Estimated savings from year two of the IRA Prescription Drug Negotiation Program

The Department of Health and Human Services announced the 15 drugs selected for the second round of negotiations that would apply to price-year 2027.¹ Table 1 lists the selected drugs, the year they were approved, the conditions they treat, the number of Medicare enrollees in Part D using each drug, and the gross Part D spending during calendar year 2024. The table shows that those drugs account for nearly $42.5 billion in gross Part D drug spending on behalf of 5.7 million Medicare beneficiaries.²

This paper provides estimates of the expected savings from the negotiation process established by the Inflation Reduction Act (IRA). The Centers for Medicare and Medicaid Services (CMS) has reported that the reduction in spending associated with the new negotiated prices amounts to $12 billion, evaluated at 2024 volumes of prescriptions. Our analysis examines the differences in the prices that resulted from negotiation by Part D plans and those obtained through the IRA’s negotiation program for the 15 drugs selected for those negotiations. We make use of publicly available data as the detailed net price measures for the pre-IRA negotiation period, which would yield more precise estimates, are confidential. We therefore combine data from multiple sources to approximate pre-IRA net prices to estimate pre-period spending. We then make use of the announced Maximum Fair Prices (MFPs) and apply them to recently reported volume for each drug in calendar 2024 to obtain expected spending for the 15 drugs at the negotiated prices. In the next section, we briefly explain our estimation methods, followed by a presentation of our estimates. The last section discusses the estimates and important issues in interpreting the numbers.

Estimation approach

We estimate the net spending by Medicare for the drugs selected for negotiation. We do so by first reviewing and combining information from several data sources using the general methods in our estimated savings for 2026.³ They include data on sales from the CMS Drug Spending Dashboard and the CMS Fact Sheet on Negotiation, rebates by therapeutic class reported by MedPAC⁴ and the Government Accountability Office (GAO)⁵, and information on the relationship between Wholesale Acquisition Costs (WAC) and the Medicare Part D gross sales process developed by the Congressional Budget Office. As a check against these data, we also examined SSR Health data on net prices for the selected drugs and reports on the gross-net sales difference reported in the literature. To establish estimates of net prices pre-IRA, we use both MedPAC and GAO reporting on pre-IRA rebates by therapeutic class. Note that GAO organizes the therapeutic categories for rebates differently from MedPAC and so offers some opportunities for greater precision in estimation of rebates, albeit for calendar year 2021. Because SSR Health estimates incorporate a variety of revenue adjustments unrelated to manufacturer rebates—like discount coupons and other consumer discounts and include purchasers like Medicaid and 340B that receive large statutory rebates—we make use of the MedPAC maximum and GAO average rebates by therapeutic class for the Part D program for each drug. Second, we use these data to create an estimate of the net sales prior to the IRA.⁶ We do so by applying the pre-IRA net prices to 2024 sales volumes for the selected drugs. By using the upper end of the rebate range estimate for prices reported in the MedPAC report, we believe that we are taking a conservative approach (resulting in lower net sales estimates in the pre-period). This is especially the case because several of the drugs selected for negotiation are specialty drugs⁷, and those have been shown to have lower than average rebates for brand-name drugs.⁸ Finally, the assembled data allow us to obtain drug-specific pre-IRA Medicare Part D net spending estimates. Those spending estimates are then compared to spending estimates that result from applying the negotiated MFPs to the 2024 volume data.

Results

Table 2 reports key data elements that were used in estimating sales resulting from negotiated prices. The table reports the following data elements for each of the 15 selected drugs:

• Gross Part D sales for the pre-IRA baseline (calendar year 2024)
• Pre-IRA manufacturer rebates to Part D plans as a percentage of gross sales
• Classification of each drug as a long or short monopoly drug
• The estimated net of manufacturer rebate sales for each drug in the pre-IRA baseline period
• The discount off of Part D gross drug costs at the Maximum Fair Price (MFP)
• Estimated sales based on reported MFP under the IRA
• Estimated savings from the pre-IRA baseline.

The estimated impact on the government negotiations compared to sole reliance on prescription drug plans for the 15 drugs was estimated at $12.5 billion, a figure close to the CMS estimate. On average, that meant that the average MFP was 44% below the pre-IRA price net of manufacturer rebates. As Table 2 highlights, 43% of the estimated savings, or $5.38 billion, is accounted for by five drugs (Trelegy Ellipta, Xtandi, Pomalyst, Ibrance, and Linzess). Several drugs had large rebates in the pre-IRA period (e.g., Trelegy Ellipta, Breo Ellipta, Tradjenta, Janumet), as noted in the third column of Table 2. It is therefore notable that the negotiations resulted in prices that were likely to be below the statutory ceiling price for those products. For example, Xtandi saw a 40-percentage-point increase in the net of manufacturer rebate price due to negotiation. The table also reveals that the largest price concessions were obtained for drugs that had the lowest pre-IRA rebates because of limited pre-IRA competition. This is the case for Xtandi and Pomalyst.

