Context: When nonretail pharmacy sales exceed 70% of sales, manufacturers of infused, injected, implanted, inhaled, or instilled (5i) drugs are required to calculate average manufacturer price (AMP) under a different methodology than that used for drugs predominantly distributed through retail channels. Specifically, the modified methodology includes pharmacy benefit manager (PBM) rebates in the calculation of AMP for 5i drugs. The modified methodology reduces manufacturers' Medicaid rebate liability and increases net costs to the Medicaid program.
Methods: The authors identified 15 5i drugs predominantly dispensed through the nonretail setting. Using 2013–2017 data from Medicaid, Medicare, SSR Health, and 340B program eligibility, they estimated differences in AMP, Medicaid rebates, and net Medicaid costs under both the standard and 5i AMP methodologies.
Findings: AMP was 42% lower, on average, under the 5i methodology than under the standard methodology. From 2013–2017, Medicaid rebates under the 5i methodology were 82% lower than under the standard methodology, resulting in manufacturers of these 15 drugs reducing their Medicaid rebate liability by $1.1 billion in five years.
Conclusions: Inclusion of PBM rebates in the calculation of AMP for 5i drugs significantly reduced Medicaid rebates, resulting in higher Medicaid spending. This may incentivize manufacturers to shift sales to nonretail channels. To remove this incentive, policy makers should consider excluding PBM rebates from the calculation of AMP for 5i drugs.
The Medicaid Drug Rebate Program (MDRP) was established in 1992 to mitigate the costs of states choosing to expand their Medicaid programs to include prescription drug coverage. The MDRP has, at its core, a trade-off between pharmaceutical manufacturers and state Medicaid programs: pharmaceutical manufacturers are required to issue rebates to Medicaid programs that substantially reduce the price of pharmaceuticals under the condition that Medicaid programs cover all outpatient drugs approved by the Food and Drug Administration. Under this arrangement, Medicaid programs reimburse pharmacies at market rates for the cost of a drug; then, on a quarterly basis, they invoice pharmaceutical manufacturers for rebates to offset the amount paid to pharmacies.
For branded drugs, the Medicaid rebate is calculated as the sum of two components: (1) the base rebate, which is the greater of a 23.1% discount off market prices or the “best price” offered to any purchaser; and (2) an inflation rebate, which offsets any price increases above general inflation since product launch. To calculate the base rebate, Medicaid uses a standard called “average manufacturer price” (AMP), which estimates the general market price at which Medicaid is reimbursing pharmacies and providers (Medicaid 2021). For drugs primarily distributed through retail pharmacies, AMP is calculated as the average price of manufacturer sales to wholesalers and retail community pharmacies (DHHS 2006). This average excludes fast-growing manufacturer rebates to pharmaceutical benefit managers (PBMs) (Hernandez et al. 2020), which are separately used in the calculation of the “best price” portion of the rebate (DHHS 2006).
However, in a statutory quirk, for drugs that are not generally distributed through retail community pharmacies the calculation of the AMP does include rebates to PBMs, meaning that rebates are calculated from net prices. Therefore, for the subset of drugs not generally distributed through retail community pharmacies, PBM rebates are included in two places in the rebate calculation: both the AMP calculation and the best price provision. These products include infused, injected, implanted, inhaled, or instilled drugs, often referred to as “5i drugs,” which have less than 30% of sales volume to retail pharmacies and wholesalers that distribute to retail pharmacies (DHHS 2016). While many of these drugs are provider-administered and are generally not subject to significant postsale PBM rebates, certain formulations of common injectable and inhaled drugs (such as insulins and preventive inhalers) are heavily distributed through mail-order pharmacies or long-term care facilities.
