Intranasal Insulin

Insulin is a peptide hormone made by the pancreas that, besides controlling blood glucose, acts in the brain [7]. Its receptors are densest in the olfactory bulb, hypothalamus, hippocampus, cerebral cortex, striatum and cerebellum [8]. Most brain insulin is thought to arrive from the blood through a saturable transport system, which works less well in obesity and perhaps in Alzheimer's disease [1]. Insulin sprayed into the nose travels along olfactory and trigeminal pathways and reaches the cerebrospinal fluid within 30 minutes while largely bypassing the bloodstream [1][2][9], so its effects on the brain can be studied without causing hypoglycemia [1]. Trials have used ordinary injectable insulin sprayed into the nose, an off-label use [10].

Healthy people. In 38 young adults, 40 IU four times a day for 8 weeks improved delayed recall of a word list (6.2 words versus 2.9 on placebo), reduced anger and raised self-confidence, with no change in blood glucose or insulin [3]. The rapid-acting analog insulin aspart improved recall more than regular insulin [11]. A single 160 IU dose improved spatial and working memory in women but not men [12], and 160 IU once daily for 8 weeks gave only slight memory gains, and only when taken in the evening [13]. A 2023 meta-analysis of 11 trials in 400 healthy people found no overall effect on cognition [14].

Mild cognitive impairment and Alzheimer's. In single-dose studies, memory improved in memory-impaired adults without the APOE ε4 allele but worsened slightly in carriers [15][16], and a single dose of rapid-acting glulisine had no effect on memory in 9 ε4 carriers with Alzheimer's [17]. In the first 4-month pilot trial (104 adults), 20 IU a day improved delayed story recall, both 20 and 40 IU a day preserved daily function, and the insulin groups lost less brain glucose metabolism on PET; the authors noted that the effects were small in absolute terms [4]. A later 4-month trial found that regular insulin improved memory but long-acting detemir did not [18]. Rapid-acting glulisine twice daily for 6 months did no better than placebo in 35 people [19].

The decisive test, the SNIFF phase 2/3 trial, randomized 289 people at 27 sites to 40 IU a day or placebo for 12 months. The original nasal device kept malfunctioning, so the main analysis used the 240 people given a second device; in them there was no difference in cognition, daily function or cerebrospinal fluid markers (ADAS-cog-12 difference 0.03 points, P = .98) [5]. The 49 people who used the first device showed a trend toward better cognition and improved Aβ42-to-Aβ40 and Aβ42-to-tau ratios in cerebrospinal fluid, which the authors cautioned may be chance or a difference in how well the devices delivered insulin [5]; a secondary analysis of that group reported slower growth of white matter lesions [20]. A 2023 meta-analysis found a small benefit on global cognition in mild cognitive impairment and Alzheimer's (standardized mean difference 0.22, 12 trials) [14], but a 2026 meta-analysis of 5 randomized trials (540 people) found no benefit on any cognitive, functional or biomarker outcome [21]. A 4-week trial of a higher dose, 160 IU a day, reported a nominal improvement on a cognitive composite, without correction for multiple comparisons [22].

Type 2 diabetes and older adults. One 40 IU dose improved visuospatial memory in older adults with and without type 2 diabetes [23]. In the 24-week MemAID trial (223 people), 40 IU daily did not improve cognition in those with diabetes, though they walked faster; in those without diabetes, executive function and verbal memory were better than on placebo [10].

Delirium. In older patients having major surgery, single-center trials in China reported less delirium: 12.5% versus 47.5% after 20 IU on the two days before gastrointestinal surgery [24], and 10.9% versus 26.6% with 40 IU daily around orthopedic or pancreatic surgery [25]. A meta-analysis of 7 such trials, all from China, found a 61% lower risk [26]. A 2026 network meta-analysis of 158 trials of delirium-prevention drugs ranked intranasal insulin among the effective options (odds ratio 0.13), with the benefit shown in elective abdominal surgery and the evidence rated moderate to very low quality [6]. In cardiac surgery, 40 or 80 IU given after induction of anesthesia did not significantly reduce delirium [27]. In 100 hospitalized older adults who already had delirium, long-acting insulin did not significantly shorten it overall, though hospital stays were shorter [28].

Other conditions. In bipolar disorder, 8 weeks improved one test of executive function out of a battery [29]; trials in treatment-resistant depression [30], schizophrenia [31] and multiple sclerosis [32] were negative. In Parkinson's disease a 14-person pilot reported better verbal fluency [33], but a 30-person dose-finding trial found no outcome that improved more than on placebo [34]. Insulin placed in the olfactory cleft twice a week improved smell in 38 people with hyposmia [35].

