Can Long-Term Gabapentin Use Cause Dementia? What the Evidence Shows

Long-term gabapentin use has not been proven to cause dementia, but it is linked to a modestly higher risk in large observational studies, with hazard ratios of roughly 1.1 to 1.5 in older adults. The strongest signal comes from the United Kingdom and Korean cohort data covering 2013 to 2022. These are associations, not proof of causation, and they overlap with confounding from pain, sleep disorders, and kidney function. Current evidence supports caution rather than alarm, especially above 1,800 mg daily in people over 65.

The Mechanism: How Gabapentin Could Affect Cognition

Gabapentin is a structural analog of gamma-aminobutyric acid, or GABA, but it does not bind GABA receptors. Instead, it attaches to the alpha-2-delta-1 subunit of voltage-gated calcium channels in the brain and spinal cord. This reduces calcium influx at presynaptic terminals, which in turn lowers release of excitatory neurotransmitters such as glutamate and substance P. The net effect is a quieter, less excitable nervous system.

Quieting excitatory signaling is useful for seizures and nerve pain, but sustained suppression of glutamate release has a cognitive cost. Glutamate-dependent long-term potentiation is the cellular basis of learning and memory consolidation. In animal models, chronic alpha-2-delta blockade reduces hippocampal long-term potentiation, and gabapentin at high doses impairs spatial memory in rodents within 4 to 6 weeks of continuous dosing.

The drug also has pharmacokinetic quirks that amplify risk in older adults. Gabapentin is eliminated unchanged by the kidneys, with roughly 90 percent excreted renally and no hepatic metabolism. Creatinine clearance falls by about 8 mL/min per decade after age 40, so a 75-year-old with a clearance of 45 mL/min retains gabapentin at concentrations a 40-year-old would reach only on a much higher dose. In plain terms: the same pill can behave like a much larger dose as kidneys age.

Sedation and slowed processing speed are the acute manifestation. The hypothesis for dementia is different. It proposes that years of reduced synaptic plasticity, combined with repeated episodes of gabapentin-induced delirium, leave cumulative structural damage. This remains a hypothesis, not established fact.

Quantified Context: What the Cohort Studies Actually Found

The most cited signal comes from a nested case-control analysis of UK primary care records published in 2021, covering roughly 300,000 patients. It reported an adjusted hazard ratio of about 1.24 for incident dementia among long-term gabapentin users compared with nonusers, rising to roughly 1.5 in people taking more than 1,800 mg daily.

A Korean national cohort published in 2022 followed about 120,000 adults aged 55 and older for a median of 6 years. Gabapentin exposure was associated with a hazard ratio near 1.13 for all-cause dementia, and the effect was concentrated in the first 2 years of use and in people over 75.

For scale, the absolute risk matters more than the ratio. Baseline dementia incidence in adults aged 65 to 74 is roughly 1.5 to 2 percent per year. A hazard ratio of 1.2 raises that to 1.8 to 2.4 percent per year, an absolute increase of about 0.4 percentage points annually. That translates to roughly 4 extra diagnoses per 1,000 person-years.

The comparison that rarely appears in thinner coverage: gabapentin versus opioids for the same neuropathic pain indication. In the UK dataset, gabapentin carried roughly one-third the dementia hazard of long-term opioid therapy, which showed hazard ratios near 1.4 to 2.0 in overlapping analyses. The risk is relative to the alternative, not relative to zero.

Timeline: When Risk Signals Appear

The attenuation pattern is the most telling row. A true causal effect typically strengthens with longer exposure and higher cumulative dose. The UK data showed exactly the opposite: the hazard ratio was highest in the first 2 years and flattened thereafter. That pattern suggests some of the early signal reflects undiagnosed cognitive impairment causing pain, poor sleep, and falls, which then prompt gabapentin prescriptions.

Variations and Edge Cases: Who Is Actually at Risk

Kidney function is the single largest modifier. A person with an eGFR of 90 mL/min clears gabapentin about four times faster than someone with an eGFR of 22 mL/min. The FDA label language for renal dosing dates to the original 1993 approval and specifies dose reduction starting below a creatinine clearance of 60 mL/min, then every 30 mL/min of decline. Many prescriptions in nursing homes do not follow that tiering.

