Summary of the Manufacturer’s Pharmacoeconomic Submission

The manufacturer submitted a cost-utility analysis (CUA) comparing dupilumab plus standard of care (SOC) compared with SOC alone in adult patients with moderate-to-severe atopic dermatitis whose disease is not adequately controlled with topical prescription therapies.² Standard of care was defined as mid-potency topical corticosteroids (TCS) or topical calcineurin inhibitors (TCIs). The model used a lifetime horizon (63 years) from the perspective of the publicly funded health care payer, with costs and clinical outcomes (QALYs) discounted at 1.5% per annum.² The model reflected a population that had baseline characteristics similar to the LIBERTY AD CHRONOS trial (58.7% males; average age: 37.6).³ The model structure included a short-term (one-year) phase for the 16- and 52-week assessment in the LIBERTY AD CHRONOS trial, and a lifetime model for the maintenance phase. The short-term phase was based on a decision tree that modelled treatment response at 16 weeks and 52 weeks. The maintenance phase was based on a Markov state–transition model with annual cycles and consisted of three health states: on maintenance treatment with dupilumab plus SOC, on SOC, and death.²

All patients started in the trial phase as a nonresponder and, at 16 weeks, patients were evaluated for treatment response. In the manufacturer’s base-case analysis, responders were defined as those who do not use a systemic immunosuppressant (such as methotrexate) and achieve an Investigator’s Global Assessment (IGA) end point of 0 or 1, with a reduction from baseline of two or more points. In the dupilumab arm, those who responded to treatment stayed on dupilumab until 52 weeks, at which point nonresponders discontinued dupilumab, as reported in the LIBERTY AD CHRONOS trial³ All nonresponders were treated with SOC alone. The proportion of patients entering each health state in the Markov model was therefore based on the patients’ treatment response at 16 weeks and 52 weeks. In the maintenance phase, patients may discontinue dupilumab and transition to SOC. In the SOC-alone arm, it was assumed that all patients remained on SOC during the maintenance phase.²

The manufacturer assumed no mortality effect from treatment; all treatment benefits were captured by an improvement in health-related quality of life (HRQoL).² Treatment-specific utilities were estimated for dupilumab plus SOC and SOC alone, rather than employing state-specific utilities. The regression was based on data from the LIBERTY AD CHRONOS trial using a mixed-model regression analysis. A forward-selection process was used to determine best fit and resulted in the following regression covariates: age, male, baseline EuroQol 5-Dimensions questionnaire (EQ-5D), total Eczema Area Severity Index (EASI) score, the weekly average of peak daily pruritus, EASI–pruritus interaction, and a treatment dummy.² The model assumed no decrement in utility due to adverse events.²

The model included acquisition costs of dupilumab, medical costs relating to responder status, and the costs of treating adverse events.² Drug costs were obtained from the manufacturer based on a 600 mg loading dose and a 300 mg dose every two weeks thereafter. It was assumed that patients would not be fully compliant to dupilumab from 16 weeks onward, resulting in lower drug-treatment costs.² SOC was not costed to avoid issues of double counting, given that health state–specific costs were present in the model based on responder status.² The manufacturer assumed that the administration cost, specifically, the cost of training patients to administer subcutaneous injections, would be covered by the manufacturer through a patient support program.² Other medical costs were attributed to responders and nonresponders based on unpublished studies undertaken by the manufacturer.^4,5 The costs of treatment-specific adverse events included a one-time injection-site reaction and other adverse events (e.g., rates of allergic conjunctivitis, infectious conjunctivitis, and oral herpes) were modelled by per-cycle incidence rates.²

Manufacturer’s Base Case

Dupilumab plus SOC was found by the manufacturer to be $112,362 more expensive than SOC alone. The estimated benefit of dupilumab plus SOC was an additional 1.25 QALYs over 64 years.² Table 2 shows the contribution of the different sources of cost to the overall total costs (the results are deterministic, as they were not reported for the probabilistic analysis). In the probabilistic base case, the incremental cost-effectiveness of using dupilumab plus SOC compared with SOC alone is $89,723 per additional QALY gained. Based on the manufacturer’s probabilistic sensitivity analysis, SOC had 99.9% probability of being the most likely cost-effective option at a cost-effectiveness threshold of $50,000 per QALY.²

Table 2

Summary of Results of the Manufacturer’s Base Case.

Summary of Manufacturer’s Sensitivity Analyses

Uncertainty was addressed using a Monte Carlo simulation, one-way deterministic sensitivity analyses, and scenario analyses. Based on the manufacturer’s one-way deterministic sensitivity analyses, the results were most sensitive to compliance to dupilumab during the maintenance phase, baseline utility weight, and dupilumab drug costs.²

Scenario analyses were used to consider a broader societal perspective, the effect of different measurements to define treatment response, different approaches to calculate treatment-specific utility values, shorter time horizons, and using the SOLO trial as an alternative data source.² The model results were most sensitive to the time horizon used, with shorter time horizons resulting in a larger ICUR. Reduction of the time horizon to 16 weeks resulted in an ICUR of $923,203 per QALY and, with a one-year time horizon to reflect the LIBERTY AD CHRONOS trial period, the ICUR was $467,208 per QALY. Use of different approaches to define treatment-specific utility weights resulted in a range of ICURs, with the approach taken in the manufacturer’s base case (regression, last observation carried forward [LOCF]) presenting the lowest ICUR ($89,723), increasing to $143,890 per QALY when utility weights were based on the observed change in baseline from the LIBERTY AD CHRONOS trial without using LOCF. An analysis that included productivity losses and out-of-pocket expenses reduced the ICUR to $62,279 per QALY. ▬▬▬▬▬ compared with the CHRONOS trial population, which only recruited patients who were inadequately controlled with topical therapies. However, the SOLO trial restricted treatment to dupilumab monotherapy, as patients were not permitted to use supportive therapies. Results from the SOLO trial resulted in an ICUR of $95,639 per QALY.²

The results of these analyses suggest that parameters pertaining to time (i.e., horizon and long-term extrapolation of clinical benefits) and utility weights had the largest impact on the ICUR.

Limitations of Manufacturer’s Submission

1)

Missing all clinically relevant comparators: Current SOC in Canada includes the use of systemic immunosuppressants such as methotrexate and cyclosporine. The benefits and costs of using these treatments have not been included in the model, although the LIBERTY AD CHRONOS trial did allow the use of these treatments in both treatment arms. The inclusion of systemic immunosuppressants in the model would likely increase the ICUR for dupilumab plus SOC, i.e., making dupilumab plus SOC less cost-effective. These treatments are much less expensive than dupilumab and are used widely to improve the health of patients with atopic dermatitis and other dermatological conditions. The CADTH Common Drug Review (CDR) was unable to conduct a reanalysis to assess the comparative cost-effectiveness of dupilumab plus SOC compared with these comparators, given the lack of comparative clinical effectiveness data. Annual or treatment-cycle costs of relevant therapies for atopic dermatitis are presented in Appendix 1.

2)

Modelling the impact of compliance with dupilumab: The manufacturer inconsistently applied the impact of compliance into the model. In the manufacturer’s submitted model, it was assumed that 81.7% of responders would remain compliant to dupilumab treatment from week 16 onwards.² This was implemented in the model by decreasing the costs of dupilumab with no effects on patient outcomes, i.e., treatment utility or response.² The approach taken by the manufacturer is optimistic as, although it may provide a more realistic cost estimate, it likely overestimates the absolute benefit of dupilumab. To match the effect of decreased compliance on costs and treatment effect, CDR used the reported compliance of 96.89% from the LIBERTY AD CHRONOS trial

3)

Estimates of treatment-specific utility values: As per current guidelines for the conduct of economic evaluations,⁶ utilities should reflect the health states within the model and not be specific to treatment. No justification was provided in support of the use of treatment-specific utilities.

Treatment-specific utility values were derived from a mixed-model regression analysis.² It was not clear why some variables were excluded from the forward-selection process, in particular, why the Dermatology Life Quality Index (DLQI) was not tested in the model, and how the utility weights were estimated (i.e., which values for the Pruritus Numerical Rating Scale and EASI were multiplied by the model coefficients to estimate the utility for each treatment). A request was made to the manufacturer for further information on how the regression model was estimated; the data provided were insufficient to assess whether the approach could have introduced bias to the estimates derived. Given concerns with the potential biases introduced by the regression model, the CDR reanalysis used the reported change in utilities, from baseline to 16 weeks as observed in the LIBERTY AD CHRONOS trial within each treatment arm, to determine the utility values at week 16.^2,3

