https://orcid.org/0000-0001-5071-9257
Introduction
Owing to the high increase in drug prices and the need to reduce overall medical costs in the United States and Europe, payers and drug manufacturers are adopting value-based approaches for drug pricingCitation1,Citation2. This approach tends to consider higher prices for innovative drugs and lower prices for lower value alternatives. Value-based pricing requires a clear definition of value and the extent to which value can be used for pricing. If the value of a treatment remains to be defined, outcome-based pricing can be used alternatively. Outcome-based pricing considers the clinical value of a treatment in determining the payback guarantee if the drug fails to achieve the outcomes demonstrated by a drug manufacturer but also to reward the manufacturer if the drug achieves exceptional outcomes.
Outcome-based contracting is not a new method. In 1994, Merck offered a payback guarantee on finasteride (ProscarFootnotei)Citation3. In the US, from 2009 to the first quarter of 2018, drug manufacturers signed more than 39 such contracts with different payers in the United StatesCitation1. In Europe, the number of outcome-based contracts reached 150 in 2013Citation4. Outcome-based contracts are expected to increase by two- to three-fold between 2017 to 2022 in the US and European G5 countries (Germany, France, United Kungdom, Italy, Spain)Citation2.
These contracts are interchangeably termed value-based contracts and outcome-based contracts. However, these two terms should be distinguished, because value-based contracting is broader than outcome-based contracting. Kaltenboeck and Bach classified value-based contracting as a category of methods that includes outcome-based contracting and other types of contractingCitation5. In their definition, value-based contracts are based on different health technology assessments including cost-effectiveness and cost-utility analyses, indication-specific pricing, outcome-based pricing, mortgage pricing, and value-based insurance designCitation5.
In the US, in all 39 outcome-based contractsCitation1 drug manufacturers and payers chose certain health metrics as payback indicators to negotiate purchasing prices. For example, Harvard Pilgrim Health Care signed a contract with AstraZeneca to cover exenatide for diabetesCitation1. Both parties chose patient adherence as a payback indicatorCitation1. Prime Therapeutics agreed to cover empagliflizin (Boehringer Ingelheim) according to its efficacy and safety performanceCitation1. To date, outcome-based contracts have been designed in different formats and including different clinical measures to enable both the drug manufacturer and the payer to achieve their respective financial goals. Although these pricing contracts are considered outcome-based, they do not include all potentially applicable indicators of outcome. For example, of the 39 contracts mentioned above, 32 contracts included efficacy, 1 focused on adherence, 1 considered safety, and 1 combined both safety and efficacyCitation1. This also explains the variability in contracts that a manufacturer may propose for a given drug to different payers.
The specifications of outcome-based contracts in Europe tend to be unclearCitation6. A survey of oncology contracts revealed that in six European countries these contracts were financial and outcome-based contracts under the definition of managed entry agreementsCitation7. However, some contracts appeared to be efficacy-based. In the UK, GlaxoSmithKline agreed to provide rebates to the National Health Services (NHS) if pazopanib was inferior to its comparator unitinib in the treatment of advanced kidney cancerCitation6. In France, Celgene agreed to refund the cost of treatment for nonresponding multiple myeloma patients on pomalidomideCitation8.
The clinical measures used as indicators for estimating the magnitude of price elasticity in the payback scenarios of outcome-based contracts is controlled by each contracting party’s financial interests. The details of how each party aligns clinical outcomes with price are not disclosed and the interests of the payer and drug manufacturer might not converge.
Some operational challenges have been observed in outcome-based contracts. One example is the ability of payers and manufacturers to capture clinical data jointlyCitation2. Data collection for outcome-based contracting is not always performed bilaterally by both parties. In some settings, patient level data may be available to payers, as shown in contracts in Sweden and SpainCitation9–11; but limited in, for instance, in ItalyCitation12. Other factors should be taken into consideration as well such as, for instancepricing policiesCitation13,Citation14; patient turnover during the course and over the duration of a contractCitation15; orphan drug statusCitation16,Citation17; and need to implement adherence programsCitation18. These operational challenges should be discussed upfront between payers and manufacturers and opportunities like enabling data sharing or/and adopting third and independent parties to perform fair assessments should be explored.
