Avai lable onl ine at www.sc iencedirect .com

journal homepage: www.elsevier .com/ locate / jva l

Payer and Pharmaceutical Manufacturer Considerations forOutcomes-Based Agreements in the United StatesJoshua D. Brown, PharmD, PhD1,2,*, Rich Sheer, BS3, Margaret Pasquale, PhD3,Lavanya Sudharshan, MS3, Kirsten Axelsen, MS4, Prasun Subedi, PhD4, Daniel Wiederkehr, MPH4,Fred Brownfield, RPh5, Sachin Kamal-Bahl, PhD4

1Institute for Pharmaceutical Outcomes and Policy, University of Kentucky College of Pharmacy, Lexington, KY, USA; 2Department ofPharmaceutical Outcomes and Policy, University of Florida College of Pharmacy, Gainesville, FL, USA; 3Comprehensive HealthInsights Inc., Louisville, KY, USA; 4Pfizer Inc., New York, NY, USA; 5Humana Inc., Louisville, KY, USA

A B S T R A C T

Background: Considerable interest exists among health care payersand pharmaceutical manufacturers in designing outcomes-basedagreements (OBAs) for medications for which evidence on real-worldeffectiveness is limited at product launch. Objectives: To build hypo-thetical OBA models in which both payer and manufacturer canbenefit. Methods: Models were developed for a hypothetical hyper-cholesterolemia OBA, in which the OBA was assumed to increasemarket access for a newly marketed medication. Fixed inputs weredrug and outcome event costs from the literature over a 1-year OBAperiod. Model estimates were developed using a range of inputs formedication effectiveness, medical cost offsets, and the treated pop-ulation size. Positive or negative feedback to the manufacturer wasincorporated on the basis of expectations of drug performancethrough changes in the reimbursement level. Model simulationsdemonstrated that parameters had the greatest impact on payer costand manufacturer reimbursement. Results: Models suggested thatchanges in the size of the population treated and drug effectiveness

had the largest influence on reimbursement and costs. Despitesharing risk for potential product underperformance, manufacturerreimbursement increased relative to having no OBA, if the OBAimproved market access for the new product. Although reduction inmedical costs did not fully offset the cost of the medication, the payercould still save on net costs per patient relative to having no OBA bytying reimbursement to drug effectiveness. Conclusions: Pharma-ceutical manufacturers and health care payers have demonstratedinterest in OBAs, and under a certain set of assumptions both maybenefit.Keywords: managed care, outcomes-based agreements, pay forperformance, reimbursement.

Copyright & 2018, International Society for Pharmacoeconomics andOutcomes Research (ISPOR). Published by Elsevier Inc. This is an openaccess article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction

Growing focus on value in health care has led to a number ofinitiatives that are designed to shift the reimbursement system tobetter align costs of services to value in real-world settings [1,2].In recent years, there has been increased interest in manufac-turer and payer agreements that tie reimbursement to productperformance. These are generally referred to as performance-based risk-sharing arrangements or outcomes-based agreements(OBAs) [3]. In an OBA, performance in a defined patient popula-tion is tracked over a specified period of time in a definedpopulation or at the individual patient level, and the amountor level of reimbursement is determined on the basis of the

outcomes achieved [3,4]. Both pharmaceutical manufacturers andpayers have been motivated to develop such agreements to moreclosely align price and value. Recent headlines regarding highercost new medicines have also likely been a catalyst for interest inOBAs [5,6].

To date, most examples of implemented OBAs come fromEurope [5]. In the United States, attempts to structure OBAs havebeen few and far between, because the contracting parties oftenstruggle to align on and define the core metrics used to assesshealth outcomes under the contract [6]. In addition, specificdetails on existing deals are limited because of the proprietarynature of these agreements between individual payers andmanufacturers [5].

1098-3015$36.00 – see front matter Copyright & 2018, International Society for Pharmacoeconomics and Outcomes Research (ISPOR).

Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

http://dx.doi.org/10.1016/j.jval.2017.07.009

Conflicts of interest: R. Sheer, L. Sudharshan, and M. Pasquale are all employees of Comprehensive Health Insights Inc. F. Brownfieldis an employee of Humana Inc. K. Axelsen, P. Subedi, D. Wiederkehr, and S. Kamal-Bahl are employees and shareholders of Pfizer Inc.There are no other perceived or actual conflicts of interest.

E-mail: joshua.brown@ufl.edu.

* Address correspondence to: Joshua D. Brown, Department of Pharmaceutical Outcomes and Policy, University of Florida College ofPharmacy, 1225 Center Dr. HPNP #3320, Gainesville, FL 32606.

