Amgen Response to ICER Plaque Psoriasis Draft Evidence Report

Amgen is a science-based company committed to developing and delivering innovative medicines. Our mission is to serve patients. We appreciate the opportunity to comment on the ICER Draft Report "Targeted Immunomodulators for the Treatment of Moderate-to-Severe Plaque Psoriasis: Effectiveness and Value."

Moderate-to-severe plaque psoriasis (PsO) is a chronic, systemic inflammatory disease associated with widespread skin involvement, significant comorbidities, and crippling physical, economic, emotional, and social consequences that accumulate over the course of a patient's life.¹ Systemic biologic treatments are often the only effective option for patients with this hard-to-treat disease.² Biologic PsO treatments have transformed the outcome of this disease in life-altering ways that short-term trials fail to capture.

While we disagree with ICER's use of cost-effectiveness thresholds and inadequate one-size-fits-all modeling methods, the findings described within the ICER Draft Report clearly demonstrate that all biologic PsO treatments, including Enbrel^® (etanercept), are good value. All targeted therapies largely fall within ICER's willingness-to-pay threshold of $100K to $150K per quality adjusted life year (QALY), and all were well under the $150K/QALY threshold after a 20% discount (a closer approximation of some products' true market price). The demonstrated value of all biologics for PsO reinforces the importance of preserving treatment choice for the patient as specific disease characteristics, clinical expertise/judgement, and patient preferences affect treatment choice.

After a careful review of the ICER Draft Report, we found the following issues warrant attention:

• The analysis lacks transparency and is difficult to interpret; the studies and study arms included in the NMA for determination of Psoriasis Area and Severity Index (PASI) responses for each product need to be explicitly identified with a clearly defined selection process.
• The phase 3b LIBERATE study comparisons between apremilast and etanercept are based on an unpublished noncomparative design and should not be used as a direct comparison for determining clinical equivalency.
• Model inputs for nontargeted agents should be based on current practice patterns and costs; ICER's use of 2003 data is inappropriate when current data are available and biases the analyses against targeted biologics.
• A wholesale acquisition cost (WAC) price with a single across-the-board discount does not reflect true patient and payer costs associated with PsO treatments; alternative price sources, such as Average Sales Price (ASP), should be used in the primary analysis.
• Administrative costs need to be corrected to reflect that typically only the first dose of self-administered products, such as etanercept, is administered in an office setting. 
• The uncertainty in PASI measures limits any justifiable differentiation in value among biologics; hence, the Draft Report should value the biologics similarly.
• Psoriatic arthritis (PsA) is a significant joint and skin comorbidity in patients with moderate-to-severe PsO that needs to be incorporated into ICER's analysis and model.
• Patient-centric outcomes, such as long-term benefits of targeted treatments, should be integrated into the economic analysis.

The analysis lacks transparency and is difficult to interpret; the studies and study arms included in the NMA for determination of PASI responses for each product need to be explicitly identified with a clearly defined selection process.

ICER should clarify trial inclusion criteria, including how the NMA studies were selected for inclusion and how between-trial heterogeneity was accounted for. Additional details regarding how QALYs were derived from PASI efficacy inputs are also needed.

The cornerstone of the model's comparative clinical effectiveness and value sections is the NMA of PASI response. The NMA methods should clearly state and provide references for the studies selected and include an assessment of effect-modifiers that is in line with accepted guidelines.^3,4

An assessment of the studies in the NMA and their fit with the analysis population should be included in the report. The NMA methods described in Appendix G in the Draft Report are unclear and fail to provide information on the rationale for the studies selected to derive PASI responses. To have valid treatment-effect estimates, the individual studies included in the NMA need to align with the population of interest for which the cost-effectiveness analysis is performed. It is unclear which randomized controlled trials (RCTs) were included in specific analyses and why only some of the treatment regimens from the RCTs were selected. In attempting to identify etanercept studies, Table 4 of the Draft Report lists 7 key trials for etanercept (p28), whereas, up to 12 trials with etanercept are identified elsewhere in the Draft Report. It is challenging to evaluate results of the NMA without specific references and clear methods.