We propose an independent platform for outcome-based contracting that aligns stakeholders’ interests; promotes inclusivity, information, and transparency; and benefits all stakeholders. Such unbiased (or at least, minimally biased) outcome-based contracting platforms should indeed serve all stakeholders. In first instance, they should enable fair market competition, sustainable price strategies, and access to treatment for those patients most likely to benefit from a new drug. In turn, such models should ensure appropriate and warranted sales volumes and patient access; supported by broad-based and inclusive stakeholder-independent scientific technologies. An independent platform that supports joint and equitable price negotiation should assure better coverage and patient access. It should stimulate payers and their providers under contract with value incentives to achieve more beneficial health outcomes and reduce the cost wastage that comes from mismatching treatments with patients, it should lower patients’ share of treatment costs.
Because methodological transparency is critical to outcome-based contracting, other factors should be ensured as well: comprehensive contracting, balancing of interests, and the use of trial data for ex ante and real-world data for ex post facto model development and improvement. Comprehensiveness can be ensured if all clinical outcomes are generalized to and integrated into the covered population. Balancing can be achieved by aligning value of treatment (as driven by its efficacy and safety performance) with its affordability (as driven by payer willingness-to-pay as well as patient willingness-to-accept). It is critical to find solutions that improve patient outcomes and reduce cost, while also maintaining incentives for innovation.
To date, outcome-based contracts typically have used real-world data to determine the value of drug and its price, in part because randomized clinical trials are typically conducted in “ideal” patients with limited co-morbidities. However, the lag time between drug approval and the availability of real-world data creates a period for opportunistic pricing to set a high starting price. Real-world effectiveness data take some years to accrue, but are critical for mitigating, upward or downward, our understanding of the effectiveness and safety of a drug in routine clinical. This could be avoided if an outcome-based contracting model were able to use clinical trial data early on until real-world data are available and both sources of data can be used in tandem.
While we are strong advocates of outcome-based contracting, there are issues of independence, stakeholder equity, transparency, timing relative to clinical development and regulatory approval, and inclusiveness of empirical evidence to address. This creates several research and innovation opportunities with significant potential impact on fair and equitable pricing, patient access, and continued innovation in pharmacotherapy. The Six Delta independent outcome-based contracting methodology and analytics platform described here aims to fill these gaps. In this (peer reviewed) special issue of the Journal of Medical Economics, we present the methods and statistical and economic underpinningsof the Six Delta platform and apply this in a proof-of-concept exercise. Our methodology is based on prediction and simulation models that, for the time being and until real-world patient-level data are available, do not require patient-level data. Despite the limitations of clinical trial data and this being a proof-of-concept exercise, we used published clinical trial data. In the future, it will be necessary to test the Six Delta platform using patient-level data to enhance accuracy and further evaluate its validity.
Overview of the six Delta platform
The pricing methodology described herein uses a price assessment approach comparing six differentiated dimensions (δ) in a comprehensive and integrated platform for outcome-based contracting that examines and reconciles price variations as a function of the six δ differentiators (hence the name Six Delta). Two-time horizons for assessing these 6 δ’s are considered, comprising 2 δ’s based on long-term assessments and 4 δ’s based on short-term assessments (). Individually, each of these 6 δ’s is designed to use different scenarios in creating price variations (also termed dispersions) at the dimensional level. The price variations in each δ can be simulated by the method of Monte Carlo Simulation (MCS) to generate a dimension-specific price (DSP), as well as a price that averages all dimensional prices (ADP). Alongside this initiative, proof-of-concept analyses are provided to apply the six pricing δ’s to the case of osimertinib for the treatment of non-small cell lung cancer (NSCLC) with epidermal growth factor receptor (EGFR) mutation. In this exercise, clinical trial data are extracted and modeled in the 6 δ’s to suggest a DSP and ADP for osimertinib. The suggested methods applied in the proof-of-concept assume random patterns of price variation because data from a clinical trial can be considered imperfect information for pricing and is best used for short term contracting. However, in the future, the proposed platform could also use real-world data, and the generated prices could be based on more predictable patterns of variation than the random ones used for clinical trial-based pricing.
Figure 1. Outcome-based price assessment framework. *DSPAdherence is subject to change based on real-world effectiveness results. In the framework, δ1 and δ2 were based on long-term assessment; δ3-δ6 were based on short-term assessment. DSP: dimension-specific price; ADP: average of all dimensional prices integrated and simulated by Monte Carlo Simulation.