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 0

Given the difficulty observed in developing successful OBAs,the primary objectives of this study were to build models tobetter understand key variables having the greatest impact onoutcomes, costs, and cost sharing, and to propose design ele-ments for the development of OBAs that may improve eachparty’s willingness to negotiate such agreements in the future.

Methods

Hypercholesterolemia was chosen as a case example for the modelgiven recent innovations in treatment paradigms that are antici-pated to have a large budget impact on payers [7]. Hypercholester-olemia is associated with multiple clinical outcomes that can bemeasured including surrogate end points (low-density lipoproteincholesterol [LDL-C] measurements and goals) as well as “hard”clinical outcomes (i.e., acute myocardial infarction [MI] andstroke). Appropriate outcomes and target populations for imple-menting a hypercholesterolemia OBA were determined on thebasis of review of end points from published literature andprescribing information for available products [8–10]. Excel-basedmodels were developed to demonstrate the impact that variousparameters could have on OBAs. An internal project advisoryboard provided input via firsthand accounts of successes andlimitations of previous OBAs from both manufacturer and payerperspectives. This input was incorporated into the model develop-ment to address areas in which OBAs may be improved.

Data on hypercholesterolemia incidence were extracted fromthe Humana Research Database (Louisville, KY). The patientpopulation included patients enrolled in Medicare or commercialplans at Humana Inc. with an index diagnosis of familial hyper-cholesterolemia or a history of atherosclerotic cardiovasculardisease (ASCVD), defined as MI, stroke, angina, peripheral arterialdisease, or revascularization procedures [11]. Diagnosis andprocedure codes used in the identification process are detailedin Appendix Tables 1 and 2 in Supplemental Materials found athttp://dx.doi.org/10.1016/j.jval.2017.07.009. Rates of ASCVD eventsper 1000 person-years were established after the index diagnosisfor this patient population.

Two model structures were developed. The first modeledpatients reaching a goal LDL-C reduction on the basis of observedLDL-C reductions in pivotal clinical trials of proprotein conver-tase subtilisin/kexin type 9 (PCSK9) inhibitors, a novel class ofmedications indicated for hypercholesterolemia (model 1) [9,10].As part of the hypothetical OBA structure, patients with LDL-Cvalues at baseline would be followed up at 12 weeks aftertreatment initiation. Given that PCSK9 inhibitors have beenshown to be highly effective at lowering the LDL-C, the percent-age of patients meeting LDL-C goal reductions at 12 weeks (“first-line successes”) was set as 80% (ranging from 70% to 90%)[9,10,12–14]. The patients who did not meet LDL-C goal reductionswere classified into one of three groups: 1) failed goal andterminated therapy (40% of first-line failures, i.e., those 20% notmeeting initial LDL-C goals), 2) met goal after a dose change andan additional 12 weeks of therapy (40% of first-line failures), and3) failed goal after an additional 12 weeks of therapy (20% of first-line failures). For this OBA structure, the manufacturer was toabsorb financial responsibility for all PCSK9 inhibitor medicationcosts until the LDL-C goal was achieved, whereas the payer wasresponsible for medication costs afterward. Thus, the manufac-turer was responsible for 12 weeks of therapy costs for 80% of thepopulation and 24 weeks for the remaining 20% of the population(those requiring additional therapy before reaching goal orterminating therapy). Thereafter, the payer would reimbursethe manufacturer at the negotiated price. Total cost of partic-ipation (i.e., responsibility for medication costs) was calculatedfor both manufacturers and payers at a 1-year time period.

The second model structure (model 2) included ASCVD-related outcomes occurring after treatment initiation and up to1 year of follow-up. An expected rate reduction of these eventswas modeled for three different scenarios: 1) when the use ofthe medication was associated with a predefined outcome eventrate goal, 2) when the outcome event rate reduction exceeded thepredefined expectations (“overperformance”), and 3) when theoutcome event rate failed to meet expectations (“underperform-ance”). The baseline event rate was determined on the basis ofthe manage care organization’s population. Expected reductionsof 15% (range 10%–20%) in outcome events were estimated on thebasis of recently published data for the Further CardiovascularOutcomes Research with PCSK9 Inhibition in Subjects withElevated Risk (FOURIER) trial [14]. For the payment structure inmodel 2, a tiered system was developed that linked reimburse-ment to the three outcome event rate scenarios describedpreviously. Reimbursement for the product was reduced by 30%to 70% if the medication underperformed, and allowed forincreased reimbursement of 105% to 125% if the medicationoverperformed. Reimbursement rates remained flat if the reduc-tion in event rates met expectations.