Transparency in the cost modeling is necessary to enable proper evaluation of the cost-effectiveness model and the life-altering benefits derived from targeted PsO treatments. There was no transparent assessment of whether the trials in the NMA are sufficiently similar regarding effect-modifiers (e.g., distribution of prior biologic use or baseline PASI) or timing of endpoint assessments. This information is essential to determine whether the estimate of treatment effects generated by the NMA are sufficiently robust to be used in the cost-effectiveness model and to ensure clinically plausible and interpretable incremental cost-effectiveness ratios.

The Phase 3b LIBERATE study comparisons between apremilast and etanercept are based on an unpublished noncomparative design and should not be used as a direct comparison for determining clinical equivalency

The report includes a LIBERATE study comparison between apremilast and etanercept that is flawed and should not be included in this analysis. Fundamentally, this study has a noncomparative design and was not powered to compare apremilast and etanercept.^5,6 Notably, the etanercept dose used in the LIBERATE study was not the labeled starting dose. Instead, the study used the etanercept 50 mg weekly maintenance dose, which biases the comparison.

The ICER Draft Report makes eight different comparative statements between etanercept and apremilast using the LIBERATE study, including almost 20% of the Comparative Clinical Effectiveness Summary section. This section, intended to describe "head-to-head" studies, states the evidence is judged insufficient between etanercept and apremilast.⁷ Despite this acknowledgement, the section includes a long discussion of this study and indirect evidence between apremilast and etanercept before concluding that the products are "functionally equivalent for all other comparisons." The "equivalence" of apremilast and etanercept is based on two highly tenuous statements in the Comparative Clinical Effectiveness Summary: (1) the conclusion of "no statistically-significant difference" based on a post hoc analysis of the poorly designed LIBERATE study and (2) that "both agents are consistently ranked together in the network meta-analysis" despite a statistical difference between apremilast and etanercept in the ICER NMA. In making the second statement, ICER undermines not only the NMA evidence that suggests differences exist between apremilast and appropriately-dosed etanercept, but also the entire NMA.

Given these issues and others detailed in Appendix A of this response, Amgen requests that this study and comparative claims be removed from the ICER Draft Report and any analyses including these data be rerun. Amgen also requests the deletion of Voting Question 5 "Is the evidence adequate to demonstrate that apremilast provides equivalent or greater net health benefit than etanercept?" as Question 6 adequately captures this query.

Model inputs for nontargeted agents should be based on current practice patterns and costs; ICER's use of 2003 data is inappropriate when current data are available and biases the analyses against targeted biologics.

The nontargeted therapy regimens and their costs are described vaguely in the Comparative Value section of the Draft Report. The acquisition costs provided are most likely a low estimate of their true costs;⁸ the nontargeted therapy regimen costs should represent current US practice in patients with moderate to severe PsO. The current nontargeted therapies available to patients include multiple oral immunomodulators, systemic retinoids, phototherapy with and without chemotherapy, and a range of topical treatments with several different mechanisms of action. Some of these therapies have recently been made available to patients as novel agents or improved dosage forms. The nontargeted therapy and costs description is from a 2003 database analysis,⁹ and does not include these newer therapies and dosage forms. This analysis also notes the limitation of not capturing all costs involved with these treatments, which would underestimate the total cost of this intervention that serves as the primary comparator for the cost-effectiveness analysis.

Additionally, the cost of the nontargeted therapy is based on a price from the pre-biologic practice era in 2003, inflated to 2016 costs using inflation factors that were not included in the report. While the nontargeted therapy sensitivity analysis of the cost-effectiveness results attempts to address the low costs and outdated data, the large variation in results of the sensitivity analysis suggests the need for a more accurate portrayal and costing of these treatments based on today's standards. ICER needs to use updated nontargeted therapy costs based on true 2016 utilization and costs. ICER should survey databases, practitioners, and patients to best determine the prescribed therapies that are acceptable to patients with moderate to severe PsO. These treatments should be acceptable for use by patients over the 10-year time horizon since the nontargeted therapy arm assumes a benefit over this period. After establishing the new nontargeted therapy regimen, its cost should be derived to represent an accurate value comparison.