δ1. Cost-effectiveness analysis and Cost-Utility analysis-based pricing
In this first dimension, cost-effectiveness analysis (CEA) and cost-utility analysis (CUA) are the basis for pricing a treatment. These analyses are integrated with methodologies to estimate the price of a drug based on probabilistic sensitivity analysis (PSA) estimates of the incremental cost-effectiveness ratio (PSA ICER) and incremental cost-utility ratio (PSA ICUR). This is done at three certainty levels (0%, turning point%, and 100%) on the cost-effectiveness acceptability curve (CEAC) for PSA CEA and PSA ICUR. To note, CEA and CUA are not new methods; rather, they are widely used to estimate the clinical and economic effectiveness of drugs. For this dimension, CEA and CUA with long-term time horizon are performed, estimating price scenarios at the three certainty levels. In this, CEA/CUA price scenarios are generated, averaged by MCS to estimate the DSPCEA/CUACitation19.
δ2. Willingness-to-pay-based pricing
Willingness-to-pay-based pricingCitation20 is an extension of dimension 1Citation19. Using base case CUA quality-adjusted life year (QALY) estimates from dimension 1, this dimension 2 considers four willingness-to-pay (WTP) scenarios based on per capita gross domestic product (GDP) per QALY gained: <1 × GDP, 1 × GDP, 3 × GDP, and >3 × GDP. Two market scenarios are considered. The monopolistic market refers to no treatment alternative(s) being available in the market and uses cost/QALY gained as metric. The competitive market, in which more than one drug option is available, applies the base case ICUR from δ1 as metric. This is complemented by one-way sensitivity analyses to generate predictive equations for price adjustment under the four WTP scenarios assumed and under the two market environments. In the end, WTP-based price scenarios are generated per contract, and averaged for MCS to estimate the DSPWTPCitation20.
δ3. Reference-based pricing
Reference-based pricingCitation21 may not be directly related to clinical outcomes. However, strategically, this dimension provides contextual pricing data by evaluating whether and how drug prices in the U.S. (home country) are aligned with prices in foreign countries that use clinical outcomes for price assessment. The value of reference-based pricing is an important third dimension from an economic point of view when integrated with the dimensions that consider the clinical value of drugs.
In our metrics, for the reference-based pricing dimension, we propose to use three economic indicators (adjusted for exchange rate) and two methods for reference pricing. The three economic indicators are: GDP per capita, purchasing power parity (PPP), and the percentage (%) of GDP spent on pharmaceuticals. The two methods for reference pricing are: average foreign prices and minimum foreign prices for a set of pre-specified countries.
As noted, the total number of reference-based pricing scenarios is five per each pairwise comparison (home country versus foreign country). For example: if we have one foreign country, we will have five price scenarios; if we have 2 foreign countries, we will have 10 price scenarios; and so on. In the end, all generated reference-based price scenarios are averaged for MCS to estimate the DSPReferenceCitation21.
δ4. Safety-based pricing
The fourth dimension, safety-based pricingCitation22, is based on the comparative safety of a given treatment. Assume that drug A is our treatment of interest and drug B is a comparator, and that adverse events associated with both treatments were evaluated in a clinical trial. In this trial, drug A and B were shown to have a respective risk of 10% and 5% for a specific adverse event. Assuming comparable sample sizes, the risk difference is 10% − 5% = 5%. Further, assume that the drug A manufacturer agrees to a payback to the payer if in the real-world setting of daily clinical practice with drug A the risk difference exceeds (or is predicted to exceed) 5%. The risk margin in excess of the prior risk difference is termed “undesirable risk “. This undesirable risk becomes a correction factor that is multiplied by the risk of an adverse event under consideration (i.e. the probability of the adverse event being considered) to estimate a corrected probability. The manufacturer of drug A will provide a payback at the magnitude of the corrected probability.
In this dimension, we aim to predict the correction factor for each adverse event by calculating the risk difference and by considering it a threshold. We then perform MCS for the probabilities of the adverse events being considered (with their 95% confidence interval) for treatment comparators within a probability distribution. MCS thus generates a probability of when the risk difference generated from MCS exceeds the threshold value. This probability is the undesirable risk and considered as correction factor. Then, we multiply this correction factor with the probability of the adverse event (as reported from clinical trial) to yield a corrected probability of the specific adverse event that could serve as a dimension metric.
