatment that improves their constipation symptoms and quality of life.

3.2 The clinical expert confirmed that all relevant comparators had been included in the key subpopulations modelled by the company (see section 3.4, table 1). The clinical expert explained that the available PAMORAs are subcutaneous methylnaltrexone and oral naloxegol. The committee was informed that methylnaltrexone is primarily used to treat severe cases of constipation when a response is needed quickly, before switching to an oral treatment. The comparators included:

3.3 The company submission included 4 pivotal randomised trials (COMPOSE-1, -2 -3 and -4) and 3 supportive open-label safety studies (COMPOSE-5, -6 and -7). The primary outcome for COMPOSE-1, -2 and -4 was the proportion of people who had spontaneous bowel movements. For COMPOSE-3, the primary outcome was measures of treatment-emergent adverse events. The proportion of people who had spontaneous bowel movements was significantly greater in the naldemedine arm compared with placebo for COMPOSE-1, -2 and -4:

3.4 The committee considered several key subpopulations revised by the company after clarification stage and after technical engagement. The committee agreed that subpopulations 1 to 4 reflect the clinical pathway in NHS practice (see table 1 below).

Abbreviations: ITC, indirect treatment comparison; ITT, intention-to-treat analysis; OIC, opioid-induced constipation; SC, subcutaneous injection.

The committee discussed the clinical plausibility of the various subpopulations modelled by the company. The clinical expert confirmed that the key subpopulations 0 to 4 reflected NHS practice. The clinical expert explained that standard practice in England often involves patients starting therapy with a conventional laxative, which will often remain as part of the treatment regimen in both pure opioid-induced constipation and mixed aetiology constipation. When there is a poor response, a PAMORA would be considered in addition to the conventional laxative, and response to therapy would be monitored. The experience of the clinical expert indicated varying NHS practice, and limited use of the NICE technology appraisal guidance on naloxegol for treating opioid-induced constipation. The committee agreed that subpopulations 0 to 4 reflected naldemedine in clinical practice and were relevant for decision making.

3.5 The company submission did not include any direct evidence comparing naldemedine with any of the active comparators (naloxegol and methylnaltrexone). It included the results from an indirect treatment comparison comparing naloxegol with naldemedine (relative risk [RR] 0.79 [95% CI 0.63 to 0.99]) which informed subpopulation 3, based on an independent publication by Luthra et al. (2018). Also, the company included the results from an indirect treatment comparison comparing methylnaltrexone with naldemedine (RR 0.88 [95% CI 0.71, 1.06]). This informed subpopulation 4, based on the COMPOSE-4 trial and a randomised controlled trial by Bull et al. (2015). The company did not provide the methods used to combine the data from the trials in the indirect treatment comparison for subpopulation 4 after technical engagement. The company also highlighted that they did not have the input data for the indirect treatment comparison used to inform subpopulation 3. Therefore, the ERG was unable to assess the appropriateness of the indirect treatment comparison analyses or verify the results. After technical engagement, the ERG did several probabilistic sensitivity analyses and concluded that the uncertainty in the indirect treatment comparisons were unlikely to have a large effect on the cost-effectiveness results. For subpopulation 4, the ERG noted that methylnaltrexone is much more expensive than naldemedine. So, even if methylnaltrexone was much more effective, naldemedine would still be cost effective. The clinical expert noted that as subpopulations 3 and 4 did not include a direct comparison of these PAMORAs with naldemedine, it was difficult to determine whether there was a true difference between treatments. The committee concluded that any uncertainty was likely to have a small effect on the cost-effectiveness results for these subpopulations. It therefore considered that the indirect treatment comparisons for subpopulations 3 and 4 were relevant for decision making.

3.6 The company's economic model structure was based on the model considered in NICE's technology appraisal guidance on naloxegol. This consisted of a decision-tree structure for the first cycle followed by a Markov-structure from the second cycle onwards. Patients enter the Markov model at either opioid-induced constipation or non-opioid-induced constipation (when having treatment) health states, with a cycle length of 4 weeks and time horizon of up to 5 years. The company made several structural assumptions in their economic model, based on the NICE technology appraisal guidance, including for stopping treatment. Patients were assumed to stop treatment with naldemedine if their constipation had not responded by week 4 or had responded but they then experienced a reoccurrence of opioid-induced constipation. After stopping treatment, people whose constipation had not responded were assumed to not resume treatment across the 5-year time horizon of the economic model. The committee discussed loss of treatment response and the clinical likelihood of having only 1 possibility of response to naldemedine. The clinical expert explained that patients with pure opioid-induced constipation often develop mixed aetiology constipation, meaning response to a PAMORA may reduce. However, the clinical expert explained that for people with pure opioid-induced constipation, a PAMORA should not stop working and people should not develop a tolerance. Any loss of efficacy is normally because of a change in the patient's underlying condition rather than because of the PAMORA itself. The committee discussed the effect of assuming that treatment would be stopped on the estimates of cost effectiveness. It noted that the company had modelled naldemedine across various time horizons between 1 and 5 years. For most subpopulations, the incremental cost-effectiveness ratios (ICERs) increased slightly with a shorter time horizon. The ERG also noted that their clinical expert confirmed the appropriateness of assuming that treatment would be stopped for people whose constipation does not respond, or those who lost response. The committee recognised that stopping treatment for constipation that stops responding to naldemedine is plausible in clinical practice and is an appropriate assumption for the model.