The wholesale acquisition cost (WAC) prices with a single across-the-board discount does not reflect true patient and payer costs associated with PsO treatments; alternative price sources, such as Average Sales Price (ASP), should be used in the primary analysis.

The use of discounted drug costs would significantly improve the accuracy of the cost-effectiveness model. Given that the discounted cost represents a closer approximation of some targeted PsO agents' true market price, it should be the primary economic analysis. The WAC is just one data point in the discussion of price and should be as up-to-date as possible before discounts are applied. While WAC is the list price to wholesalers, usually wholesalers and other purchasers of drugs (such as pharmacies, hospitals, payers, and physician groups) negotiate considerable discounts and rebates for the products they purchase. These discounts and rebates drive down drug prices and are not reflected in the WAC. The publicly available ASP used in Medicare Part B is one source that can be used to estimate the drug discounts.¹⁰ For example, the ASP of etanercept is 74.5% of the WAC. With discounts varying across drugs, applying a single discount rate does not accurately reflect prices paid in the U.S. ICER should seek to understand the differences in price discounting and refine the price discounts to achieve accurate prices for each drug. 

The economic analysis has a 10-year time horizon, but only accounts for the time discounting of costs. During this time horizon and over a product lifecycle, drug prices are dynamic. The price at a single point in time only answers a static question of value today. A good economic evaluation of a product would understand the value across time from launch of a product through biosimilar entry to full replacement by more efficacious treatment options. The economic model, at a minimum, needs to include the entry of biosimilars in the market and use appropriate costs in those years. ICER should anticipate the changing treatment costs and incorporate them into the model to more appropriately value these interventions.

Administrative costs need to be corrected to reflect that typically only the first dose of self-administered products, such as etanercept, is administered in an office setting. 

ICER makes a series of assumptions around biologic administration that are inconsistent with common practice. The Enbrel U.S. Prescribing Information instructs that the first injection should be performed under the supervision of a qualified healthcare professional. Patients rarely return to the office for repeat injections and generally self-administer etanercept utilizing the product's Instructions For Use; educational material and services are also provided by the manufacturer to help with self-administered injections (i.e. the Amgen Enbrel Support^TM Nurse Partners^TM program). Because dermatologists rarely administer injections to patients, the assumption that all injections during the induction period occur in the office is inappropriate for etanercept. Amgen recommends adjusting the assumption from all etanercept-induction injections performed in the office to one etanercept injection performed in the office.

The uncertainty in PASI measures limits any justifiable differentiation in value among biologics; hence, the Draft Report should value the biologics similarly.

Given the complexity of psoriatic disease, we appreciate ICER's statement¹¹ that "This uncertainty hinders our understanding of the relative effectiveness of these agents." Uncertainty is seen predominantly in the PASI measurements, especially when body surface area (BSA) involvement is low; PASI is a relatively insensitive measure in patients with low BSA involvement. Also, patients can often achieve satisfactory responses based on their signs and symptoms that are not adequately captured by PASI scores. As a result, use of PASI scores translated into utilities further expands the uncertainty. Additional examples of uncertainty in the costs were noted previously with market-based drug costs and administration fees. While the sensitivity analysis attempts to explain some of the uncertainty, it suggests that these biologics are similarly valued. The report should acknowledge the uncertainty and conclude these drugs have similarly good value and are cost effective.

PsA is a significant joint and skin comorbidity in patients with moderate-to-severe PsO that needs to be incorporated into ICER's PsO analysis and model.

ICER missed a significant opportunity by not addressing PsA, a common PsO comorbidity that affects up to 30% of patients with moderate to severe PsO. In patients with both PsO and PsA, up to $2184 additional costs are incurred annually.¹² PsA can contribute to differential effects on QALYs and costs as the targeted therapies have variable effectiveness for treating PsA. Given these differences and the high prevalence of PsA in patients with PsO, PsA as a comorbidity should be included in the ICER PsO analysis.

