A First Month Above 50 MME Raised OUD Risk 147 Percent. Gabapentin Landed at 102.

Most people who show up at a doctor's office with a new ache in the lower back do not leave with an opioid prescription. In a cohort of nearly 400,000 Medicaid patients published this week, 79 percent of them did not. The most common first-line choices were anti-inflammatory drugs, at 48 percent, and muscle relaxants, at 44 percent — which is roughly what the guidelines have been asking for since 2017.

The interesting part is what happened inside the other 21 percent.

Among the patients who did fill an opioid prescription in that first month, the dose and the duration of a single month of treatment sorted them into groups whose twelve-month outcomes differed by more than a factor of two. Patients whose first-month treatment included opioid doses above 50 milligrams of morphine equivalent per day carried a 146.6 percent higher risk of a later opioid use disorder diagnosis or overdose than patients who did not start that treatment, with a confidence interval of 121.5 to 171.7 percent. Low doses held longer — 50 MME or less but for more than seven days — came in at 105.2 percent. And gabapentinoids, the drugs most often offered as the opioid-sparing alternative, landed at 101.6 percent.

What the Cohort Could See

The study, led by Lisa Doan at NYU Grossman School of Medicine with Kara Rudolph at Columbia Mailman as senior author, drew on the Transformed Medicaid Statistical Information System analytic files from 25 states that had expanded Medicaid under the Affordable Care Act by 2014. The window runs from January 2016 to December 2019, with a one-month exposure period followed by twelve months of outcome tracking. Everyone in it was opioid-naive, under 65, and had received treatment within a month of a new acute low back pain diagnosis. Median age was 39, and 59.5 percent were female. The National Institute on Drug Abuse funded the work under R01DA053243.

The outcome was a composite: a diagnostic code for opioid abuse or dependence, an opioid-involved overdose, or a prescription for a medication used to treat opioid use disorder. Cumulative incidence reached 1.4 percent at seven months and 2.3 percent at 13 months.

That composite is worth pausing on, because it determines what the percentages mean. It is a claims-based measure, so it captures what got coded, not what happened. A patient who developed opioid use disorder and never returned to care does not appear in it. Neither does a patient who died; death certificates are not in these files. The 2.3 percent is therefore a floor on the treated and diagnosed population, and the relative risks are comparisons between groups whose contact with the health system is not identical.

The Ladder Inside a Single Month

Initial treatment in the first month	Change in risk of OUD or overdose diagnosis (95% CI)
Opioids above 50 MME/day	+146.6% (121.5 to 171.7)
Opioids at 50 MME/day or less, more than 7 days	+105.2% (89.0 to 121.4)
Gabapentinoids	+101.6% (84.9 to 118.2)
Opioids at 50 MME/day or less, 7 days or less	+31%
Chiropractic care	−32.8% (26.8 to 38.9)
Physical therapy	−31.1% (22.1 to 40.1)

Read the bottom of that ladder carefully. Even the most conservative opioid exposure in the study — a week or less, at 50 MME or below, the pattern most prescriber education has converged on — carried a 31 percent higher relative risk. There is no row here where starting an opioid is neutral. What the data describe is a gradient, not a cliff, and the gradient starts on the safe side of every threshold the CDC guideline names.

The Finding That Did Not Make the Headline

Gabapentin and pregabalin were not the subject of the press release. They are, on these numbers, the most consequential result in the paper.

These drugs are widely prescribed as opioid adjuvants and as first-line therapy for neuropathic pain, and the logic of using them in acute low back pain is to reduce opioid exposure. In this cohort, starting one was associated with a risk increase statistically indistinguishable from starting low-dose opioids for more than a week. The point estimate sits one percentage point below the low-dose opioid figure, and the confidence intervals overlap almost entirely.

There is a separate safety reason to notice this. The FDA added a warning in December 2019 about serious breathing difficulty in patients taking gabapentinoids alongside opioids or other drugs that depress the central nervous system, and the agency required the risk to be studied further in the populations most exposed. A cohort that treated gabapentinoids and opioids as separate exposures does not tell us what happens when they are combined, which is the most common real-world pattern.