The inclusion of PBM rebates in the calculation of both AMP and best price for 5i drugs, which we call the “5i loophole,” leads to a reduction in Medicaid rebates, thereby increasing total Medicaid spending on pharmaceuticals. Consider a hypothetical $100 drug with an average $30 PBM rebate to commercial purchasers. The manufacturer sells the drug to pharmacies for $100, insurers (and Medicaid) reimburse the pharmacy $100 plus a dispensing fee, and the manufacturer pays the PBM a $30 rebate. Under the standard AMP methodology, AMP would be calculated as the average price to retail pharmacies, here $100. The Medicaid base rebate would be $30, because the difference between the AMP and the PBM net price ($100 – $70 = $30) is greater than 23.1% of AMP ($23.10), triggering the “best price” requirement (DHHS 2019). Therefore, the net price would be $70. However, if 70% of sales took place through mail-order pharmacies, the 5i AMP methodology would apply, and the PBM rebates would be included in the AMP calculation, resulting in an AMP of $70. Because the AMP minus the PBM net price ($70 – $70 = 0) is less than the Medicaid standard rebate (23.1% * $70 = $16), the PBM rebate does not trigger the best price requirement. Therefore, the Medicaid base rebate would be 23.1% of AMP, or $16. Under this scenario, Medicaid would reimburse the mail-order pharmacy $100 (reflecting the pharmacy's acquisition cost) but only receive a rebate of $16, resulting in a net price of $84. This is $14 higher than the net price of $70 estimated under the first scenario, when rebates were excluded from the AMP calculation.
The increased cost to Medicaid under 5i AMP is magnified when rising prices and PBM rebates are considered, as Medicaid is supposed to receive an additional rebate when market prices, represented by AMP, increase faster than inflation. Drug manufacturers frequently offset list price increases with additional PBM rebates to limit net cost growth to insurers. Assume that our hypothetical drug increased in price in one year to $110, PBM rebates increased to $38, and general inflation is 2%. Under the standard AMP methodology, AMP would be $110. Medicaid would receive a base rebate of $38, because the best price provision would apply again, since $38 is greater than 23.1% of $110. The inflation rebate would be $8, because the new AMP is $110, and the inflation-adjusted AMP would be $102 (given 2% general inflation). This would result in a net price of $64 ($110 pharmacy reimbursement – $38 base rebate – $8 inflation rebate). Under 5i AMP methodology, however, the AMP is $72 ($110 list price – $38 PBM rebates). Medicaid base rebate would be 23.1% of AMP, or $16. The inflation rebate would be $72 – $70 * 1.02, or $0.60. This would result in a net price of $93.40 ($110 pharmacy reimbursement – $16 base rebate – $0.60 inflation rebate), or $29 higher than under the standard AMP methodology.
To our knowledge, no study has estimated how AMP and Medicaid rebates for 5i drugs would differ if concessions to PBMs were excluded from AMP. Our objective was to estimate the differences in Medicaid rebates and net Medicaid costs if PBM rebates were excluded from AMP calculations for a set of 5i drugs predominantly dispensed through the nonretail setting with sufficient data to estimate these differences. We use the methodology for 5i AMP calculation established by the Centers for Medicare and Medicaid Services in 2016 and as described above. Aside from the Medicaid impacts, the use of 5i AMP could also affect the Medicare program, as the Elijah Cummings Lower Drug Costs Now Act based Medicare Part D inflation rebates on AMP.
We used the following data sources: (1) the Food and Drug Administration Orange and Purple books; (2) monthly wholesale acquisition costs, which represent list prices and were obtained from AnalySource (reprinted with permission by First Databank, Inc.) (FDB 2019); (3) net drug sales and total units obtained from SSR Health (SSR Health 2019); (4) Medicare Parts B and D claims for a 5% random sample of Medicare beneficiaries; (5) Medicaid state drug utilization data (Medicaid 2020); (6) Medicare Part D prescriber utilization files (CMS 2020); (7) Medicare Part D, Part B, and Medicaid spending dashboards; and (8) a Health Resources and Services Administration list of 340B-eligible institutions (HRSA 2020).
We selected the study sample in six steps. First, we used the Medicaid spending dashboard to select products with more than $500,000 in spending in 2018. Using the dosage form from the Orange and Purple Books, we identified infused, injected, implanted, inhaled, or instilled products (n = 661 products) (figure 1). Second, using the Orange and Purple Books, we calculated the number of quarters that a product was marketed in 2013–2018. We extracted Part D claims in 2013–2018 and selected products with at least 60% of Part D units accounted for by nonretail pharmacies (including mail order, specialty, hospital, and long-term care pharmacies, among others) in at least half of the quarters marketed (n = 249 products). While 5i AMP eligibility is triggered when 70% of sales are through nonretail pharmacies, the Medicare Part D program requires plans to ensure that beneficiaries can access drugs through retail pharmacies within a certain driving distance (DHHS 2011). Therefore, Part D retail percentage may be higher than overall retail percentage. Third, we constrained sampling to brand-name products that had SSR health data available (n = 75 products). Fourth, we excluded products that had a larger number of Medicare Part B users than Part D users. This is because physician-administered drugs generally do not have PBM rebates (Visante 2020), so these drugs would be unlikely to have significant differences between standard and 5i AMP. Fifth, we excluded products that were marked as outliers in Medicare Part D data or that had unstable SSR Health estimates, as previously done in the literature (Hernandez et al. 2020). Finally, we constrained sampling to products that had at least three years of consecutive data available across all data sources. The final sample included 15 products.
The primary outcomes included (1) estimated Medicaid rebates under the 5i AMP methodology (inclusive of PBM rebates); (2) estimated Medicaid rebates under the standard AMP methodology (exclusive of PBM rebates); and (3) the difference between the first two outcomes, which represents the reduction in Medicaid rebates associated with inclusion of PBM rebates in the calculation of AMP.
We estimated Medicaid rebates under the 5i rule (outcome #1) as Medicaid units * (23.1% * 5i AMP + inflation penalty). In this case we assumed PBM rebates would not trigger best price because 5i AMP approximates the net price after PBM rebates. Medicaid rebates under standard AMP (outcome #2) were estimated as Medicaid units* (PBM rebate + inflation penalty), if the PBM rebate triggered best price. If the estimated PBM rebate was lower than 23.1% of regular AMP, Medicaid rebates under regular AMP (outcome #2) were estimated as Medicaid units * (23.1% * AMP + inflation penalty). Medicaid rebates were capped at the respective AMP, per statute.
Estimation of AMP and PBM Rebates
We used invoice price as a proxy for standard AMP, as previously done (Dickson and Reynolds 2019). The online appendix explains the derivation of formulas for the calculation of PBM rebates and of 5i AMP, which were calculated using the following parameters: net sales, total units, Part B units, Part D units, Medicaid units, invoice price, units sold to 340B-eligible institutions, and the inflation rebate (input parameters).
In brief, the invoice price was estimated as the weighted average reimbursement in Medicare and Medicaid per unit of product and was calculated using annual spending per product from the dashboards and the standardized Part B units, Part D, and Medicaid units. Total sales and units were obtained from SSR Health (SSR Health 2019). SSR Health is an investment firm that estimates net prices from publicly reported drug sales and the number of units sold in the United States obtained from Symphony Health, which captures 92% of prescriptions dispensed across the United States, including sales to nonretail institutions (Symphony Health 2019). Medicare Part B, Part D, and Medicaid units were obtained from the spending dashboards and were standardized because units are expressed in different quantities across data sources (details in the online appendix). The Medicaid inflation rebate penalizes increases in drug prices above inflation and was calculated as the difference between the invoice price and a utilization-weighted inflation-adjusted launch price, which was calculated each year of the study period (details in the online appendix).
From 2013 through 2017, across the 15 study drugs, 5i AMP was 42% lower than standard AMP on average, ranging from 28% in 2013 to 52% in 2016 (table 1). For the set of 15 study drugs, we estimate that 5i Medicaid rebates over the period totaled $243.0 million. Had the AMP been calculated under the standard AMP methodology, Medicaid rebates over the period would have been $1.3 billion. This means manufacturers avoided paying $1.1 billion in Medicaid rebates over the period for the selected 15 drugs, reducing their rebate liability for these drugs by 82%. The difference in rebates calculated under each methodology is equivalent to 6.7% of Medicaid spending on the selected sample.
Of the 15 study drugs, hormones accounted for 49% of avoided rebates, or $424.1 million. Respiratory therapies were the second-largest group of avoided rebates, totaling $304.9 million (28% of avoided rebates). Of the individual drugs, Genotropin led the avoided Medicaid rebates ($346.7 million), followed by Pulmicort ($302.3 million) and Lovenox ($206.8 million). From 2013 through 2017, 82% of claims for the study drugs were dispensed through long-term care pharmacies, 8% from retail pharmacies, 8% through managed care organization pharmacies, and 2% through mail-order pharmacies (table 2).
Figure 2 illustrates the differences in AMP and Medicaid rebates per unit under the standard and 5i methodologies for Humira in 2017, the drug with highest Medicaid expenditures that year. Under the standard AMP methodology, AMP reflects the invoice price ($2,205 per unit), and Medicaid rebates include a base rebate of $509 per unit (23.1% of AMP) and an inflation rebate of $1,451 per unit, totaling $1960 per unit. Under the 5i methodology, PBM rebates reduce AMP to $1,699 per unit and thus the base rebate to $392 per unit (23.1% of $1,699). Lower AMP under the 5i methodology reduces the inflation penalty ($945 as opposed to $1,451); total rebates are $1,337 per unit, or $623 less per unit than under the standard AMP methodology.
Figure 3A illustrates differences in AMP per unit under the standard and 5i methodologies for Humalog over the study period, the drug with the second-highest Medicaid expenditures in our study. Standard AMP for Humalog increased from $16.20 per unit in 2013 to $30.40 per unit in 2017. In parallel, PBM rebates increased from $5.80 per unit to $22.50 per unit. Including PBM rebates in 5i AMP caused 5i AMP to decrease over the study period, from $10.60 per unit in 2013 to $9.10 per unit in 2017. Commercial net prices followed a similar downward trend, from $10.30 per unit in 2013 to $9.30 per unit in 2015 and $8 per unit in 2017. Importantly, this represents a 13.9% decrease in net price in 2015–2017, which is consistent with reports by the drug's sponsor (Lilly 2019). Figure 3B illustrates how these pricing dynamics with increasing PBM rebates reduce Medicaid rebates under the 5i methodology. In 2013, the Medicaid rebate per unit of Humalog was $8.10 lower under the 5i methodology ($8.10 vs. $16.20). In 2017, the Medicaid rebate per unit of Humalog was $25.50 lower under the 5i methodology ($4.90 vs $30.40).
To our knowledge, our study is the first to quantify avoided Medicaid rebates for drugs that meet the 5i criteria. Our findings demonstrate that calculating AMP based on net price (including PBM rebates) rather than list price (excluding PBM rebates) substantially reduced Medicaid rebates and increased Medicaid net costs. While the likely majority of 5i AMP drugs are physician-administered drugs without significant PBM rebates, the growth of mail-order pharmacy for self-administered drugs will likely result in greater avoided Medicaid rebates under the current 5i AMP methodology. The significantly lower rebates under 5i AMP may incentivize manufacturers to shift distribution to nonretail channels to take advantage of 5i AMP, creating a 5i loophole that deprives Medicaid programs of significant rebate dollars used to offset ever-growing drug prices.
Insulin manufacturers may be particularly incentivized to shift distribution to nonretail channels, such as mail-order or long-term care pharmacies, given the significantly lower Medicaid rebate liability faced and the increasingly lower net prices realized by insulin manufacturers (Hernandez et al. 2020). As illustrated in figure 3, Medicaid rebates under the standard AMP methodology offset increases in invoice price. However, the lower rebates under the 5i methodology do not offset price increases, resulting in a greater net cost to Medicaid. While only vial formulations of Humalog currently appear to meet the 5i nonretail distribution threshold, the significant difference in revenue could encourage the manufacturer to move additional formulations to the nonretail channel. Because the 340B drug discount program uses the same methodology as Medicaid to calculate prices, manufacturers with significant 340B exposure would have additional incentive to shift sales to the nonretail channel, raising the 340B price for a drug eligible for the 5i methodology.
Under the Inflation Reduction Act of 2022 (IRA), AMP will be used to calculate an inflation penalty for the Medicare program beginning in October 2022. Combined with growth under both the Medicaid program and the 340B program, manufacturers will now have substantial incentives to take advantage of the “5i loophole” to reduce their mandatory price concessions. A manufacturer with a 5i drug that currently uses standard AMP may avoid future inflation penalties by shifting distribution to nonretail channels to allow use of the 5i AMP calculation, limiting inflation penalties while continuing to take list price increases. While lawmakers did not address 5i AMP in the IRA, rectifying the 5i loophole would be a relatively easy task, as the 5i AMP definition would only need to be slightly modified to exclude PBM rebates from being included in both the AMP calculation and the best price calculation. This would address the 5i loophole while still including sales outside of retail community pharmacies, allowing manufacturers to calculate an AMP for drugs with limited retail pharmacy sales.
Our study is subject to seven main limitations. First, because AMP is a proprietary calculation, we must estimate AMP and subsequent Medicaid rebates based on available data. However, our estimates for the net price of Humalog closely mirror those publicly reported by the manufacturer in congressional testimony (Mason 2019), demonstrating validity (figure 3). Second, because we do not have access to full manufacturer sales channel data, we used Medicare Part D data to estimate the proportion of nonretail sales for each drug. This could have underestimated the proportion of nonretail sales, since units administered at physician offices were not included. To partially mitigate this issue, we allowed in the sample products with at least 60% of Part D units accounted for by nonretail pharmacies in at least half of the quarters marketed, as opposed to the 70% threshold defined by the statute (DHHS 2011). Still, our estimates represent a conservative estimation of difference in Medicaid rebates under two methodologies, also because a large number of drugs were excluded because of unavailability of SSR Health data. Third, we estimate average price concessions at the product level, rather than at the National Drug Code (NDC) level, because the net pricing data from SSR Health are available at the product level. However, not all NDCs for each product met the 5i criteria. For instance, only one NDC of Humira met these criteria (00074–9374–02). This NDC corresponds to a syringe kit of low-dose Humira (20mg/0.4ml), which only accounts for 0.6% of Humira sales in part D. Because we estimate average PBM rebate for the product, we likely underestimate true Medicaid rebates under standard AMP, as any individual PBM rebate above the average would establish a higher best price and thus a greater Medicaid rebate. Fourth, as a conservative measure, we estimated Medicaid rebates at the NDC-11 level rather than the NDC-9 level that manufacturers do, which may underestimate the difference in Medicaid rebates under two methodologies (as the NDC-9 approach averages AMP across all NDC-11s that share the same NDC-9 and assesses the rebate across all products, potentially magnifying the impact of any difference). Fifth, we assumed that manufacturers did not provide PBM rebates when launching drugs, which could have underestimated inflation rebates. Sixth, in our analyses we used SSR Health data, which obtains volumes of sales from Symphony Health. Symphony Health data capture 92% of prescriptions dispensed across the United States (Symphony Health 2019); however, it is possible that the company's coverage of nonretail channels is not as comprehensive as its coverage of retail channels. This could have impacted our estimates, since our sample focused on drugs predominantly distributed through nonretail channels. Seventh, we used prescriber information to identify units subject to 340B discounts, and we assumed that all units prescribed at 340B entities were subject to 340B discounts. This may have resulted in an overestimation of the proportion of units subject to 340B discounts.
We believe that while these limitations may affect the magnitude of reduced Medicaid rebates reported, they do not affect the overall presentation of the unintended consequences of the 5i loophole. Additional analyses from institutions with access to proprietary AMP data, such as the Congressional Budget Office or the Department of Health and Human Services Office of Inspector General, are necessary to estimate the impact of the 5i loophole with greater accuracy.
Drug manufacturers are estimated to have reduced their Medicaid rebate liability by $1 billion in 2013–2017 for a set of 15 drugs because of the inclusion of PBM rebates in the 5i AMP calculation methodology. The continued inclusion of PBM rebates in the 5i AMP methodology may encourage manufacturers to shift their sales to nonretail channels to reduce their Medicaid liability, an incentive we call the 5i loophole. To discourage this practice and avoid further reductions in Medicaid rebates, policy makers should consider legislation that excludes PBM rebates from the 5i AMP calculation methodology.