Patient-centric outcomes, such as long-term benefits of targeted treatments, should be integrated into the economic analysis

The report qualitatively addresses key patient issues unique to moderate-to-severe PsO in Section 2.4:  Insights Gained from Discussions with Patients and Patient Groups of the Draft Report. The economic analysis should quantify these insights and the treatment benefits of avoiding the long-term cumulative economic, emotional, and social consequences of PsO. Short-term PsO clinical trials capture neither these cumulative life impairments nor the long-term benefits of therapeutic interventions. In order to represent the long-term PsO patient experience identified in the report in the economic model, ICER should integrate these concepts into the model and further show the value of these lifelong therapies beyond the standard clinical trial time horizon in its cost‑effectiveness analysis.

As patients can often also achieve satisfactory responses based on their signs and symptoms that are not adequately captured by PASI scores, ICER should consider adjusting short-term trial-based utility with a long-term patient-based utility augmentation factor to test what omitting longer-term outcomes means for the base model. For example, one could report in a sensitivity analysis the costs per QALY where the long-term utility was 20%, 40% or even 50% higher in long-term use than that captured in short-term trial metrics.

Conclusion

The ICER Draft Report demonstrates good value at market prices for all PsO treatments in this analysis despite using the one-size-fits-all economic model that relies too heavily on short‑term studies and lacks patient-specific impacts. ICER has a responsibility to provide the evidence, costs, and patient perspective in their NMA in a highly transparent and credible manner. Development of a proper economic model that accounts for the significant physically and emotionally life-altering patient factors associated with PsO and reflects U.S. culture and practice patterns would best represent the value of PsO treatments in the U.S. Despite these limitations, the ICER analysis reinforces the importance of preserving patient treatment choice across all PsO treatments based on specific disease characteristics, clinical expertise and judgement, and patient preference.

References

Appendix A

ICER should eliminate or significantly reduce the extensive assessment of the LIBERATE study comparing apremilast and etanercept due to the noncomparative study design and unpublished sources. The ICER Draft Report noted this was grey literature and has cited an abstract recently presented at the 2016 American Academy of Dermatology meeting (Reich AAD 2016-cited in ICER references as #65). The report further makes comparative statements about this study in 8 different places in the "Comparative Clinical Effectiveness" section including almost 20% of the "Comparative Clinical Effectiveness Conclusion" language. This conclusion section has a long discussion about this study and the indirect evidence between apremilast and etanercept, and concludes that the products are "functionally equivalent for all other comparisons" within the Draft Report. There are a number of problems with this conclusion:

• The purpose of this Phase 3b Safety and Efficacy Study of Apremilast to Treat Moderate to Severe Plaque Psoriasis is to test the clinical effectiveness and safety of apremilast compared with placebo as well as etanercept compared with placebo in the same group of patients with moderate-to-severe plaque PsO. (ClinicalTrials.gov NTC01690299). An etanercept comparison is not an objective of this study as it was meant to be an apremilast vs placebo comparison. In fact, none of the primary or secondary outcome measures listed in the clinicaltrials.gov file list a comparison between apremilast and etanercept, only comparisons of apremilast with placebo.
• The grey literature abstract from AAD specifically states "This study was not powered for APR vs ETN comparisons." This statement in a published source should make ICER pause at the credibility of any statements in the report regarding comparative results.
• The study lists, and ICER has noted, that the etanercept dose used in the LIBERATE study was not the labeled starting dose. Instead, the study used the etanercept 50 mg maintenance dose, which further makes this comparison biased. Additionally, while a 50 mg syringe dosage form is available, the study used 2 - 25mg syringes unduly increasing the patient burden and likely affecting the quality-of-life assessments.
• In the conclusion of Section 4, ICER states "While the addition of the indirect evidence suggests an incremental benefit for etanercept, we feel that indirect findings can only confirm or downgrade certainty in evidence ratings." The difference between etanercept and apremilast in the indirect NMA for PASI 75 is significant as shown in Table G3. Dismissing this significant difference between these two drugs would invalidate this table for all drugs, and should make ICER conclude they are all equivalent, which clearly is not the case. This insignificant finding in the indirect NMA may be due to the appropriate etanercept dose being evaluated in the NMA studies compared to the LIBERATE study. The arbitrary nature of the decision to assume the indirect comparison is not valid negates the entire credibility of the indirect comparison and its use in the economic model.

Given these study design issues, the unpublished data, and improper interpretation of the indirect comparison, Amgen requests ICER remove this study from the Draft Report, including the following: