FormalPara Key Summary Points

Existing literature on the financial impact of biosimilar use is limited to budget impact models from the payer’s perspective.

The impact of biosimilar adoption on reducing total cost of care (TCOC) and provider financial risk for exceeding TCOC benchmarks in oncology value-based payment models is unclear.

We estimated the impact of biosimilar substitution for reference products on TCOC and provider financial performance in 6-month episodes of care as defined in the Medicare Oncology Care Model.

Biosimilar substitution decreased TCOC by an average of $1193 per episode, and resulted in a 15% reduction in the risk of providers needing to pay recoupments to Medicare for exceeding TCOC benchmarks.

For oncology providers participating in value-based payment programs, substitution of biosimilars for reference agents may act as a key mechanism for reducing total cost of care without diminishing quality and improving financial performance within the terms of oncology value-based payment programs.

Introduction

Oncology is a key area of focus for healthcare payment reform efforts in the USA, with total medical costs associated with cancer expected to grow from $184 billion in 2015 to $246 billion in 2030 [1,2,3]. Over the past several years, numerous public and private sector payers have implemented new models that transition reimbursement from fee-for-service (FFS) to alternative value-based payment (VBP) models [4, 5]. These models generally involve tracking of total cost of care (TCOC) for oncology treatment, comparing those costs to TCOC benchmarks set by the given VBP model, and providing performance-based bonus payments for achieving TCOC below the benchmark and/or owing recoupment payments for exceeding the TCOC benchmark. The largest VBP model to date is Medicare’s Oncology Care Model (OCM), which was implemented from 2016 to 2022, has had significant influence on oncology payment reform efforts in the USA, and informs the design of Medicare’s new VBP called the Enhancing Oncology Model (EOM), which began in July 2023 [6].

Biosimilars are an important group of therapies in the context of oncology VBP models because they can reduce TCOC without diminishing quality of care. This is because biosimilars are approved by the US Food and Drug Administration (FDA) on the basis of having no clinically meaningful differences from an existing FDA-approved biologic, and are typically priced more than 25% lower relative to their biologic reference products [7,8,9]. Additionally, biosimilars can be particularly impactful in oncology VBPs because they are currently more prevalent in oncology indications than many other therapeutic areas. At the time of this study, there were biosimilars for six molecules FDA-approved in oncology—the antineoplastic agents bevacizumab [10], rituximab [11], and trastuzumab [12], and the supportive care agents epoetin alfa [13], filgrastim [14], and pegfilgrastim [15]. As such, biosimilars may be applicable in a large proportion of care episodes in oncology, and thereby could prove to be critical tools to manage TCOC and reduce provider risk of paying recoupments for exceeding TCOC benchmarks or quality standards set by VBPs [16, 17].

However, transitioning from reference products to biosimilars can be complex for some providers [18]. Key gaps in knowledge exist about biosimilar outcomes in “real-world” settings, particularly after switching from a reference product, which can hamper the adoption of biosimilars by providers [19]. Available studies evaluating biosimilars within VBP largely focused on the impact to payers rather than providers [20, 21]. Additionally, the existing literature largely relies upon general budget impact methods rather than modeling the specific payment methodologies of a specific VBP initiative [22]. In particular, evaluating the impact of biosimilars without comparison to an expected cost of care limits the estimation of how biosimilars may impact provider risk of recoupments for exceeding TCOC benchmarks [23, 24]. Furthermore, many budget impact models focus primarily on comparisons between a particular reference product and one or more products biosimilar to it, rather evaluating impacts to an overall portfolio of oncology reference products and biosimilars that providers could potentially use.