Furthermore, the manufacturer’s model took a treatment-specific approach to estimate and apply utility weights after week 16. In the dupilumab plus SOC arm, utility values were based on whether a patient responded to treatment, with a higher utility weight applied to responders (0.9029) compared with nonresponders (0.8175); whereas in the SOC arm, the utility values reflected the week 16 utility value estimated for all patients on SOC (0.8175).² As there is no justification for different approaches to model the utility values after 16 weeks of treatment, a CDR reanalysis was conducted in which identical methods were used to estimate and model utility for SOC after 16 weeks. At week 16, those on SOC were separated into responders (probability = 12.4%^2,3) or nonresponders and, similarly, utility values were not only treatment-specific but also reflected responder status. The utility of responders on SOC was adjusted as follows:

• Utility of responders on SOC = utility of all patients on SOC at week 16 + (utility of responders on dupilumab at week 16 minus utility of all patients on dupilumab at week 16)
• Utility of nonresponders on SOC = (utility of all patients on SOC at week 16 minus utility of responders on SOC × probability of response on SOC at week 16) ÷ (1 minus probability of response on SOC at week 16)

4)

Durability of response beyond trial duration: In the manufacturer’s model, treatment-specific assumptions were made regarding the persistence of the treatment response.² For responders on SOC, treatment response would be lost at 52 weeks (i.e., treatment waning), as — regardless of responder status — all patients would return to their baseline utility value (0.6400), adjusted by age. The manufacturer justified this assumption based on feedback from their key opinion leaders that the effects of treatment on SOC last four to six weeks.² However, the clinical expert consulted as part of this review disagreed with this assumption, as it is expected that patients responding to SOC would have improved quality of life over those not responding to treatment. Furthermore, the LIBERTY AD CHRONOS trial reported that 45.4% of the all-observed SOC population would still meet the EASI-50 (50% or greater improvement in EASI from baseline) criteria at 52 weeks.² This suggests that not all patients should revert to the baseline utility at 52 weeks. The manufacturer assumed no waning effect for patients on dupilumab plus SOC.² Patients who responded at 16 weeks to dupilumab plus SOC were assumed to have a utility weight reflective of treatment responders (0.9029), adjusted by age, for the rest of their life unless treatment was discontinued. This does not align with the LIBERTY AD CHRONOS trial findings that reported a reduction in the proportion of responders for dupilumab plus SOC between week 16 and week 52 (proportion of responders defined by EASI-50: 85.8% and 81.1% at week 16 and week 52 respectively; defined by EASI-75 (75% or greater improvement in EASI from baseline): 73.6% and 60.4% at week 16 and week 52, respectively).² The decline in effectiveness between 16 weeks and 52 weeks suggests that durability of treatment response continues to decline after 52 weeks. The manufacturer’s approach underestimates the absolute benefit of SOC while it overestimates the absolute benefit of dupilumab plus SOC.

Current guidelines for the conduct of economic evaluations state that it is not acceptable to assume that the relative effectiveness will be maintained for the duration of the intervention.⁶ No justification was provided to support the assumption of no treatment waning with dupilumab. According to the clinical expert consulted as part of this review, treatment waning is a realistic assumption. CDR’s reanalysis was informed by the opinion of the clinical expert consulted as part of this review. The clinical expert did not agree that those who continue to respond to SOC at one year would have a utility value identical to nonresponders. However, the clinical expert did state that a high proportion of patients who respond to dupilumab and stay on treatment would maintain their improved quality of life, and stated that this would also be true for those responding to SOC. In the CDR reanalysis, it was assumed that the utility weight of both dupilumab and SOC responders would decrease over time, but at the same rate (i.e., that utility value for responders would return to the baseline utility after 40 years of treatment).

5)

Annual discontinuation estimates for dupilumab: The manufacturer’s submitted model applied a discontinuation rate of 2.4% to treatment responders starting from the second year of treatment, which was based on a discontinuation rate for responders of between 16 weeks and 52 weeks in the LIBERTY AD CHRONOS trial.^2,3 This meant that, among patients who proceeded to maintenance therapy, 11% would have discontinued use of dupilumab after five years. This value does not align with those reported in the literature, in which the discontinuation rate of biologics in patients with psoriasis at five years has been shown to range from 25% to 89%.¹⁰ The manufacturer provided an alternative discontinuation rate for responders based on the findings from the SOLO trials (6.3%).^2,8,9 In the CDR reanalysis, an annual discontinuation rate of 6.3% would result in 28% of responders discontinuing after five years. Applying this rate better aligns with the discontinuation rate observed in patients with psoriasis on biologics.

6)

Full indication not addressed: The economic analysis did not adequately assess the cost-effectiveness of dupilumab in patients in which topical prescription therapies are not advisable. The clinical expert consulted for the CDR review noted this is likely to represent a small portion of patients who would be receiving this treatment.