Comments on interpretation

The estimates presented represent savings stemming from the differences in sales based on prices negotiated by prescription drug plans pre-IRA and prices negotiated by the government due to the IRA’s creation of the drug negotiation program. The estimates are not meant to provide a full budgetary impact of the IRA but to focus on how allowing the government to negotiate affects prices relative to relying solely on Part D prescription drug plans.

The estimated percentage of savings off of list price from 2027 prices was roughly double those for 2026. It appears that the increased savings stem from obtaining larger price concessions for drugs that faced significant therapeutic competition, more drugs in “Protected Classes” (four oncology drugs, one drug for mental illnesses), and the larger number of specialty drugs being negotiated. In that sense, the estimates show that the government negotiations are especially significant for drugs where market forces were most limited and therefore had the least impact on producing price concessions. This was the intent of the policy design.

Finally, because the estimates rely on publicly available data and not the actual rebates and non-federal manufacturer prices, they are meant to provide an approximation of how the $12.5 billion in savings is distributed across the 15 selected drugs. Our methods of analysis are similar to those used by CMS, and so it is notable that the reliance on publicly available information leads us to roughly the same estimates of aggregate savings and the percentage reduction in net spending as those reported by CMS. Like CMS, we also take into account the coverage gap discount program. Our estimate is that the discount program reduced 2024 net spending on the 15 negotiated drugs by $3.4 billion. That implies that savings under the negotiated prices (MFP) relative to prices negotiated by Part D plans are estimated at $12.5 billion and $9.1 billion after accounting for the coverage gap discounts.⁹

Appendix

Appendix A

Assumptions used for estimating rebates for Part D drugs based on average Part D rebates reported by therapeutic class (MedPAC and GAO reports).

1. Use the high end of the range in the MedPAC therapeutic class if there is an exact match on therapeutic class [Three drugs: Breo Ellipta, Trelegy Ellipta, and Otezla].¹⁰
   1. Exception: For the three drugs that treat diabetes, we instead combined rebate data results from GAO and MedPAC (even though the GAO rebate data is for a broader therapeutic class—endocrine metabolic agents). MedPAC shows rebates >= 50% for drugs that treat diabetes. GAO shows rebates of 47% for Endocrine Metabolic Agents. Last year, we assumed 60% rebates for drugs that treat diabetes based on a Senate Finance Committee report that showed rebates in this range (for insulin drugs, insulin products were in the 2026 cohort). This year, we are using the lower end of the MedPAC range for two diabetes drugs, Tradjenta and Janumet, which are not insulin products. For Ozempic/Wegovy, we use the GAO broad therapeutic class (endocrine metabolic agents), which shows 47% rebates (see also #3 below).
2. Combine MedPAC and GAO data together:
   1. MedPAC and GAO both have an antineoplastic class. GAO point estimate is 2% and MedPAC range is < 10%. Compromise by using 5% estimate on rebates for antineoplastics. [Four antineoplastic drugs in the 2027 cohort].
   2. Another exception was antipsychotics. MedPAC said < 10% for antipsychotics, and GAO provides an estimate for central nervous system agents of 8%. Because antipsychotics are a protected class, and because GAO provides a point estimate for central nervous system agents, we used the lower GAO estimate of 8%. [One drug: Vraylar].
3. Use GAO rebates by therapeutic class. If there is no match on the MedPAC therapeutic class, then use the GAO therapeutic class (which provides a point estimate on average rebates for what is usually a broader therapeutic class than MedPAC). [Ozempic/Rybelsus/Wegovy, Ofev, Linzess, Austedo, Xifaxan].

Assumptions Used to Estimate Manufacturer Discounts

We used the MedPAC report, which gives the total amounts collected in coverage gap discount for each of the 15 top therapeutic classes in Part D, as well as total gross spending in the therapeutic class and the brand share of gross spending by therapeutic class.

For example, in the case of Diabetic therapies, coverage gap discounts totaled $8.2 billion, total spending on brand-name and generic drugs in the therapeutic class was $60.7 billion, and brand share of spending in the therapeutic class was 98% (see Charts 10-24 and 10-25 in the MedPAC report). Therefore, coverage gap discounts as a share of gross spending on brand-name drugs in this therapeutic class were estimated as ($8.2 billion/($60.7 billion)*98%) = 13.8% of gross spending.

For drugs in the 2017 cohort that did not fall within a therapeutic class included in the MedPAC report, we assumed that the coverage gap discounts as a share of gross spending was similar to another drug in the 2017 cohort that had a roughly similar level of annual spending per beneficiary (and for which we were able to calculate the coverage gap discount as a share of gross spending based on data in the MedPAC report). This assumes that drugs with a similar level of annual spending per beneficiary will have a similar share of gross spending that falls within the coverage gap discount phase of the Part D benefit.