Medication costs were assumed to be $1,167 per month, or$14,000 annually, on the basis of estimated prerebate costs forPCSK9 inhibitors currently on the market [15]. Costs were calcu-lated with an assumed adherence of 100% to therapy. Althoughthis is unlikely in real-world situations, we assumed the simplestcase considering that no real-world adherence data are availablefor our case example of PCSK9 inhibitors. Moreover, loweradherence rates would result in proportionally lower cost esti-mates (e.g., 80% adherence would result in 80% of medicationcosts). Because of the lack of real-world evidence on the exactrelationship between adherence levels and effectiveness, theeffect of lower adherence on effectiveness could be assumed tobe proportional as well. Many existing OBAs have reportedly usedpredefined cohort inclusion criteria specifying minimum adher-ence to therapy to assess the agreed upon outcomes of interest.Estimated medical offsets due to a reduced rate of ASCVD eventswere also included in the payer value calculations in model 2.These offsets were calculated on the basis of annual hospital-ization rates for MI, stroke, and unstable angina, as well ascoronary artery bypass graft and percutaneous coronary inter-vention procedures, taken from the Humana Research Database.The derived rate estimates were then multiplied by the incre-mental 1-year cost of care for each type of event, includingfollow-up outpatient care, on the basis of available literature(see Appendix Table 3 in Supplemental Materials found at http://dx.doi.org/10.1016/j.jval.2017.07.009) [8]. For each model, payercosts were calculated as total costs per treated patient, andpayments to manufacturers were defined as the total reimburse-ment paid to the manufacturer.

In each model, two simulations were included: “without OBA”and “with OBA.” We assumed that payers who implemented anOBA with a pharmaceutical manufacturer would deploy fewerbarriers to access the medication because the uncertainty regard-ing the medication’s value was shared between both parties.Thus, a scenario in which no OBA was present, formularyrestrictions, previous authorizations, higher patient co-pay-ments, or other barriers would be in place, which meant thatfewer patients would have access to the medication. Theseconsiderations incorporate either of the following two generalscenarios: 1) a medication is first to market or is unique in itsclass or 2) a medication has one or more comparators, with themanufacturer competing for market share. In scenario 1, restric-tions to market access for a given patient population may be dueto the nonformulary status of the drug. For scenario 2, tieredformularies may influence market share, and a relative positionversus competitors is also important to consider. These models

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 034

do not incorporate scenarios in which market access restrictionslimit certain patient populations from accessing the medication;rather, these models focus on the market influences for a generalpopulation for whom the medication is indicated.

In the “with OBA” scenario, a tiered reimbursement structurewas deployed. Manufacturers would receive less reimbursementwhen patients failed to meet the OBA goal, contractual reim-bursement when an outcome was met, and higher reimburse-ment via a bonus system when the medication performanceexceeded expectations. Sensitivity analyses were conducted forkey parameters including the size of the treated populationthrough improved market access, the effectiveness of the med-ication in achieving the desired outcome, and the amount of thetiered negative or positive reimbursement payments using arange of values for each parameter discussed earlier. OBA con-tracts were developed under a 1-year time horizon, and implica-tions beyond the 1-year time horizon are discussed. Modelinputs, sensitivity ranges, and sources are presented in Table 1.

Results

In the LDL-C goal model (model 1) without an OBA in place, thepayer’s cost per treated patient was static and the amount paid tothe manufacturer was the simple product of the drug cost andthe number of patients treated (Table 2). For a treated populationof 500 without an OBA, this translated to a total population costof $5,040,000 (range $4,935,000–$5,145,000), or $10,080 per treatedpatient (range $9870–$10,290), being paid to the manufacturer(see Table 2; range values reflect different assumptions regardingpopulation size and efficacy of treatment). With an OBA in place,some risk was taken on by the manufacturer, because the treat-ment cost was paid by the manufacturer until patients reachedthe goal. This reduced the payer’s costs to $7840 per treatedpatient (range $6510–$9170). After adjusting for those who failedtreatment, the manufacturer’s total reimbursement was$7,840,000 (range $6,510,000–$9,170,000). The model indicatedthat this increase in overall manufacturer reimbursement per-sisted as long as the size of the treated patient population, that is,the increase in market access, was at least 28% (52%–11%), forexample, from 780 (range 660–900) patients to 1000 patients in thebase model, larger with the OBA versus without the OBA (Fig. 1).