3.7 The company submission included the probabilities for loss of treatment response to naldemedine, which were based on extrapolated time-to-event data from the relevant trials (see section 3.4, table 1). The company did not explore the clinical plausibility of their preferred parametric curves to model loss of treatment response at the clarification stage or after technical engagement. Instead, it highlighted that for all subpopulations, the choice of survival distribution has a minimal effect on the ICERs for each subpopulation. The ERG agreed with the choice of parametric curve in the company submission for subpopulations 1 to 4 but concluded that the Gompertz model was more appropriate for subpopulation 0. This was based on clinical opinion, which suggests that loss of response is likely to plateau at a certain level. The committee was aware that the choice of the curve has a minimal effect on the ICERs for all the subpopulations. It agreed that, while the clinical plausibility of the time-to-event curves is not known and that it would have been helpful for the company to provide this information, the effect on the ICERs is likely to be small.

3.8 The EQ-5D was not used in the COMPOSE trials, and so utility values from NICE's technology appraisal guidance on naloxegol were used. The company used treatment-specific utilities for the non-opioid-induced constipation (when having treatment) health state in the base case. The ERG noted that each health state should be homogeneous enough that the utility does not differ between different treatments. Therefore, it would have preferred a refined Markov model to which health state-specific utility values could be applied. The ERG's clinical expert did not expect differences in quality of life between people having naldemedine or naloxegol. The committee was aware that the ICER was sensitive to assuming treatment-specific utilities. Using health state-specific utilities increased the company's base case ICERs for subpopulations 0, 1 and 2 to £28,131, £27,484 and £15,020 per quality-adjusted life year (QALY) gained, respectively. The ERG noted that while it was not ideal to use treatment-specific utilities, the non-opioid-induced constipation (when having treatment) health state was probably quite heterogeneous in terms of spontaneous bowel movements. The committee agreed that using treatment-specific utilities was reasonable based on the approach in the technology appraisal guidance on naloxegol, and on the clinical expert opinion that naldemedine would improve a range of opioid-induced side effects, in addition to increases in spontaneous bowel movements seen in the COMPOSE trials. The committee noted that the company's model may not have captured these additional health benefits of naldemedine, and therefore accepted the use of treatment-specific utilities in the economic model.

3.9 NICE's guide to the methods of technology appraisal notes that above a most plausible 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. Because of the uncertainty in the indirect treatment comparisons and the impact on the choice of utility values, the committee agreed that an acceptable ICER would be below £20,000 per QALY gained. The committee recognised that the company's cost-effectiveness estimates for naldemedine using treatment-specific utility values were below £20,000 per QALY gained for all subpopulations (see table 2 below) and considered this to be a cost-effective use of NHS resources.

The ICERs in table 2 have been calculated using incremental costs and QALYs from the company's economic model.

The committee noted that the use of health state-specific utilities increased the ICERs for some of the subpopulations above this range, but that these were still under £30,000 per QALY gained. The committee was also persuaded that using health state-specific utilities did not capture all the broader benefits of treatment with naldemedine as highlighted by the clinical expert. If these were taken into account, the ICERs were likely to be under £20,000 per QALY gained. The committee was reassured by the results of the ERG's probabilistic sensitivity analysis for subpopulation 1. This indicated that naldemedine had probabilities of being cost effective of 74.8% and 86.3% at £20,000 and £30,000 per QALY gained, respectively. The committee agreed that treatment with naldemedine will likely result in an ICER below £20,000 per QALY gained compared with the relevant comparators for all the subpopulations.

3.10 No equality or social value judgement issues were identified.

3.11 The company considered naldemedine to be innovative because of its permanent binding capacity and higher receptor affinity compared with other PAMORAs. The committee agreed that these were important benefits of naldemedine. But, it concluded that it had not been presented with evidence of any additional benefits that could not be captured in the QALYs.