Further, in our method, we consider that the payback is also conditioned by drug discontinuation due to one or more adverse events; that is, we multiply the yielded corrected probability with the probability of drug discontinuation due to specific adverse events. Note that, as per our methodology, if drug A was associated with more than one adverse event versus drug B, the payback must be based on the probability of at least one undesirable adverse event occurring and causing discontinuation.
This dimension includes a 4-step method: (1) pooling adverse events, standardization, and adjustment for contracting time; (2) estimating correction factors and corrected probabilities; (3) estimating the probability of at least one adverse event occurring and leading to drug discontinuation; and (4) estimating the payback percentage and its range, performing MCS, and deriving the DSPSafetyCitation22.
δ5. Risk of efficacy failure-based pricing
The fifth dimension, pricing based on the risk of efficacy failureCitation23 is itself a novel, sophisticated, and innovative methodology to detect the percentage of patients for whom the superior treatment (assuming there are at least two treatments in the market) may not result in additional clinical efficacy benefits compared to the standard of care (inferior treatment). In this case, the drug manufacturer would provide payback to patients for whom the treatment fails to be effective and does not exceed what could be achieved with the standard of care. To determine this percentage of non-responding patients, a 7-step method is applied: (1) identifying the efficacy endpoints to be used in an outcome-based contract; (2) extracting Kaplan-Meier curves from published evidence using a digitizing technique (unless patient level data are available); (3) selecting a distribution to fit the extracted data, based on the shape and scale functions generated by regression analyses and goodness-of-fit metrics such as the Akaike information criterion; (4) performing MCS to estimate risk of efficacy failure and this by using the mean, lower, and upper bound probabilities for the treatment of interest and its alternatives at a given time (contract duration); (5) estimating ranges for payback; (6) adjusting for inflation rates; and (7) performing MCS to estimate the DSPRisk of efficacy failure.
This dimension considers the percentage of price that should be paid back by the drug manufacturer to the drug payer. Technically, this percentage may not fully or accurately quantify the uncertainty generated by the analysis. We encourage using the term of “up to” the estimated percentage as a range for payback; specifically, from zero percent to the estimated percentCitation23. Thus, for each efficacy measure, two price bounds are developed: one using the lower bound of the percentage of superiority failure; and one using the upper bound of 0% price change. The lower and upper bounds are then averaged for MCS to derive the DSPsuperiority failureCitation23.
Importantly, the ranges for payback are validated by estimating the restricted mean survival time (RMST) from the area under the curve. The RMST is measured in limits of time and represents the accumulated survival time gained by the standard of care comparator. Thus, if there will be a payback to the payer, it will be at the RMST gained by the standard of care that can be considered as a failure for the drug of interest to gain. In our validation for the ranges assumed, the RMST-based payback is located within the ranges assumed in this dimensionCitation23.
δ6. Adherence-based pricing
The sixth and final dimension, adherence-based pricingCitation24, is a derived from dimension 5Citation23 as patient adherence to a medication is a major determinant of treatment efficacy and outcome. In this dimension, adherence is defined as maintaining patients, and especially the highly benefiting patients, on their drug by incentivizing payers with price adjustment in return for demonstrated adherence. We hypothesize that drug manufacturers would provide incentives to payers in two stages: in-advance (ex ante based on efficacy data from clinical trial measured at specific adherence rate) and in-arrear (ex post facto upon the availability of real-world data). With the in-advance incentive, the payer would receive payback for adhered patients who are expected to show a greater benefit compared to the general population studied in the clinical trial. In other words, a percentage of patients for whom the treatment is more suitable would benefit from an in-advance price reduction. To estimate this percentage of patients, we digitize the Kaplan-Meier curves and extract data as explained in the fifth dimension, to estimate the mean, the lower, and the upper estimates. For in-advance incentives, the percentage of patients within the upper and mean estimates would receive a full payback of the drug price in advance.