Appetite and weight. In healthy men, 40 IU four times a day for 8 weeks reduced body fat by 1.4 kg, while women gained 1 kg as body water [36]. The same regimen did not reduce weight or fat in obese men, though their memory and mood improved [37]. A single 40 IU dose cut breakfast intake by about 12% in men [38], and 160 IU after lunch reduced chocolate-cookie snacking in women [39].

Insulin binds the insulin receptor, a tyrosine kinase that recruits insulin receptor substrate proteins and signals through the PI3K–Akt and MAPK pathways [7][8]. Brain insulin receptors are enriched at synapses, where insulin regulates the trafficking of AMPA and NMDA receptors and the long-term potentiation and depression that underlie memory; it also enhances neurite outgrowth and modulates catecholamine release [1][8]. Neurons take up glucose mainly through GLUT3, which does not need insulin, so insulin influences rather than controls brain glucose use [8]. Insulin also affects how the amyloid β peptide is cleared and how tau is phosphorylated, and it acts on blood vessels and inflammation [7].

"Brain insulin resistance", a failure of brain cells to respond to insulin, has been shown in post-mortem hippocampus from people with Alzheimer's, which is the rationale for giving insulin to the brain; there is still no agreed way to measure it in living people [7][8].

After intranasal dosing in rats, labeled insulin traveled along the outside of the trigeminal nerve and through the spaces around brain blood vessels within 30 minutes, reached the cortex, hippocampus and hypothalamus, and activated insulin receptors [9]. In healthy men, 40 IU raised ATP and phosphocreatine in the brain within 10 minutes, and the size of the rise predicted how much less they ate afterward [38]. 160 IU lowered synaptic dopamine in the striatum [40] and, in lean but not obese men, improved whole-body insulin sensitivity, an effect linked to hypothalamic activity and parasympathetic output [1][41].

Insulin type. Over 8 weeks in healthy men, intranasal aspart improved recall more than regular insulin [11]. Long-acting detemir improved memory over 3 weeks in APOE ε4 carriers but worsened it in non-carriers [43]; over 4 months it did no better than placebo and was linked to smaller right hippocampal and cuneus volumes, while regular insulin improved memory [18]. Glulisine had no effect on cognition [19], and lispro could not be detected in human cerebrospinal fluid [44].

Device. Many trials used the ViaNase electronic atomizer, which releases a metered mist into a chamber over the nose that is breathed in over about 2 minutes [4]. In the SNIFF trial its added timer failed often, and the replacement device (Impel's I109 Precision Olfactory Delivery) was used in the null primary analysis [5]. Other trials used a cartridge pump system [22], the SipNose device [34] or simple 0.1 mL spray pumps [31][39]. The SNIFF investigators concluded that future devices should be shown to deliver insulin to the brain before efficacy is tested [5].

Additives. Regular insulin, detemir and similar products contain m-cresol, which has a noticeable coal-tar smell, so placebo sprays use the insulin diluent to keep participants blinded [1]. Porcine insulin given with an absorption enhancer (laureth-9) at about 1 unit per kilogram entered the blood and caused hypoglycemia in healthy volunteers [45].

Across 38 single-dose studies (1,092 people) and 18 repeated-dose studies (832 people, 21 days to 9.7 years), a 2018 review found no symptomatic hypoglycemia and no severe adverse events; transient nasal symptoms were common with both insulin and placebo sprays, and the authors judged long-term safety data insufficient [47]. In healthy adults taking 60 IU daily for 3 weeks, rhinoscopy showed no nasal irritation and nasal clearance and airflow were unchanged, though 11 of 20 reported transient stinging or an unpleasant smell [48].

In the 4-month pilot trial in Alzheimer's, adverse events were more frequent on 20 IU a day than on placebo, mostly rhinitis and nosebleeds [4]. In the 12-month SNIFF trial adverse events were equally common on insulin and placebo; vascular disorders were 12.4% versus 5.0% (not significant, P = .07), and hippocampal volume fell slightly more on insulin [5]. A 2026 meta-analysis found more gastrointestinal adverse events (risk ratio 1.57) and fewer cardiovascular events with insulin [21]. In a Parkinson's trial, nasal irritation or nosebleeds affected nearly everyone, including on placebo, and one person on 80 IU a day stopped because of persistent nosebleeds [34].

Blood glucose is little affected at usual doses. In the 24-week MemAID trial, asymptomatic low glucose readings were similarly common on insulin and placebo, including in people with diabetes [10]. During cardiac surgery, 40 and 80 IU caused no hypoglycemia [27], but 240 IU caused it in 2 of 3 patients and 160 IU raised serum insulin [46]. HbA1c rose very slightly over 4 weeks of 160 IU a day [22].

Effects on memory were not always positive: single doses worsened recall in some memory-impaired APOE ε4 carriers [15][16], and 40 IU a day worsened memory in APOE ε4-negative women with mild cognitive impairment or Alzheimer's [49].