CADTH Common Drug Review Reanalyses

The results of the CDR reanalysis are reported in Table 3. The reanalysis addressed the limitations identified above by:

• assuming a 96.89% compliance rate for dupilumab, for consistency
• adjusting treatment-specific utility values to reflect the utility change observed in the CHRONOS trial (based on LOCF)
• incorporating identical assumptions pertaining to waning effects for dupilumab and SOC (utility weights decline at a linear rate over 40 years)
• adjusting long-term clinical benefits for responders on SOC
• revising the annual discontinuation rate for dupilumab of 2.4% to 6.3% in order to better align with a discontinuation rate reported in a similar clinical area

Table 3

CDR Reanalysis of Limitations.

Compared with the manufacturer’s results, the CDR reanalysis reported lower expected costs for SOC, but higher expected costs for dupilumab plus SOC, while QALYs for dupilumab plus SOC and SOC alone were higher than reported by the manufacturer’s results. Under the CDR base case, the ICUR for dupilumab plus SOC was estimated to be $579,672 per QALY compared with SOC alone (Table 3).

A price-reduction analysis (Table 4) demonstrates that, in using the manufacturer’s base-case analysis, dupilumab would be considered cost-effective at $50,000 per QALY at a 40% price reduction. Using the CDR reanalysis, dupilumab would be cost-effective at $50,000 per QALY following a price reduction of 84% and, in order to be cost-effective at $100,000 per QALY, a price reduction of 75% would be required.

Table 4

CDR Reanalysis Price Reduction Scenarios.

Issues for Consideration

• As noted by the CDR clinical expert, there is no clear and objective definition of response. The clinical expert stated that the EASI score is more likely to be used to define response to treatment.
• Although dupilumab is indicated for use as a second-line drug in the treatment of moderate-to-severe atopic dermatitis following inadequate control with topical therapies, and as a first-line treatment in patients for whom topical therapies are not advisable, the clinical expert indicated it may, in fact, be used as a second- or third-line drug, after failing systemic therapy or phototherapy.¹¹ There is, however, no comparative clinical effectiveness data between dupilumab plus SOC compared with these therapies, which would inform the potential cost-effectiveness of dupilumab when compared against these alternative therapies.

Patient Input

Input was received by one patient group, the Eczema Society of Canada (ESC). The ESC reported many consequences of atopic dermatitis that affect patients’ quality of life, including: itching, pain, anxiety, depression, social isolation, productivity effects, poor self-esteem, and suicidal thoughts. Patients who participated in the clinical trial reported a reduction or elimination of flare-ups and itch. Their comments support the lower baseline utility value that was used in the model for moderate-to-severe atopic dermatitis. Although comments from patients who had experienced successful treatment with dupilumab suggest an improvement in quality of life compared with alternative treatments in terms of reducing disease severity, this was not adequately addressed in the manufacturer’s submitted economic model; treatment efficacy was driven solely by the dichotomous outcome of treatment response rather than explicitly modelling change in disease severity. The ESC also reported that caregivers also suffer sleep loss, anxiety, and depression, although the effects on the caregiver were not reported by the manufacturer or captured in the submitted economic model.

Conclusions

The key limitations were the assumptions on the sustained treatment effect of dupilumab plus SOC, the use of the regression analysis to estimate treatment-specific utility values, and the partial incorporation of the effects of poor compliance. The CDR reanalysis addressed the aforementioned limitations within the manufacturer’s economic analysis. In patients with moderate-to-severe atopic dermatitis whose disease is not adequately controlled with topical prescription therapies, the ICUR was estimated at $579,672 per QALY for dupilumab plus SOC compared with SOC alone.

The difference in incremental utilities was driven largely by the approach taken to model utilities and the assumptions around the waning of the treatment effect for both treatment arms in the model. The model was sensitive to both sets of assumptions.

A price reduction of 84% would be required for the ICUR of dupilumab plus SOC to fall below the $50,000 per QALY when compared with SOC alone. CDR was unable to assess the cost-effectiveness of dupilumab plus SOC compared with the alternative comparators that are presently used by patients with moderate-to-severe atopic dermatitis whose disease is not adequately controlled with topical prescription therapies. The magnitude of clinical benefit that dupilumab may offer compared with these alternative treatments is uncertain, given the lack of comparative effectiveness evidence.

CADTH notes that this model does not explicitly address the population of patients for whom topical therapies are not advisable, and the ICUR is presently unknown in this patient population.