Why the Comparison Is Harder Than It Looks

Claims data record what was prescribed, not why. Patients who start gabapentin for a new episode of back pain are not a random subset of the cohort: they are more likely to have pain that a clinician judged neuropathic, or severe enough to warrant a second drug, or to already be on an opioid that the gabapentin was meant to supplement. The cohort was opioid-naive at baseline, but baseline naivety is a snapshot, and the twelve-month follow-up contains every subsequent prescription.

So the 101.6 percent is an association that survives adjustment for measured variables and cannot rule out the unmeasured ones. The authors do not claim otherwise. What makes it worth a paragraph anyway is the asymmetry: the same limitation applies to the opioid figures above it, and only one of those sets of numbers got repeated in every write-up.

What the Nonpharmacologic Numbers Are Worth

Every nonpharmacologic, noninterventional treatment in the study was associated with a reduction in risk. Chiropractic care produced the largest, at 32.8 percent, followed by physical therapy at 31.1 percent.

These are the numbers a reader is most likely to over-read. Sixteen percent of the cohort received physical therapy and 10 percent received chiropractic care, which means the comparison group for each is the 84 and 90 percent who did something else — including the patients who were never referred anywhere. Referral to physical therapy is itself a marker: it means a clinician decided the problem was mechanical, took the time to write a referral, and had a clinic to send the patient to. Patients on the receiving end of that decision differ from patients who were not, in ways no claims file measures.

The direction still matters. It matches the guideline, it matches trial evidence for exercise-based care in acute low back pain, and it is the one set of estimates in the paper that points toward a change in practice rather than a restriction on one.

The Twenty-Five-State Problem

The cohort covers states that expanded Medicaid by 2014. Texas, Florida and Georgia did not, and are absent. So are the commercially insured and the uninsured, along with everyone 65 and older.

That is not a small caveat for a paper whose headline number is a risk ratio. The states left out of the sample are not a random draw: they include several of the largest contributors to the national opioid death toll, and their Medicaid programs cover a different, narrower slice of low-income adults. The 2.3 percent absolute figure should not be transplanted to them without a second study. The relative gradient is more portable, because it is a comparison within the same population and the same benefit design, but portability is a hypothesis here, not a finding.

There is also a timing problem that the paper's design cannot address. The cohort ends in December 2019. The illicit supply that most overdose deaths now involve — fentanyl and its analogs, increasingly cut into counterfeit pills — reached its current dominance after the follow-up window closed. The overdose half of the composite therefore describes a drug market that no longer exists, which likely makes the absolute overdose risk in this cohort an underestimate of what the same prescribing pattern would produce today.

What a Prescriber Can Take From It

Nothing here changes the first-line recommendation, which was already nonpharmacologic care followed by non-opioid medication. What the study adds is a quantified reason to hold that line, and two specific cautions that are easy to lose in a summary.

The first is that short-course, low-dose prescribing is not a free option. It is the least risky opioid exposure measured here, and it still moved in the wrong direction. The second is that adding a gabapentinoid to reduce opioid exposure is not obviously a risk-reducing move on these data, and the combination carries its own FDA warning.

For patients who do go on to develop opioid use disorder — and 2.3 percent of this cohort did within thirteen months — the path forward is well established. Opioid addiction treatment begins with assessment and a level of care matched to severity, and medication-assisted treatment with buprenorphine or methadone remains the intervention with the strongest evidence behind it. The prescribing question and the treatment question are the same continuum seen from two ends.

Where the Evidence Goes Next

A randomized trial of opioid dose in acute back pain is not going to happen, and it should not. The realistic next step is a target trial emulation that uses prescriber-level variation as an instrument — clinicians differ widely in how they treat an identical presentation, and that variation is closer to random than anything a claims cohort can construct on its own.

The other useful study is a replication in the current supply environment, with the follow-up window placed after 2021, in expansion and non-expansion states alike. Until one exists, the honest summary of this paper is narrower than its headline: in a large Medicaid population treated before fentanyl dominated the supply, the dose and duration of one month of opioid therapy, and the decision to add a gabapentinoid, all tracked with what happened over the following year.