Michigan put gabapentin on its controlled substance list in January 2019, the fourth state to do so, and began requiring every gabapentin prescription to be reported into the state's monitoring system. The drug had not changed. Its federal status had not changed. What changed was the paperwork and the license a prescriber needed to write it.

Five years later the state reversed itself. On May 28, 2024, the Pharmacy-Controlled Substances Rules issued by the Michigan Department of Licensing and Regulatory Affairs took effect and removed gabapentin from the state schedule, restoring it to the same footing as an ordinary prescription drug. A prescriber without a controlled substance license could now write it, and dispensers no longer had to report it to the Michigan Automated Prescription System. The reversal was not prompted by a finding that gabapentin's risks had shrunk. The regulatory impact statement filed by the state's Bureau of Professional Licensing says the change was driven partly by a wish to align Michigan's classification with neighboring states and with the federal government, which has never scheduled gabapentin.

That leaves an unusually clean natural experiment. A research letter published September 21 in JAMA Network Open reports what happened to prescribing after the brake came off, and the answer is that it rose almost immediately. Gabapentin dispensing to Michigan Medicaid enrollees jumped by 4.00 prescriptions per 1,000 enrollees (95% CI, 2.33 to 5.67) at the moment of the policy change, measured against a comparison drug that stayed under federal control. The authors, led by Katherine Gora Combs of the University of North Carolina at Chapel Hill and Juan Hincapie-Castillo of the University of Pittsburgh, read the result as evidence that scheduling works as a supply-side check on prescribing rather than only as a tool aimed at diversion.

The state ran the experiment in both directions

Most drug scheduling debates are one-way arguments. A legislature adds a substance to a controlled list, advocates for the change predict that prescriptions will fall and misuse with them, and nobody runs the tape backward to see what the list was worth. Michigan did run it backward. The state scheduled gabapentin effective January 4, 2019, then rescinded the scheduling five years later, which means the same state, the same Medicaid program, and the same prescribing base were observed under both rules within a single dataset.

The policy environment matters here. Gabapentin sits in an odd place in American drug regulation. It is one of the most prescribed medications in the country, approved for postherpetic neuralgia and partial-onset seizures, and used far more widely off-label for chronic pain, a use that grew as prescribers came under pressure to write fewer opioids. Its close relative pregabalin has been a Schedule V controlled substance since it reached the market. Gabapentin never followed. Between 2016 and 2024, eight states enacted laws classifying gabapentin as a Schedule V drug, each one responding to the same concern: a medication that is not especially dangerous on its own can carry real risk when it is combined with opioids and other sedatives, and its wide prescribing footprint makes small per-patient risks add up across a population.

Michigan's 2019 decision made it an early mover among those states. Its 2024 decision made it a case study in what happens when a state takes the step back.

The counts moved, and the comparison drug did not

The study draws on the Medicaid State Drug Utilization Data, a public dataset of quarterly outpatient dispensing, paired with enrollment figures from the Centers for Medicare and Medicaid Services. The authors assembled quarterly dispensing rates for Michigan Medicaid enrollees from the first quarter of 2023 through 2025, expressed as prescriptions per 1,000 enrollees, and analyzed them with a controlled interrupted time-series model, a standard design for evaluating a policy that arrives on a known date.

The comparison is the part that gives the finding its force. Pregabalin was used as an internal control, chosen because it treats similar conditions, sits in the same drug family, and stayed a federally controlled substance throughout. If gabapentin dispensing rose because pain was being treated more aggressively across the board, or because enrollment or data reporting shifted, pregabalin should have moved too. It did not move in any meaningful way. Pregabalin held between roughly 10 and 13 prescriptions per 1,000 enrollees for the whole period. Gabapentin ran between 33 and 43 per 1,000, a range consistent with the 21.98 per 1,000 gap between the two drugs that existed before the policy changed and that the model used as its baseline.

What the model estimates is a level shift, not a trend break. Dispensing rose at the moment of the change and then continued along roughly the path it had been on. The authors report the post-policy trend change as 0.49 prescriptions per 1,000 enrollees per quarter with a confidence interval that crosses zero, which means the data cannot distinguish that slope from no change at all. The honest reading is that the policy produced a step up in prescribing and not a lasting acceleration, and the authors say so.

What prescribers were responding to

Gabapentin's controlled status in Michigan never blocked a patient from getting the drug. It added steps. A prescriber needed a controlled substance license, a prescription had to be reported into a monitoring system that a clinician could expect to be reviewed, and the prescription carried a legal signal that the drug was one to think twice about. The study's authors describe the mechanism as prescriber hesitancy, and their account is behavioral rather than pharmacological.

The researchers wrote that the findings point to drug scheduling as something that "may serve as a supply-side policy to prevent overprescribing," a framing that treats the controlled list as a governor on how readily a medication flows out of pharmacies. In an interview with MedPage Today, Gora Combs described the same mechanism in both directions, saying that prescriber caution under tighter rules tends to loosen once those rules are lifted.

The published evidence on that reading is not thin. Earlier work by overlapping groups found that prescribing fell in states that scheduled gabapentin, and the new letter's contribution is showing the relationship running backward, which is a stronger test than either direction alone. A state that had scheduled a drug and then unscheduled it, with the same population and the same claims pipeline on both sides of the date, gives the mechanism less room to hide behind other explanations.

The federal government looked at the same evidence and said no

The national version of this argument has been settled for now, and settled against scheduling. Public Citizen petitioned the FDA and the DEA in 2022 to place gabapentin and gabapentin enacarbil in Schedule V under the Controlled Substances Act, citing abuse potential, dependence, and overdose risk. The DEA accepted the petition for gabapentin in October 2022 but declined to start proceedings for gabapentin enacarbil, and on January 18, 2023, the FDA denied the petition, leaving gabapentin unscheduled at the federal level. Physician organizations including the American Medical Association have opposed federal scheduling, arguing that the drug's record does not justify the access costs that a controlled classification imposes.

That refusal is a decision about magnitude as much as about principle. Nobody in the federal proceeding argued that gabapentin is harmless. The argument was that its abuse liability is modest relative to the drugs already scheduled, that its misuse tends to occur alongside opioids and other sedatives rather than on its own, and that a Schedule V listing would burden millions of patients managing pain and neuropathy to prevent a harm the evidence did not clearly show the listing would stop. Michigan's numbers now put a size on the other side of the ledger. Removing the listing produced something on the order of four additional prescriptions per 1,000 enrollees right away, in a program whose rate had been running in the thirties.

Access sits on the other side of the ledger

There is a case for the reversal that the study's framing does not capture, and it deserves the same hearing the scheduling argument gets. Gabapentin is a workhorse drug for conditions where the alternatives are worse. For postherpetic neuralgia and for some seizure disorders it is standard therapy, and for chronic pain it is one of the few non-opioid options prescribers can reach for. Brian Callaghan, a University of Michigan neurologist who was not involved in the study, told MedPage Today that removing the scheduling requirement makes the drug easier to prescribe and that its downsides do not rise to the level of justifying a controlled classification. A patient who needs gabapentin and has to find a prescriber with a controlled substance license, or who faces a pharmacy queue that moves more slowly because of reporting requirements, pays a real cost for a policy aimed at an aggregate risk. Controlled status also changes what happens after the prescription is filled. Every fill lands in a state database that law enforcement can query, refills can be limited by the schedule rather than by the clinical picture, and a patient who moves between states can find that a prescription written in one is treated differently in another. Those costs fall on everyone who takes the drug, while the benefit is measured in the prescriptions that never get written.

The study's authors acknowledge that tension in their own conclusions, calling for future work on how the policies affect access for patients with legitimate need. That is the question a prescribing rate cannot answer. A rise in prescriptions is consistent with more appropriate treatment of pain and with more prescribing that will not help anyone, and nothing in dispensing data separates the two.

What the numbers cannot answer

The study has limits the authors list plainly, and they matter for how far the finding travels. Dispensing data record what left the pharmacy counter. They say nothing about what patients swallowed, and they miss any gabapentin obtained through channels that never generate a claim. The analysis window is short by the standards of policy evaluation, which is why the trend estimate is inconclusive. And a controlled interrupted time series can only adjust for the comparison it was given, not for every other change in Michigan's Medicaid program or in pain medicine generally across the same quarters.

What the letter does establish is narrower and more useful than a verdict on gabapentin. It measures the braking force of a scheduling decision in a setting where the brake was applied and then released, and it finds that the brake was holding back a measurable volume of prescribing. That finding cuts in both directions at once. It is evidence for the states that scheduled gabapentin to reduce use, and it is evidence for the states that have not, since it implies that keeping gabapentin uncontrolled kept a drug with real risks flowing more freely. The federal government decided in 2023 that the access cost of scheduling exceeded the harms it would prevent. Michigan's data does not overturn that judgment, but it prices it, and a price is what has been missing from a debate that has run for a decade on assertions about deterrence. If a state schedules gabapentin and prescribing falls, that is the policy working as designed. If a state unschedules it and prescribing rises, that is the same mechanism seen from the other end. Both readings describe a controlled list that changes clinical behavior, which is either the point or the problem depending on what a patient needs from the drug.

Primary sources

  1. JAMA Network Open for the study design, data sources, and all reported estimates: Gora Combs K, Lo-Ciganic W, Yang J, Hincapie-Castillo JM. Gabapentin Use in Medicaid After Descheduling. JAMA Netw Open. 2026;9(9):e2635059. Published online September 21, 2026.
  2. Michigan Department of Licensing and Regulatory Affairs (LARA), Bureau of Professional Licensing, for the Pharmacy-Controlled Substances Rules that removed gabapentin from the state schedule effective May 28, 2024, and for the regulatory impact statement reasoning.
  3. Centers for Medicare and Medicaid Services, Medicaid State Drug Utilization Data, for the quarterly dispensing dataset analyzed in the study.
  4. U.S. Food and Drug Administration regulations.gov docket FDA-2022-P-0149 for the Public Citizen citizen petition on federal scheduling of gabapentin, the Drug Enforcement Administration's October 7, 2022 response, and the FDA denial letter of January 18, 2023.
  5. MedPage Today for the interview with study co-author Katherine Gora Combs and with Brian Callaghan of the University of Michigan (read in the RheumNow republication).
  6. Psychiatric News Alert, American Psychiatric Association, for its summary of the study's findings and conclusions.