In this study, we estimated the impact of oncology biosimilar substitution for reference products from the perspective of providers participating in the Medicare OCM as it is the largest oncology VBP to date and its methodology is fully defined [25]. To better inform provider decision-making, our study estimated the impact of biosimilar adoption on TCOC and provider risk of owing recoupments for exceeding TCOC benchmarks in the final OCM performance period (second half of 2021). We used observed Medicare claims data and simulation techniques to evaluate adoption of a broad portfolio of biosimilars that were available to providers at that time. Our findings may provide insights about the impact of biosimilars in managing cost of care in the largest oncology VBP to date and inform treatment planning for providers considering participation in the new iteration of the Medicare oncology VBP—the Enhancing Oncology Model.

Methods

Data

The Medicare Limited Data Set (LDS) analytic files from 2020 were used to evaluate the impact of biosimilar adoption on TCOC and provider performance relative to TCOC benchmarks [26]. The LDS analytic files are Medicare FFS claims from a random 5% sample of Medicare FFS beneficiaries that include information about claims for inpatient and outpatient care, skilled nursing facilities, home health agencies, hospice, and durable medical equipment. We implemented major components of the Medicare OCM methodology in LDS claims from 2016 to 2020 that used biologic products meeting our study criteria. Prior-year claims data are particularly important for framing prior episodes of care for patients in the study population, as these episodes can impact the timing of the patients’ episodes in the study period. Additionally, risk adjustment of episodes in the study population required prior-year claims data in order to determine applicable comorbidities that impact the expected cost of care—technically referred to as the OCM-Modified Hierarchical Condition Categories (HCCs) [27].

Biologic product prices and market share data were derived from sources that aligned with the simulation timeframe of the second half of 2021. Prices for biologic reference products and biosimilars were taken from Medicare’s (Q4 2021) Average Sales Price (ASP) Drug Pricing File [28]. Market share data (Q3 2021) was derived from IQVIA Drug Distribution Data (DDD) [29] for biologic products and used to construct weighted average pricing for biosimilars where there was more than one biosimilar to the reference product.

The study was conducted using Medicare Standard Analytic Files (a limited dataset) in accordance with a data use agreement between Tuple Health and Medicare. An institutional review board was exempted because the study only accessed this deidentified limited dataset.

Study Drug Inclusion Criteria

We evaluated the impact of biosimilar adoption as a portfolio, meaning that our analyses compared a setting where included patients received treatment with only biologic reference products relative to an alternative scenario where biosimilars were substituted for all of those reference products. Biosimilars that met the following criteria were selected: (1) biologic products designated by the FDA as biosimilar to a corresponding reference product and commercially available prior to July 2, 2021 (the start date for the Performance Period (PP) 11—the final OCM PP) [30]; and (2) biosimilar products with an FDA indication related to at least one of the 21 OCM-eligible cancers [30]. The following six sets of biologic products met our inclusion criteria: bevacizumab biosimilars [31, 32], pegfilgrastim biosimilars [33,34,35,36], rituximab biosimilars [37,38,39], filgrastim biosimilars [40, 41], epoetin alfa biosimilars [42], and trastuzumab biosimilars [43,44,45,46,47] (see Table 1).

Table 1 Population-level characteristics of Medicare Limited Data Set (LDS)-framed episodes
Full size table

Unit of Analysis and Study Outcomes

The unit of analysis was a 6-month patient episode of care as defined by the OCM methodology. The OCM methods have specific claims-based criteria for when the episodes begin, when the episodes end, determining the cancer being treated in the episode, and overall inclusion/exclusion criteria (e.g., no beneficiaries who were enrolled in Medicare Advantage, no beneficiaries who received the End-Stage Renal Disease benefit) [25].

The primary outcome for this study was change in the 6-month care episode TCOC in the scenario with biosimilar use vs. the scenario with only reference product use. This TCOC outcome was also used to calculate oncology provider financial performance relative to OCM TCOC benchmarks in episodes in which those biosimilars were applicable. Specifically, we evaluated the OCM two-sided risk model in which providers could earn a performance-based payment for achieving less than 97.5% of the TCOC benchmark, would owe a performance-based recoupment for achieving more than 100% of the TCOC benchmark, and would neither receive nor owe payment if in the 97.5–100% of TCOC benchmark “safe zone.”