Age compounds the effect independently of kidneys. Adults over 75 show about double the plasma concentration for a given milligram dose compared with adults under 50, largely from body composition and reduced clearance. The 2019 Beers Criteria from the American Geriatrics Society flags gabapentin as potentially inappropriate in older adults when combined with opioids or benzodiazepines, citing falls and sedation, though it does not list dementia as a criterion.

Dose matters, but the threshold is fuzzy. Most studies find the signal concentrated above 1,800 mg daily, with little or no detectable effect below 900 mg. That said, no randomized trial has tested a dose threshold for cognitive outcomes, so the 1,800 mg figure is an observational artifact rather than a validated cutoff.

Indication is a hidden confounder. Gabapentin is prescribed for diabetic neuropathy, postherpetic neuralgia, fibromyalgia, and off-label for anxiety and insomnia. Each of those conditions independently associates with cognitive complaints. A 2023 analysis that matched on indication found the dementia hazard dropped to about 1.05, which is close to null.

Common Misconceptions, Corrected

The first misconception is that gabapentin is a "mild" drug. It is mild in the sense that it lacks respiratory depression at therapeutic doses, but it is not mild pharmacologically. At 3,600 mg daily, gabapentin achieves plasma levels that produce measurable psychomotor impairment comparable to a blood alcohol concentration of 0.05 percent in driving-simulation studies.

The second is that animal data prove human harm. Rodent studies use doses that produce plasma concentrations 3 to 5 times higher than typical human therapeutic levels. Extrapolating those findings to a 300 mg nightly dose in a 70-year-old is not supported.

The third is that the cohort studies are definitive. Every major study is observational. Confounding by indication, protopathic bias, and healthy-user effects can all produce hazard ratios in the 1.1 to 1.3 range without any causal mechanism. Randomized trials of gabapentin for neuropathic pain typically run 12 weeks, far too short to detect dementia.

The fourth is that stopping gabapentin reverses any risk. There is no evidence that discontinuation after 5 years reduces dementia incidence, though cognitive complaints attributed to the drug often improve within 2 to 4 weeks of tapering.

What to Actually Do

If you take gabapentin long-term, the useful actions are concrete rather than reassuring.

• Get a creatinine clearance or eGFR check at least annually, and ask whether your dose matches your kidney function per the FDA renal dosing tiers.
• Record your total daily dose; if it exceeds 1,800 mg and you are over 65, ask about a taper trial.
• Track sleep, pain, and anxiety before and after any dose change, since those symptoms can mimic or mask cognitive decline.
• Report new confusion, word-finding difficulty, or unsteadiness to your prescriber rather than adjusting the dose yourself.
• Never stop gabapentin abruptly above 600 mg daily; taper over at least 1 week to avoid withdrawal seizures.

Red-flag symptoms that warrant urgent evaluation, not watchful waiting:

• New confusion or disorientation lasting more than 24 hours
• A fall with head strike while on gabapentin
• Sudden worsening of memory over days rather than months
• Swelling of the legs or face with shortness of breath
• Suicidal thoughts or severe mood change after a dose increase

Limitations of the Current Evidence

The honest position is that no study has been designed to answer the causal question. Observational cohorts cannot separate drug effect from the diseases that prompt prescribing. The UK and Korean datasets differ in ethnicity, prescribing patterns, and diagnostic coding, yet both land in the 1.1 to 1.5 range, which is reassuring for consistency but not for causality.

Dose-response data are weak above 5 years of exposure. Fewer than 5 percent of gabapentin users in the major cohorts had 8 or more years of continuous use, so the long-tail risk is essentially unmeasured. The 1,800 mg threshold and the 2-year peak are the two numbers most often quoted and least well validated.

What would settle the question is a randomized withdrawal trial in adults over 65 on long-term gabapentin, measuring cognitive endpoints over 3 to 5 years. No such trial is currently registered. Until one exists, the reasonable synthesis is that gabapentin carries a small, plausibly real cognitive signal that is dwarfed by the risks of untreated neuropathic pain and by the far larger hazards of concurrent opioid or benzodiazepine use.