The ASCVD outcome model (model 2) included a tieredreimbursement strategy for a medication underperforming oroverperforming in the treated population for ASCVD-relatedclinical outcome event rates (Table 3). For the base case, in which1000 patients were treated with a 15% reduction in ASCVD events(worst case o10%, best case 420%), the payer cost per patientwas $14,000 ($7,000,000 for the total population). Annual healthcare costs for ASCVD events were estimated to be $9,220,234, withnet costs after medical offsets of $7,837,199 if a 15% reduction inoutcome event rates was achieved (Table 3). Taking into accountthe lower health care costs associated with fewer incurredASCVD events because of treatment, this translated to a net costper patient of $12,617 (range $13,170–$12,064).

With an OBA in place, the treated population was increased to1000 patients through assumptions regarding improved marketaccess as a driving motivation for a manufacturer to enter anOBA. If the medication failed to meet expectations, the manu-facturer received 50% (worst case 30%, best case 70%) lowerreimbursement for drug costs, resulting in the total manufacturerreimbursement of $7,000,000 (range $4,200,000–$9,800,000). Netcost per patient for the payer, taking into account the reduceddrug cost and the reduced cost associated with fewer ASCVDevents, was $6170 (range $3370–$8970). When the medicationexceeded expectations, the manufacturer received increasedreimbursement of 115% (105%–125%). This payment strategy

raised the total manufacturer reimbursement to $16,100,000(range $14,700,000–$17,500,000) with per-patient payer costs of$14,164 (range $12,764–$15,564).

In this example, under an agreement of 125% reimbursementpaid to the manufacturer when the medication exceeds outcomeexpectations, the payer incurred higher net cost per patientcompared with when an OBA was not in place even whenconsidering medication offsets (Fig. 2). The manufacturer reim-bursement remained higher than the “without OBA” reimburse-ment threshold of $7,000,000 so long as the maximum decreasein reimbursement remained higher than 50% in the case of adoubling of the treated population (Fig. 2). If there was a greaterthan 50% reduction in reimbursement, manufacturer reimburse-ment would be lower than under the scenario without an OBA inplace with a smaller treated population but no decrement inreimbursement because of underperformance. Ultimately, thefactors having the greatest impact on reimbursement and costswere the number of patients treated and drug effectiveness(Fig. 3).

Discussion

Recent interest in OBAs has emerged largely from payers’ desireto reduce the risks associated with high-cost pharmaceuticalswith uncertain real-world outcomes and effectiveness [6]. Byusing OBAs, payers can shift some risk to manufacturers withthe potential incentive of improving market access for themanufacturer’s product, thus improving patient access to noveltherapies and potentially generating important real-world evi-dence. Nevertheless, the use of OBAs in the United Statesremains nascent in practice and there remains considerableinterest among payers and manufacturers in understandinghow to structure successful OBAs in the United States [6]. Thiswas shown most recently by OBAs between PCSK9 inhibitormanufacturers and health plans [15,16]. These agreementsfocused on the medication’s ability to lower LDL-C as a surrogatefor effectiveness, which is analogous to model 1 in this study[15,16]. It will be of considerable interest to observe the actualfinancial impact of these OBAs and to track over time the valuegained by both parties. Incorporating the use of clinical outcomes(in this case, ASCVD events) versus surrogate outcomes (LDL-C)as part of the OBA structure should also be evaluated for futureinnovation and overall performance for both parties. As manu-facturers continue to publish data demonstrating the impact ofPCSK9 inhibitors on actual ASCVD events, the effectivenessmeasures can continue to be updated, relying on ASCVD eventreduction rather than on surrogate outcomes. How these newfindings impact future OBA structure will be of continuedinterest.

In this study, models were developed to examine importantconsiderations in structuring an OBA. Specifically, we sought tounderstand which variables had the largest impact on the dollarflow between the payer and the manufacturer on the basis oflearnings from the project team and the advisory board. Twocommon themes emerged for OBAs including the motivation toshare risk across both parties as well as inclusion of positivefeedback to the manufacturer. This bidirectional reimbursementconcept was subsequently modeled as a percent change in thepayment for either underperforming or overperforming on thebasis of the specified outcome of the OBA. This concept stemmedfrom discussions with the project’s advisory board in reference toexperiences with OBAs that focused only on off-loading risk tomanufacturers with no positive reinforcement for product suc-cesses. Such past practice may limit pharmaceutical manufac-turers’ desire to explore OBAs because of the increased risk.Similarly, managed care organizations are unlikely to offer

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 0 35

reimbursement bonuses because of prescription drug costs in theUS health care system. From our models, we were able to observebreakeven thresholds for both parties when an OBA would nothave detrimental budgetary impact compared with the base casein which an OBA does not exist. This is important to establish innegotiations of an OBA so that some perception of value ispresent to bring both parties to the table. In this model, thethreshold for the payer allows a 4% higher reimbursement paid tothe manufacturer for a high-performing product. This thresholdcould be higher, for more effective medications, on the basis ofhigher rate reductions in outcome events, which would generatemore medical cost offsets and allow for a larger positive reim-bursement policy to the manufacturer.

We further observed that changing assumptions around thenumber of treated patients had a significant impact on the

economic impact of the OBA to the manufacturer. If the size ofthe treated population does not increase with the OBA, there isno direct motivation for a manufacturer to enter such an agree-ment in which the reimbursement would be subject to new risk,but access to the medication is not improved. Thus, OBAs makeinherently more sense for medications that are subject toformulary management through adverse tier placement, previousauthorizations, or step edits whether in absolute changes for themedication or in relative standing compared with competitormedications. From a payer perspective, focusing on medicationswith high drug costs and/or high prevalence will be of the mostinterest to reduce the overall budget impact [17]. Furthermore, anincrease in the size of a treated population would also preventmore outcome events, thereby benefiting both the payer and thepatient. Our models assumed a homogeneous patient population

Table 1 – OBA model inputs, sensitivity ranges, and sources.

Model input Basecase

Sensitivityrange

Source

Treated population (with OBA)* 1000 600–900 Assumption for increased market access vs. “without OBA”population of 500

Annual drug cost (per patient) $14,000 – Gray literatureLDL-C modelDrug effectiveness

LDL-C reduction (% of populationthat met the LDL-C goal at 12 wk)†

80% 70%–90% PCSK9 inhibitor pivotal clinical trials [12,13]

Percent reimbursement (LDL-C) 100% – Negotiated medication cost between payer and manufacturer. Actualpayment is determined by the proportion of patients meeting orfailing goal per LDL-C outcome definition.

ASCVD outcome event rate modelASCVD incidence rate reduction

with lowered LDL-C15% 10%–20% [14]

Percent reimbursementBase case 100% – Negotiated contract price if medication meets expectations

Conditional on medicationunderperformance (o40% ASCVDrate reduction)

50% 30%–70% Assumed range to determine breakeven points

Conditional on medicationoverperformance (460% ASCVDrate reduction)

115% 105%–125% Assumed range to determine breakeven points

CV event incidence (per 1,000persons)

– Humana Research Database

Myocardial infarction 51.0Stroke 55.8Unstable angina 1.7Angioplasty 19.1CABG 17.6

Annual cost of care per person‡ – [8]Myocardial infarction $64,035Stroke $56,352Unstable angina $31,780Angioplasty $54,934CABG $97,154

ASCVD, atherosclerotic cardiovascular disease; CABG, coronary artery bypass graft; CV, cardiovascular; LDL-C, low-density lipoproteincholesterol; OBA, outcomes-based agreement; PCSK9, proprotein convertase subtilisin/kexin type 9.⁎ Assumed that the population size would be increased with an OBA in place because of improved market access.† Among the 20% who did not meet the LDL-C goal after 12 wk, it was assumed that some discontinued treatment (40%), some continued at ahigher dose and met goal (40%), and some continued at a higher dose and did not meet goal (20%).

‡ The total cost of care, taken by multiplying incidence by the annual cost of care and summing for all CV events, was $9,220,234. Costs of carewere incremental from the Bonafede study and represented quarterly cost estimates for the first quarter after an event plus three additionalquarters of costs to equal 12 mo of follow-up costs. Confidence intervals for a sensitivity range were not used because of the narrow rangeand negligible effect on OBA results (~$50 per-patient costs).

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 036

with or without the OBA, that is, similar risk of outcome events.Conceivably, access restrictions may limit medication use tothose patient groups with the highest risk, and each additionalincrease in access may introduce patients with incrementallylower risk. Further consideration in OBA design is needed toincorporate how formulary management directs medication useto certain risk groups and whether those additional patients whoare granted access through an OBA are representative or wouldhave lower medical offsets.

Because our models encompassed only a 1-year time horizon,they did not account for the potential longer term ramificationsof these agreements and the real-world evidence they couldgenerate. If a product performed well in the population, it would

likely have positive, long-term market access implications,potentially further increasing market share and revenues forthe life cycle of the medication. Conversely, underperformanceof a product in an OBA may jeopardize future market accessthrough increased formulary management by the payer, maybeeven beyond the previous level the OBA was intended to over-come. From a payer’s perspective, evidence generated through anOBA could drive further competition within the therapeutic class.For other high-cost medication classes, for example, specialtyoncology products, this evidence can help inform formularydecision making, decreasing costs by allowing more effectivemedications to be placed in preferred tiers. These considerationshave been modeled in depth in a previous study that considered

Table 2 – Base-case and sensitivity results for LDL-C goal attainment outcome (model 1).

Simulation Treatedpopulation

Drug effectiveness*

(%)Manufacturer reimbursement

($)Payer per-patient cost

($)

WithoutOBA

500 80 5,040,000 10,0801000 80 10,080,000 10,080500 70 4,935,000 9870500 90 5,145,000 10,290

With OBA 1000 80 7,840,000 78401000 70 6,510,000 65101000 90 9,170,000 9170

Note. Boldfaced figures represent the base case without OBA.LDL-C, low-density lipoprotein cholesterol; OBA, outcomes-based agreement.⁎ Percentage of patients who achieve the LDL goal at 12 wk.

Fig. 1 – Change in patient population needed for manufacturer reimbursement breakeven point with an OBA versus withoutan OBA in model 1. Note. Dashed horizontal lines represent the manufacturer reimbursement with an OBA, including 1000treated patients, calculated with base-case and sensitivity thresholds for medication effectiveness from model 1 (Table 2).Worst case and best case represent scenarios in which the manufacturer is responsible for treatment failures. Solid diagonallines represent reimbursement by varying the number of patients treated without an OBA. The intersections of lines of thesame color represent the breakeven points when an increase in the number of patients through an OBA results in the samereimbursement despite some of the medication costs paid by the manufacturer. For example, with the base case, the OBAwould be beneficial to the manufacturer so long as the treated population increased from 780 to 1000 patients with an OBA, anincrease of 28% (intersection of green lines). OBA, outcomes-based agreement.

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 0 37

formulary decision making by the payer in terms of net monetarybenefit of a drug rebate as well as pricing motivation for themanufacturer [18]. Our study added consideration of medicaloffsets for the payer as a result of the medication’s effectiveness,and factors these into the value calculation of an OBA. From thisanalysis, payers are likely to always benefit from OBAs viareduced drug costs, especially in therapeutic classes with mar-ginal differences in effectiveness between competitors. Never-theless, incorporating a positive incentive to the manufacturercould increase payer costs dependent on the overall effectivenessand total medical cost offsets. In the example of PCSK9 inhibitors,

this increase in cost to the payer rises to a threshold of 5% whenthe highest effectiveness levels are achieved. Manufacturers canalso benefit and receive positive feedback through increasedreimbursement for sharing the risk with the payer when themedication is ultimately demonstrated to be more effective thanexpected. More flexibility in this scheme is possible for medica-tions with a combination of higher medical cost offsets and largermargins of effectiveness.

Another study surveyed key opinion leaders and acknowl-edged that OBAs are underutilized with limitations related to thecosts and additional effort to implement the OBA [6]. Participants

Table 3 – Base-case and sensitivity results for model 2 with outcome event reduction tied to reimbursement.

Simulation Treatedpopulation

Ratereduction*

(%)

Percentreimbursement†

Manufacturerreimbursement

($)

Payerper-

patientcost ($)

Medicaloffsets‡

($)

Payer netper-

patientcost ($)

WithoutOBA

500 15 100 7,000,000 14,000 691,518 12,6171000 15 100 14,000,000 14,000 1,383,035 12,617500 o10 100 7,000,000 14,000 414,911 13,170500 420 100 7,000,000 14,000 968,125 12,064

With OBA 1000 15 100 14,000,000 14,000 1,383,035 12,6171000 o10 50 7,000,000 7000 829,821 61701000 420 115 16,100,000 16,100 1,936,249 14,1641000 o10 30 4,200,000 4200 829,821 33701000 o10 70 9,800,000 9800 829,821 89701000 420 105 14,700,000 14,700 1,936,249 12,7641000 420 125 17,500,000 17,500 1,936,249 15,564

Note. Boldfaced figures represent the base case without OBA and a reduced market access population size.OBA, outcomes-based agreement.⁎ Expected rate reduction of 15% for cardiovascular events including myocardial infarction, stroke, unstable angina, and percutaneouscoronary intervention/coronary artery bypass graft procedures with a range of 10%–20%.

† Percent of total drug payment by the payer to the manufacturer.‡ Based on the total (acute and follow-up) cost of care for prevented cardiovascular disease events and procedures [8] (see Appendix Table 3 inSupplemental Materials).

$2000

$4000

$6000

$8000

$10,000

$12,000

$14,000

$16,000

$18,000

$2,000,000

$4,000,000

$6,000,000

$8,000,000

$10,000,000

$12,000,000

$14,000,000

$16,000,000

$18,000,000

30% 40% 50% 60% 70% 80% 90% 100% 110% 120%

Payer net cost per patient

Man

ufac

ture

r rei

mbu

rsem

ent

Manufacturer reimbursement with OBA Manufacturer reimbursement without OBA

Payer cost with OBA Payer cost without OBA

Fig. 2 – Breakeven thresholds of positive and negative reimbursements for payers and manufacturers in OBAs. Note. Solidhorizontal lines represent the manufacturer reimbursement (red) and payer cost per patient (blue) without an OBA, including500 treated patients, calculated with base case (50% rate reduction) medication effectiveness in model 2. Dashed lines are thetotal reimbursement and costs at varying percentages of reimbursement on the basis of the medication performance in thetreated population. The intersection of same colored lines represents the breakeven points (denoted by the vertical blacklines) for an overperforming product (payer breakeven 104% bonus payment to manufacturer) and an underperformingproduct (manufacturer breakeven 50% reduced reimbursement penalty). OBA, outcomes-based agreement.

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 038

in that study noted other relevant considerations such as dataquality and availability, measurement of meaningful and rele-vant real-world outcomes, and potential regulatory or legalimplications. In general, a formal collaboration agreementbetween manufacturers and payers could help to reduce uncer-tainty in establishing OBAs. The cost to evaluate an OBA wasconsidered to be equivalent to conducting an observational studyincluding research personnel, data, and so forth, for outcomesthat can be measured using claims data. Costs associated withpatient- or clinician-reported outcomes would inherently behigher, but there were no real-world examples of OBAs that haveused such an agreement. The cost to administer the agreement atthe plan level is established given that the payer would bemanaging the therapeutic class and need only to make adjust-ments to the formulary.

This work is limited in considering only one type of innovativecontracting agreement in OBAs. The work by Walker et al. [18]and an International Society for Pharmacoeconomics and Out-comes Research task force [4] established a taxonomy of coverageoptions including outcomes-based, nonoutcomes-based, andevidence-generating agreements. Our study considered only out-comes-based contracts given that, by name, these are the onlyforms of agreement that measure outcomes and tie those to

reimbursement. Other forms of coverage options are negotiationsaround product price and not necessarily measurable outcomes,more typical of those used in settings outside the United States.Similarly, coverage with evidence development generallyinvolves payers allowing market access at a reduced medicationprice so that evidence can be collected on a new medication forfuture decision making, but without payment being tied to theattainment of an outcome [18]. Although our models are specifiedfor a single therapeutic area, the concepts of identifying thepotential change in market access with an OBA, breakeventhresholds for both parties, and overall general structure can beextrapolated and applied to multiple therapeutic areas by alteringthe assumptions around incidence rates, treatment effectiveness(i.e., event rate reduction), and the associated pharmaceuticaland medical costs of treatment.

Conclusions

Although a few OBAs have been established in the United Statesto date, more are likely to be initiated because of the desire tobetter align costs and performance associated with a medicine.By structuring agreements with shared risk between the payer

Fig. 3 – Sensitivity analysis of key parameters in OBAs and impact on manufacturer reimbursement and payer cost comparedwith base case. Note: Diagram shows change in reimbursement or costs from the base case for the best-case scenario (red) andthe worst-case scenario (blue) for each of the parameters. OBA, outcomes-based agreement.

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 0 39

and the pharmaceutical manufacturer, positive and negativeincentives via reimbursement levels, and careful considerationof potential breakeven points, payers may incentivize manufac-turers to enter OBAs, decrease risk in the short-term, and providenew evidence to inform long-term formulary management.Manufacturers entering OBAs can benefit from improved patientaccess while supporting evidence generation for the real-worldeffectiveness and safety of their product. By incorporating theseconcepts into an OBA, along with a formal collaboration agree-ment between the payer and the pharmaceutical manufacturer,more OBAs and future innovation may be possible. Furtherevaluations of obstacles and benefits of real-world OBAs areneeded to establish the true short-term and long-term implica-tions for both parties.

Source of financial support: This study was funded by PfizerInc. and conducted in collaboration with Comprehensive HealthInsights, a wholly owned subsidiary of Humana Inc. J. D. Brownreceives fellowship funding from both Humana Inc. and PfizerInc. There are no other financial disclosures.