TCOC in OCM consisted of standardized payer responsibility from Part A and Part B, and a portion of Part D costs (80% of gross drug cost above the catastrophic threshold plus the low-income cost sharing subsidy). Evaluation of TCOC under the terms of the OCM involves two prices: (1) the OCM “benchmark” price, and (2) the OCM “target” price. The “benchmark” price for each episode (i.e., Medicare’s projected cost for the episode) is calculated using a two-component regression model consisting of a baseline price projection (reflective of the expected episode cost during the baseline period of the model) and a trend factor (which estimated the amount of medical cost inflation for that episode in a specific performance period). This “benchmark” is used to define the TCOC threshold above which providers owe a performance-based recoupment. Alternatively, the “target” price for each episode (i.e., Medicare’s financial goal for the episode to determine performance-based payment or recoupment) is calculated by applying a 2.5% discount to the benchmark price that is described above. This “target” is used to define the TCOC threshold below which providers qualify for performance-based payment.

The secondary outcome of the study was to assess how provider financial performance in the OCM changed when biosimilars were substituted for reference products. The secondary outcome was defined as the average change in OCM risk category after biosimilar substitution for sampled cohorts of 100 eligible episodes randomly drawn from the study population to account for variability in individual episode performance. Cohort sampling was repeated over 10,000 simulation runs to account for variability in cohort cancer distributions. Across the simulation runs, three risk categories were evaluated in accordance with OCM’s alternative two-sided risk methodology: (1) Below Target, where TCOC is less than the predicted OCM target; (2) Inside the Safe Zone, where TCOC is greater than or equal to the OCM target, but less than or equal to the predicted OCM benchmark price; and (3) Outside the Safe Zone, where the TCOC is greater than the predicted OCM benchmark price [25].

Simulation Model

The study used the OCM simulation module in the Tuple Health Real World Evidence (RWE) Technology Platform which included operationalizing all key facets of the OCM methodology including OCM episode framing, cancer attribution, risk adjustment, TCOC calculation, as well as OCM’s explicit inclusion/exclusion criteria [25, 48]. The implementation of the OCM methodology and the Tuple Health simulation module was previously cross validated using empirical OCM data, independent of this study. Modeled predictions for OCM TCOC targets were greater than 99.9% accurate when compared to the episode targets provided by Medicare [48, 49].

As an extension to this core OCM methodology implementation, Part D costs and Part D enrollment for OCM episodes were simulated in order to address key gaps in publicly available research datasets [50]. Part D cost ranges were initially estimated from a random sample of OCM data, and zero-inflated gamma distributions for sampling were assumed to construct the models. The simulation module was adapted for the purposes of the study, and Monte Carlo methods were used to estimate the impact of biosimilar substitution [25, 51].

From the initial set of framed episodes, we identified all episodes utilizing reference products corresponding to the biosimilars of interest and any episodes from the LDS data using one of the study biosimilars. We referred to these as Biosimilar Applicable Episodes (BAEs). Claims for biosimilars in episodes were replaced with claims for their corresponding reference product in order to standardize claims in the initial population of episodes. These reference products and their corresponding biosimilars were considered to be Biosimilar Applicable Products (BAPs).

Episodes in the study population initiated between January 2, 2020 and July 1, 2020, corresponding to OCM PP8. For each episode in the study population, we computed the TCOC and performed episode cost truncation (winsorization) in accordance with the OCM methodology. Costs were computed first under the assumption of the use of reference products, then again under the assumption of the use of biosimilar products. Costs for reference products were computed using the number of billed units in the episode (inclusive of drug waste) for each BAP multiplied by the per-unit prices for the reference products from Medicare’s Q4 2021 ASP Drug Pricing File [28]. Costs for biosimilars were calculated by substituting reference products with biosimilars at dose equivalence, except for claims for subcutaneous formulations of rituximab and trastuzumab. In these instances, the number of substituted units of biosimilar rituximab or trastuzumab were estimated by using the median dose of intravenous rituximab or trastuzumab in LDS claims from patients with the same gender. Some reference products have multiple biosimilars, and in these cases we weighted the ASP calculation by the market share of each applicable biosimilar observed in IQVIA data for the per-unit price impact of biosimilar substitution.

To simulate episodes of care and provider financial performance, TCOC targets were computed in accordance with the terms of the OCM alternative two-sided risk arrangement. In this arrangement, providers could receive a performance-based payment of up to 16% of their annual Part B revenue (as defined by OCM payment methodology) if their TCOC was below 97.5% of the total benchmark amount. If TCOC exceeded the benchmark amount, providers could owe a recoupment to Medicare of up to 8% of their annual Part B revenue.

Using Monte Carlo methods, we drew cohorts of 100 episodes from the study population by random sampling with replacement [51]. To estimate the distribution of the outcomes, the simulation was repeated 10,000 times with new draws of 100 episodes. In the primary study outcome, we computed the average per-episode TCOC. In the secondary study outcome, we compared the simulated TCOC to the TCOC target (97.5% of benchmark) and TCOC benchmark prices.

Biosimilar substitution of BAPs in an episode only occurred under specific assumptions that were dependent on the type of product. For antineoplastic BAPs, the patient was required to be treatment naïve to the BAP prior to the current episode of care. This was motivated by prior studies, which find mixed (but increasing) levels of physician support for switching to biosimilar antineoplastics in patients currently treated with reference products [19, 52]. Treatment-naïve status to a BAP was defined as the absence of claims for the BAP in all framed episodes of care for the patient in all prior framed episodes (initiating between July 1, 2016 and January 1, 2020). For supportive therapy BAPs, biosimilar substitution was conducted for all applicable episodes regardless of treatment-naïve status. We assumed practices adopting biosimilars would do so for all potential corresponding reference products (i.e., a portfolio strategy).

Results

From the observed population of 3,296,240 beneficiaries in the 2020 LDS analytic files, a total of 8281 episodes of care were identified with antineoplastic treatment trigger events occurring in the study period of January 2, 2020 to July 1, 2020 (see Table 1). Of these, 1586 (19.2%) episodes met the criteria for BAP use and potential OCM eligibility. A total of 80% of episodes were assigned to one of the seven OCM cancers with the highest prevalence in the study population: breast (19.4%), lymphoma (18.5%), lung (12.9%), colorectal/small intestine (12.7%), prostate (6.5%), pancreatic (5.3%), and ovarian (4.9%). A total of 49.9% of the study population had use of a BAP antineoplastic in the episode, 66.0% had use of a BAP supportive therapy, and 16.0% had use of a BAP antineoplastic and a BAP supportive therapy.

Use of BAPs by OCM-assigned cancer type is summarized in Table 2. Use of both a BAP antineoplastic and BAP supportive therapy was most frequent in lymphoma (37.8% of episodes), ovarian (32.1%), and colorectal/small intestine (23.4%), and less frequent in breast (9.7%).

Table 2 Use of biosimilar applicable products (BAPs) by cancer typea
Full size table

Biosimilar substitution resulted in a $1193 reduction in TCOC per episode in the cohort (95% CI $583–1840). This reduction in TCOC represents an average of a 2.4% reduction in TCOC relative to the aggregate benchmark TCOC for the cohort (Table 3). Biosimilar substitution led to a reduction in TCOC relative to benchmark in 9996 of the 10,000 simulated cohorts (Table 3).

Table 3 Simulation outcomes
Full size table

Biosimilar substitution was effective at reducing the proportion of cohorts of applicable episodes in which providers owed a performance-based recoupment for exceeding TCOC benchmarks (Fig. 1). With the use of reference products only, in 10,000 simulation runs, 4440 (44.4%) runs had cohort TCOC above the safe zone—i.e., owing performance-based recoupments. In 1350 of the 4440 (30.4%) simulation runs above the safe zone, alternative use of biosimilar products reduced the cohort TCOC to be in the safe zone—thereby neither receiving nor owing an additional amount. In 191 of the 4440 (4.3%) simulation runs above the safe zone, use of biosimilar products reduced the cohort TCOC to an amount below the TCOC target—thereby earning a performance-based payment.

