We identified state-specific Medicaid coverage guidelines for nusinersen and onasemnogene abeparvovec using targeted web searches. Two researchers (IE and BK) independently searched state Medicaid websites for coverage guidelines from February 1, 2020, through March 16, 2020. We focused on fee-for-service (FFS) Medicaid state websites rather than those for third-party managed care organizations (MCOs). This choice was motivated by our belief that it would be easier for state Medicaid directors to implement policy change for select drugs in the FFS domain compared with implementing a policy change in the MCO domain, which typically involves established contracts that manage the MCO.

Once a coverage policy was identified, it was independently documented using an agreed-upon protocol (see Supplementary Material, available in online article). The documentation process included categorizing key coverage attributes and recording them in a master file. The 2 independently created master files with coverage attributes were then compared, and differences were resolved by a third researcher (JB), using a consensus-driven adjudication process. Comparison of coverage guidelines between the 2 genetic therapies were statistically analyzed using chi-square tests.

Additional information on state Medicaid programs was collected from restricted-access and publicly available data as follows. State Medicaid drug utilization for nusinersen and onasemnogene abeparvovec was estimated using the branded drug names in restricted-access state Medicaid drug utilization files.⁹ State Medicaid births were estimated using the Centers for Disease Control and Prevention’s (CDC) National Vital Statistics System and March of Dimes data.^10,11 The CDC data provided total births by state, and the March of Dimes provided the proportion of births for whom Medicaid was the primary payer. We then estimated the proportion of these births to be allocated to FFS Medicaid based on the proportion of FFS Medicaid enrollees compared with total Medicaid beneficiaries in each state.¹² These population statistics yielded estimates of total FFS Medicaid births. We then estimated nusinersen utilization per 10,000 Medicaid FFS births using the Medicaid drug utilization data. We concluded that utilization estimates were premature for onasemnogene abeparvovec, given that it was approved in May 2019,¹³ which left only 7 months of observable utilization data in our dataset.

We also compared our sample state Medicaid characteristics with out-of-sample state Medicaid characteristics using 2 metrics: (1) whether a state expanded Medicaid and (2) per child Medicaid spending.^14,15 We used a Fisher exact test to compare Medicaid expansion in our sample with out-of-sample states, and a Wilcoxon rank sum test to compare sample with out-of-sample per child state Medicaid spending.

We developed a Medicaid restriction score based on 4 coverage guideline components: SMN2 count required, restriction based on ventilator status, baseline motor function required, and whether a specialist was required. These 4 components were selected based on a review of nusinersen and the key clinical considerations for nusinersen utilization.¹⁶ For each of these components, we assigned a binary value based on whether the state Medicaid guideline for that component was restrictive or unrestrictive. Restrictive components were assigned values of one, and unrestrictive components were assigned values of zero. These values were aggregated for each state to create a state’s Medicaid restriction score. We then compared restriction scores to nusinersen utilization per 10,000 Medicaid FFS births.

We identified 19 states with FFS Medicaid that had their coverage guidelines for nusinersen on their websites, along with their corresponding Medicaid characteristics (Table 1). These states represented approximately 166,000 Medicaid births in 2018, with an average per child Medicaid spending of $3,445. States with identified guidelines were spread out geographically and represented many of the more populous states. States in our nusinersen sample were not more likely to be a Medicaid expansion state (73.6% of states in our sample expanded Medicaid vs 70.8% of the out-of-sample states [P = 0.52]) nor did they have significantly different Medicaid spending per child ($3,445 Medicaid spending per child for states in our sample vs $3,626 Medicaid spending per child for the out-of-sample states [P = 0.32]).

The majority (95%) of the 19 states required prior authorization and had detailed prior authorization requirements (Table 2). There was more variation in the coverage policies involving genetic characteristics. Nearly all (95%) states required a formal SMA diagnosis; however, only 14 (74%) of the 19 states required an SMN2 gene count. Of those 14 states, all but Colorado restricted coverage based on the SMN2 count. Two states restricted access based on a patient’s age. Eight states (42%) restricted access based on ventilator status. Seventeen states (90%) had a baseline motor function requirement. Ten (53%) states specified that a child must see a specialist, and 4 (21%) states required a child to see a subspecialist, such as a pediatric neurologist, before gaining access to nusinersen. One state, Wisconsin, deemed eligibility only if a child had no previous treatment using onasemnogene abeparvovec. In a similar vein, the state of Utah deemed that a child was not eligible for nusinersen if the child was receiving concurrent onasemnogene abeparvovec treatment.

We identified 17 states with Medicaid coverage guidelines for onasemnogene abeparvovec on their websites in 2020, along with their corresponding Medicaid characteristics (Table 3). These states represented approximately 156,000 Medicaid births in 2018 with an average per child Medicaid spending of $3,868. States in the onasemnogene abeparvovec sample were not more likely to be a Medicaid expansion state (82.4% of states in our sample expanded Medicaid vs 70.8% of the out-of-sample states [P = 0.41]) nor did they have significantly different Medicaid spending per child ($3,868 Medicaid spending per child for states in our sample vs $3,626 Medicaid spending per child for the out-of-sample states [P = 0.22]).

Of the 17 states with coverage guidelines for onasemnogene abeparvovec, 15 required prior authorization. Nearly all (94%) states required a genetic SMA diagnosis. Eleven (65%) states required SMN2 counts for onasemnogene abeparvovec. Thirteen (77%) states limited access to onasemnogene abeparvovec based on ventilator status. Five (29%) states restricted access to onasemnogene abeparvovec. Approximately half the states required a specialist consultation, and 5 (29%) required a subspecialist. Four (24%) states expressly stated no previous treatment with nusinersen, and 8 (47%) states did not permit concurrent nusinersen treatment. Twelve (71%) states stated specifically no previous treatment with onasemnogene abeparvovec.

States had similar coverage policies for onasemnogene abeparvovec and nusinersen. However, there were some differences. States were more likely to restrict onasemnogene abeparvovec (77%) based on ventilator status compared with nusinersen (42%); this difference was statistically significant (P = 0.049). States were more likely to restrict onasemnogene abeparvovec (29%) based on symptom onset, which was not a coverage guideline for nusinersen. Coverage policies around previoius treatment of either drug were much less restrictive for nusinersen compared with onasemnogene abeparvovec.

We aggregated restriction scores for each state based on our coverage guidelines and gave each state a score of 1 to 4. A score of 1 suggested that a state had relatively less restrictive coverage guidelines, whereas a score of 4 suggested that a state had relatively more restrictive coverage guidelines (Table 4). We treated each coverage restriction equally in our composite measure, since we determined that none of the 4 coverage domains warranted a significantly higher weighting.

We then examined if coverage restrictions were associated with other characteristics of the Medicaid program. We did not find any obvious associations with coverage policies for genetic therapies. For example, there were no perceived discernible patterns across state restrictive scores and state Medicaid program characteristics, such as whether a state expanded Medicaid or average Medicaid spending per child.

Next, we examined whether coverage policies were associated with utilization rates for nusinersen. Our results suggested a correlation such that states with higher restriction scores had lower nusinersen utilization rates per 10,000 Medicaid FFS births. States with the most restrictive scores had an average of 51.9 nusinersen prescriptions per 10,000 Medicaid FFS births compared with states with the least restrictive scores, which had 92.9 nusinersen prescriptions per 10,000 Medicaid FFS births (Table 4).

Four policies that state Medicaid programs could use to address ultra-expensive drugs, such as genetic therapies, in addition to the coverage restrictions already discussed are as follows: preferred drug lists, joint purchase agreements, multistate pooled negotiations for supplemental rebates, and a pooled subscription, commonly referred to as the “Netflix” model.^18,19

A 2012 study using Michigan’s Medicaid experience with the first 3 of these strategies demonstrated that the preferred drug list (PDL) generated the greatest cost savings.¹⁹ However, the PDL strategy generated its savings primarily by switching from expensive drugs to less expensive drugs on a drug list. The use of a PDL strategy may not yield significant savings for state Medicaid programs in the case of genetic therapies because there tends to be limited pharmacotherapy options for the underlying condition. Without competition, either from alternative branded drugs or from generics, the ability for a state to extract rebates, vis-à-vis savings, from manufacturers based solely on placement on a PDL is limited.

A more promising cost containment strategy for genetic therapies and Medicaid programs could be a joint purchase agreement, which occurs when 2 states join together to purchase a drug, thereby (and ideally) increasing the buying power of the states to achieve lower prices. Michigan Medicaid entered into a joint purchase agreement with the state of Vermont in 2003. This agreement resulted in lower brand-name drug prices, although the magnitude of savings was small.

A more robust variation of a joint purchase agreement is a multistate pooled negotiation. As the name implies, this type of agreement includes multiple states, which further increases the monopsony of the buyers. Michigan’s experience suggests that multistate negotiation generated significant savings by negotiating lower unit prices for branded drugs. To implement an effective multistate pooled negotiation cost containment strategy, states should agree on a uniform set of coverage policies for a specific drug.

By agreeing to a uniform coverage policy, the states increase the importance of each coverage domain, which can be leveraged with the supplemental rebate negotiations with drug companies. For example, if states agree on a specific SMA drug coverage policy around a patient’s ventilator status, they could have better bargaining power, and better bargaining power could yield additional rebates from the drug manufacturer.

Drug companies would want a less restrictive policy to ensure that more infants have access to the SMA drug. If all states agree on a less restrictive policy, then they could extract a larger rebate from drug companies for adopting this policy. This is because, in an all-or-none negotiation, a monopsonist (ie, states collectively purchase the drug) is in a better position to achieve lower prices than if the states acted individually.²⁰

In joint purchase and multistate arrangements, there are additional advantages to a more uniformed coverage policy across state Medicaid programs. These types of arrangements reduce the potential inequity for Medicaid patients based on geography and could minimize the incentive for people to move to a different state to seek health care. Another advantage is the reduced administrative burden for the state Medicaid program and the drug company. Currently, states are using resources to develop their own coverage policies for genetic therapies. This requires an extensive review of evidence, which is constantly evolving for genetic therapies. Also, drug companies have administrative expenses associated with coordinating with each individual state for access to their drugs. That said, it is worth noting that many states are moving toward Medicaid pharmacy benefit standardization within each state by requiring Medicaid MCOs to follow the PDL of state FFS Medicaid PDL.⁸

Another state drug purchase arrangement that could help Medicaid programs address the high cost of genetic therapies is a “Netflix” model.²¹ This model is subscription based in that the purchaser (Medicaid program) negotiates with the supplier (drug manufacturer) for unfettered access to a drug using a monthly subscription. This type of model appeals to manufacturers, since it produces a predictable and sustained revenue stream for manufacturers. This model has been successfully implemented by the state of Louisiana for the high-cost specialty drugs needed for hepatitis C treatment.²²

It has also been proposed to combine a multistate purchase agreement with a subscription arrangement for genetic therapies.²³ This approach could provide additional leverage for the buyers in the negotiation of prices with increased monopsony; however, it is contingent on state Medicaid programs agreeing to uniform coverage guidelines.

This study has some limitations to consider. First, our study focused on nusinersen and onasemnogene abeparvovec coverage guidelines that were published on state websites. States may have guidelines that are not available on their websites. However, it is worth noting that websites are perhaps the most accessible form of information for Medicaid enrollees and physicians. Second, our estimates of drug utilization were limited to nusinersen due to data availability, since onasemnogene abeparvovec was only approved in May 2019. Third, our restriction scores were unvalidated and assumed that each component should be equally weighted. Our estimates of Medicaid FFS births were based on Medicaid FFS total enrollment proportions, which may not be similar for newborns and may exclude those born in managed care programs funded by the Medicaid program.