For in-arrear incentives (upon the availability of real-world data), the drug manufacturers provide payback to patients who did not benefit from the drug more than expected from the clinical trial. This in-arrear incentive requires real-world data for comparing the differential probability between the lower estimates of the real-world distribution and the lower estimates of the clinical trial distribution. If the differential is less than zero, the payer should pay the manufacturer according to this differential (the treatment’s value was underestimated in the clinical trial). If the differential is greater than zero, the drug manufacturer would return money to the payer according to that differential (the treatment’s value was overestimated in the clinical trial).
As in the fifth dimension, this percentage may not fully or accurately address the uncertainty generated by the analysis. Here too, we encourage using the term of “up to” the estimated percentage as a range for payment; specifically, from zero percent to the estimated percentCitation24.
For each efficacy measure, two price bounds are developed: one using the lower bound of the differential between upper and mean probability; and one using the upper bound of 0% price change. The lower and upper bounds are then averaged for MCS to derive the DSPAdherence Citation24.
As a caution, in our proof-of-concept exercise, real-world data were not yet available for the drug of interest and in-arrear incentives cannot be estimated at this time. Thus, the DSPAdherence reflects the price of the drug of interest considering the in-advance incentivesCitation24.
Integration analysis of the six dimensions
Our analytical platform concludes with an integrated analysis of the DSP metrics generated by each of the six dimensions to estimate a single price for a drug. This single price estimate is denoted as the ADPCitation25. Here, the ADP is a more informative and balanced price estimate than the DSP of a single dimension, as the ADP represents the average of all price variations within and among the dimensions. This variation is due to a higher degree of random patterns than at the DSP level. The random pattern of price variation is a methodology applied in cases when there is imperfect knowledge about a productCitation26,Citation27. With random pattern of price variation, different prices for a product are suggested under varying degrees of uncertainties, until less uncertain information becomes available and useful for pricing. Note that, at this time, the averaging is unweighted though manufacturers and payers could negotiate relative weights for each dimension in the finalization of the ADP.
Conclusion
Our method of estimating the ADP is characterized by the following: (1) methods are transparent and reproducible; (2) the ADP can be used for pricing during the lag-time between drug approval and availability of real-world data, thus, pre-empting opportunistic pricing; (3) both trial and, once available, real-world data can be used in pricing and contracting for pricing; (4) it is agnostic to disease and therapeutic area; and (5) it can be optimized to any payer’s and manufacturer’s data and priority decisionsCitation10.
Transparency
Declaration of funding
The work reported herein was performed without sponsorship or grant funding.
Declaration of financial/other interests
The authors have no financial relationships to declare.
JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.
Author contributions
All named authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship of this manuscript.
Acknowledgements
No assistance was received in the preparation of this article.
Notes
i Proscar is a registered trademark of Merck & Co., Inc., Kenilworth, NJ, USA.
References
- PhRMA. Value-based contracts: 2009-Q1. 2018. USA 2018; [cited 2019 Dec 1]. Available from: http://phrma-docs.phrma.org/files/dmfile/PhRMA_ValueBasedContracts_V9.pdf.
- Nazareth T, Ko JJ, Sasane R, et al. Outcomes-based contracting experience: research findings from U.S. and European stakeholders. J Manag Care Spec Pharm. 2017;23(10):1018–1026.
- Tonks A. Merck offers money back guarantee on finasteride. BMJ. 1994;309(6964):1252–1253.
- Piatkiewicz TJ, Traulsen JM, Holm-Larsen T. Risk-sharing agreements in the EU: a systematic review of major trends. Pharmacoecon Open. 2018;2(2):109–123.
- Kaltenboeck A, Bach P. Value-based pricing for drugs: theme and variations. JAMA. 2018;319(21):2165–2166.
- Garrison LP, Jr, Towse A, Briggs A, et al. Performance-based risk-sharing arrangements-good practices for design, implementation, and evaluation: report of the ISPOR good practices for performance-based risk-sharing arrangements task force. Value Health. 2013;16(5):703–719.
- Espin J, Rovira J, Garcia L. Experiences and impact of European risk-sharing schemes focusing on oncology medicines; [cited 2020 July 1]. Available from: https://pdfs.semanticscholar.org/d137/b85fc61a66b981a99249db9507f49eef30d1.pdf.