Fig. 1
figure 1

Change in aggregate cohort risk under biosimilar substitution

Full size image

In 1674 (16.7%) of overall simulation runs using reference products only, cohort TCOC was inside the safe zone. In 1440 (86.0%) of the 1674 simulation runs in the safe zone with reference products only, alternative use of biosimilars reduced the cohort TCOC to being under the TCOC target. A total of 234 (14.0%) of the 1674 simulation runs in the safe zone with reference products remained in the safe zone after biosimilar substitution.

The 3886 (38.9%) simulation runs with cohort TCOC under the TCOC target with reference products all remained under the TCOC target after biosimilar substitution.

We analyzed the primary study outcome by OCM cancer type for the 12 cancer types comprising 2% or more of the study population (Table 4). We found that biosimilar substitution was most beneficial in OCM cancers where treatment with the BAP antineoplastics was prevalent and had higher proportions of treatment-naïve patients. In breast, comparatively high survival rates and expected long-term treatment with trastuzumab meant that only 31.8% of the study population with trastuzumab use were treatment naïve at the beginning of the episode. Conversely, lymphoma episodes with use of rituximab were more often treatment naïve (60.7% of the study population with rituximab use), leading to significantly higher biosimilar substitution impact per episode ($3700 cost reduction per episode for lymphoma vs. $730 for breast).

Table 4 Base case model: primary outcome—average reduction in episode total cost of care relative to Oncology Care Model (OCM) benchmark price, by cancer type
Full size table

OCM cancer types where episodes primarily received BAP supportive therapies without BAP antineoplastics tended to have lower aggregate cost reduction relative to benchmark attributable to biosimilar substitution (female genitourinary cancer other than ovarian), or even cost increases relative to benchmark attributable to biosimilar substitution (pancreatic, prostate). During the study time period (second half 2021), OCM costs for pegfilgrastim and epoetin alfa were such that biosimilar substitution of these agents resulted in increased costs relative to benchmark versus the use of reference products and cancers in which biosimilar utilization was primarily associated with one or both of these agents resulted in increased average TCOC. Biosimilar substitution of filgrastim led to significantly reduced per-unit costs associated with the drug.

Discussion

Our findings estimated that adoption of biosimilars can lead to $1193 in TCOC savings per biosimilar applicable 6-month oncology care episode. We estimate that this shift in TCOC from use of biosimilars would increase the proportion of cohorts of biosimilar applicable episodes receiving performance-based payment bonuses from 38.87% to 55.17%, and reduce the proportion of cohorts owing recoupment payments for exceeding TCOC benchmarks from 44.40% to 28.99% under the terms of the Medicare OCM. Biosimilar substitution resulted in an average reduction in TCOC equivalent to 2.4% of the benchmark price, which is large enough that it is nearly equal to the TCOC reduction required to go from owing a recoupment to receiving a performance-based payment bonus in the OCM two-sided risk model (2.5%) [6]. These findings demonstrate the potential importance of biosimilars for providers managing TCOC as oncology payment in Medicare and other health plans increasingly transition from FFS to VBP. Our results have demonstrated that biosimilars can act as one of the key tools to enable providers to manage risk and improve financial performance.