Supplemental Materials

Supplemental material accompanying this article can be found inthe online version as a hyperlink at http://dx.doi.org/10.1016/j.jval.2017.07.009 or, if a hard copy of article, at www.valueinhealthjournal.com/issues (select volume, issue, and article).

R E F E R E N C E S

[1] PwC Health Research Institute. Unleashing Value. The ChangingPayment Landscape for the US Pharmaceutical Industry. PwC, 2012.

[2] Forcier K, Model N. The potential for outcomes-based contracting.Available from: http://pharmaceuticalcommerce.com/index.php?pg=information_technology&articleid=27385. [Accessed October 1,2014].

[3] Carlson JJ, Sullivan SD, Garrison LP, et al. Linking payment to healthoutcomes: a taxonomy and examination of performance-basedreimbursement schemes between healthcare payers andmanufacturers. Health Policy 2010;96:179–90.

[4] Garrison LP Jr, Towse A, Briggs A, et al. Performance-based risk-sharingarrangements—good practices for design, implementation, andevaluation: report of the ISPOR Good Practices for Performance-BasedRisk-Sharing Arrangements Task Force. Value Health 2013;16:703–19.

[5] Carlson JJ, Gries KS, Yeung K, et al. Current status and trends inperformance-based risk-sharing arrangements between healthcarepayers and medical product manufacturers. Appl Health Econ HealthPolicy 2014;12:231–8.

[6] Garrison LP Jr, Carlson JJ, Bajaj PS, et al. Private sector risk-sharingagreements in the United States: trends, barriers, and prospects. Am JManag Care 2015;21:632–40.

[7] Tice JA, Kazi DS, Pearson SD. Proprotein convertase subtilisin/kexintype 9 (PCSK9) inhibitors for treatment of high cholesterol levels:effectiveness and value. JAMA Intern Med 2016;176:107–8.

[8] Bonafede MM, Johnson BH, Richhariya A, Gandra SR. Medical costsassociated with cardiovascular events among high-risk patients withhyperlipidemia. Clinicoecon Outcomes Res 2015;7:337–45.

[9] Amgen Inc. REPATHA (evolucumab): prescribing information. Availablefrom: http://pi.amgen.com/united_states/repatha/repatha_pi_hcp_english.pdf. [Accessed April 15, 2016].

[10] Sanofi-Aventis. PRALUENT (alirocumab): prescribing information.Available from: http://products.sanofi.us/praluent/praluent.pdf.[Accessed April 15, 2016].

[11] Stone NJ, Robinson JG, Lichtenstein AH, et al. 2013 ACC/AHA guidelineon the treatment of blood cholesterol to reduce atheroscleroticcardiovascular risk in adults: a report of the American College ofCardiology/American Heart Association Task Force on PracticeGuidelines. Circulation 2014;129(Suppl. 2):S1–45.

[12] Robinson JG, Farnier M, Krempf M, et al. Efficacy and safety ofalirocumab in reducing lipids and cardiovascular events. N Engl J Med2015;372:1489–99.

[13] Sabatine MS, Giugliano RP, Wiviott SD, et al. Efficacy and safety ofevolocumab in reducing lipids and cardiovascular events. N Engl J Med2015;372:1500–9.

[14] Sabatine MS, Giugliano RP, Keech AC, et al. Evolocumab and clinicaloutcomes in patients with cardiovascular disease. N Engl J Med2017;376:1713–22.

[15] Loftus P, Mathews AW Health insurers push to tie drug prices tooutcomes: Cigna reaches value-based contracts for entire new class ofcholesterol drugs. 2016. Available from: http://www.wsj.com/articles/health-insurers-push-to-tie-drug-prices-to-outcomes-1462939262.[Accessed July 2, 2016].

[16] Avalere. Health plans are interested in tying drug payments to patientoutcomes. Available from: http://avalere.com/expertise/life-sciences/insights/health-plans-are-interested-in-tying-drug-payments-to-patient-outcomes.[Accessed July 10, 2016].

[17] Zaric GS, Xie B. The impact of two pharmaceutical risk-sharingagreements on pricing, promotion, and net health benefits. ValueHealth 2009;12:838–45.

[18] Walker S, Sculpher M, Claxton K, Palmer S. Coverage with evidencedevelopment, only in research, risk sharing, or patient access scheme?A framework for coverage decisions. Value Health 2012;15:570–9.

V A L U E I N H E A L T H 2 1 ( 2 0 1 8 ) 3 3 – 4 040