- APMNews.com. Celgene a conclu un accord “efficace ou remboursé” avecle CEPS pour Imnovid [article in French]. 2015; [cited 2020 July 1]. Available from: https://www.apmnews.com/Celgene-a-conclu-un-accord-efficace-ou-rembourse-avec-le-CEPS-pour-Imnovid-FS%20256638.html.
- Frisk P, Aggefors K, Cars T, et al. Introduction of the second-generation direct-acting antivirals (DAAs) in chronic hepatitis C: a register-based study in Sweden. Eur J Clin Pharmacol. 2018;74(7):971–978.
- Eriksson I, Wettermark B, Persson M, et al. The early awareness and alert system in Sweden: history and current status. Front Pharmacol. 2017;8:674.
- Clopes A, Gasol M, Cajal R, et al. Financial consequences of a payment-by-results scheme in Catalonia: gefitinib in advanced EGFR-mutation positive non-small-cell lung cancer. J Med Econ. 2017;20(1):1–7.
- Garattini L, Curto A. Performance-based agreements in Italy: ‘trendy outcomes’ or mere illusions? Pharmacoeconomics. 2016;34(10):967–969.
- World Health Organization. WHO guideline on country pharmaceutical pricing policies; [cited 2020 July 1]. Available from: https://apps.who.int/iris/bitstream/handle/10665/153920/9789241549035_eng.pdf;sequence=1.
- Uyl-de Groot CA, Löwenberg B. Sustainability and affordability of cancer drugs: a novel pricing model. Nat Rev Clin Oncol. 2018;15(7):405–406.
- Gray C, Kenney JT. Outcomes-based contracting for disease-modifying therapies in multiple sclerosis: necessary conditions for paradigm adoption. Am Health Drug Benefits. 2019;12(8):390–398.
- Luzzatto L, Hyry HI, Schieppati A, et al. Outrageous prices of orphan drugs: a call for collaboration. Lancet. 2018;392(10149):791–794.
- Godman B, Bucsics A, Vella Bonanno P, et al. Barriers for access to new medicines: searching for the balance between rising costs and limited budgets. Front Public Health. 2018;6:328.
- Adamski J, Godman B, Ofierska-Sujkowska G, et al. Risk sharing arrangements for pharmaceuticals: potential considerations and recommendations for European payers. BMC Health Serv Res. 2010;10:153.
- Alkhatib NS, Ramos K, Erstad B, et al. Pricing methods in outcome-based contracting-δ1: cost-effectiveness analysis and cost utility analysis-based pricing. J Med Econ. 2020;23(11):1215–1222.
- Alkhatib NS, Ramos K, Erstad B, et al. Pricing methods in outcome-based contracting-δ2: willingness-to-pay-based pricing. J Med Econ. 2020;23(11):1223–1229.
- Alkhatib NS, Erstad B, Ramos K, et al. Pricing methods in outcome-based contracting-δ3: reference-based pricing. J Med Econ. 2020;23(11):1230–1236.
- Alkhatib NS, Bhattacharjee S, McBride A, et al. Pricing methods in outcome-based contracting-δ4: safety-based pricing. J Med Econ. 2020;23(11):1237–1245.
- Alkhatib NS, McBride A, Bhattacharjee S, et al. Pricing methods in outcome-based contracting-δ5: risk of efficacy failure-based pricing. J Med Econ. 2020;23(11):1246–1255.
- Alkhatib NS, Slack M, Bhattacharjee S, et al. Pricing methods in outcome-based contracting-δ6: adherence-based pricing. J Med Econ. 2020;23(11):1256–1265.
- Alkhatib NS, McBride A, Slack M, et al. Pricing methods in outcome-based contracting: integration analysis of the six dimensions. J Med Econ. 2020;23(11):1266–1272.
- Griffith D. Four types of price variation. Presented in March 2013; Laramie, WY; [cited 2019 Dec 1]. Available from: http://www.uwagec.org/rightrisk/presentations/2013_03_26_DGriffithMarketing/1_Introduction/1_FourTypesPriceVariation.pdf.
- Bodoh-Creed A, Boehnke J, Hickman B. Using machine learning to predict price dispersion. Scholars at Harvard; [cited 2020 Dec 1]. Available from: https://scholar.harvard.edu/files/boehnke/files/bcbh_machine_learning_price_dispersion.pdf.