Our findings are consistent with prior research. Across multiple studies, the adoption of oncology biosimilars over the reference product has shown significant savings with respect to cost expenditures [23, 53, 54]. In one study, researchers found that biosimilar substitution between July 2020 and April 2021 at a community oncology practice for rituximab, bevacizumab, and trastuzumab resulted in 27%, 19%, and 18% savings, respectively [16]. However, these studies estimated cost savings primarily from the perspective of the payer through the use of budget impact methods, often considering a limited set of biosimilars for substitution [23, 55, 56]. Our findings expand the literature by specifically modeling risk and financial performance from the perspective of provider participants in a particular VBP model and looking at the impact of multiple biosimilar substitutions as a portfolio of interventions. By filling these gaps in the existing literature, our study may better inform evidence-based decision-making related to biosimilars on the part of providers participating or considering participation in VBP models and reduce barriers to biosimilar adoption [57,58,59,60].

In addition to insights relevant to providers, our findings may have important implications for policymakers and payers. The Department of Health and Human Services (HHS) as well as the Centers for Medicare and Medicaid Innovation (CMMI) recently announced drug pricing strategies prioritizing subspecialty pharmaceuticals in oncology for future payment reform efforts [61, 62]. Biosimilars are an increasing source of focus for policymakers [63]. The most recent OCM Evaluation Report found statistically significant differences between adoption of biosimilars from OCM participants compared to non-OCM participants; however, these differences were small [56]. The evaluation report only looked at biosimilars for pegfilgrastim and biosimilars in this context were used only 57% of the time, suggesting that the impact on risk mitigation of biosimilar substitution may not have been clear to OCM provider participants. Our findings suggest the benefits of biosimilar adoption can be high impact and can act as critical tools to mitigate risk and improve financial performance for providers. Aligning provider perspectives on risk mitigation from biosimilars with the intent of policymakers is likely critical to successful design and implementation of future VBP models.

Limitations

Our model has several limitations. First, we chose Medicare’s OCM as the foundation for our analyses and, as such, results may not directly generalize to other VBP models. We feel the trade-off for increased focus is warranted because without modeling a specific VBP initiative, quantifying how biosimilars impact the risk transferred to providers is not feasible. OCM is the largest oncology VBP initiative implemented to date in the country and, as CMMI’s first major cancer model, it will likely influence future cancer models. Importantly, much of the published payment methodology for Medicare’s EOM is consistent with the methodology used in OCM.

Second, we assumed providers would implement biosimilars as a portfolio with uniform therapeutic substitution for all reference products in which there was a corresponding biosimilar. The economic impact might differ according to different adoption strategies.

Third, our study estimated the impact of biosimilar substitution on episodes initiating in OCM’s final performance period (second half of 2021). As such, OCM costs for BAPs were calculated using Medicare’s Q4 2021 ASP Drug Pricing File. Projecting our results forward in time may result in underestimates of the impact of biosimilar substitution, as the price of biosimilars relative to reference agents may decline further over time.

Fourth, we modeled the substitution of all biosimilars available by July 2, 2021—the date of initiation for the last OCM Performance Period. However, the most recent LDS data available ran through the end of 2020 creating a 6-month gap. It is unlikely that the rates of utilization for reference products would have changed significantly over this 6-month period and no significant new indications emerged during this gap that would have high enough prevalence to significantly alter our findings.

Fifth, the study was scoped to episodes in which one or more reference biologic products were applicable. These episodes likely represent a minority of an OCM provider’s population of episodes. Conclusions about overall provider risk and performance in value-based models are dependent on performance across the entirety of a provider’s episodes. While this issue limits the generalizability of our findings for our secondary outcome, it does not impact the generalizability of our primary outcome.

Conclusion

Our simulation study, using observed Medicare claims, demonstrated that oncology biosimilars can play an important role for US providers in managing TCOC in the context of value-based payment models like the Medicare OCM. Substitution of biosimilars for the reference products for bevacizumab, rituximab, trastuzumab, epoetin alfa, filgrastim, and pegfilgrastim reduced TCOC to an extent that 16.31% more cohorts of applicable episodes would qualify for performance-based bonus payments and 15.41% fewer cohorts of applicable episodes would owe recoupments for exceeding TCOC benchmarks. Our findings can inform strategy for oncology providers attempting to manage TCOC while using biologic therapies in the context of value-based payment models in the USA.