diabetes meds and incident dementia

 Another recent article just came out on positive benefits for GLP-1 receptor agonists as well as SGLT-2 inhibitors, finding a decreased risk of Alzheimer’s disease and related dementias (see dm GLP1 SGLT2 less dementia JamaNeuro2025 in dropbox, or doi:10.1001/jamaneurol.2025.0353)

 

Details:

-- this was a target trial emulation study using electronic health record data from 2014-2023, from the OneFlorida+ Clinical Research Consortium of patients who were at least 50 years old with type II diabetes and no prior diagnosis of Alzheimer’s disease and related dementias (ADRD)

    -- this database integrates longitudinal electronic health records linked with the National Death Index for multiple healthcare partners in Florida, with more than 21 million persons covering approximately 86% of Florida’s population, also extending into Georgia and Alabama

-- 33,858 eligible patients were included in the GLP1 versus the other second-line glucose lowering drug (GLD) cohort (these individuals were on sulfonylureas, thiazolidinediones, DPP-4s, alpha-glucosidase inhibitors, and meglitinide)

    -- insulin therapy was excluded from the comparisons given that it might have been associated with longer duration of diabetes

-- 34,185 eligible patients were included in the SGLT2 versus GLD cohort

-- 24,117 were in the GLP-1 versus SGLT2 cohort

-- this emulation compared people at the time of initiation of treatment with GLP1s, SGLT2s, or GLD

-- ADRD was identified using clinical diagnosis codes from the database (they included people as ADRD with only one ADRD diagnostic code, which might have overstated the ADRD label)

-- to equalize patients in the different groups, they used inverse probability of treatment weighting (IPTW) to adjust for potential confounders

 

-- GLP1 versus GLD cohort: mean age 65, 53% female, 41% white/28% Black/13% Hispanic, 43% private insurance/45% Medicare, family income 45% <$40K/23% $40-60K, BMI 34

    -- diabetes/complications: A1c 7.8%, neuropathy 12%, PAD 8%, retinopathy 5%

    -- comorbidities: cardiovasc disease 22%, heart failure 8%, hyperlipidemia 60%, depression 12%, anxiety 12%, hypertension 73%, CKD 13%, obesity 23%, anemia 18%, sleeping disorder 22%, thyroid disease 16%

    -- medications: ACE-I 25%, b-blocker 24%, diuretics 30%, ARBs 22%, statins 50%, NSAIDS 20%, PPIs 22%, antidepressants 13%, benzos 16%, oral steroids 27%, opioids 27%, aspirin 16%, insulin 34%, metformin 42%

-- SGLT2 versus GLD cohort: mean age 66, 49% female, 43% white/26% Black/13% Hispanic, 38% private insurance/47% Medicare, 43% family income <$40K/28% $40-60K

    -- diabetes/complications: A1c 7.7%, neuropathy 12%, PAD 12%, retinopathy 12%

    -- comorbidities: cardiovasc disease 30%, heart failure 18%, hyperlipidemia 64%, depression 11%, anxiety 12%, hypertension 74%, CKD 18%, obesity 52%, anemia 20%, sleeping disorder 22%, thyroid disease 16%

    -- medications: ACE-I 26%, b-blocker 32%, diuretics 33%, ARBs 25%, statins 52%, NSAIDS 18%, PPIs 25%, antidepressants 12%, benzos 18%, oral steroids 28%, opioids 29%, aspirin 22%, insulin 34%, metformin 42%

-- GLP1 versus SGLT2 cohort: mean age 64, 52% female, 42% white/28% Black/13% Hispanic, 46% private insurance/43% Medicare, 45% family income <$40K/25% $40-60K

     -- diabetes/complications: A1c 7.8%, neuropathy 13%, PAD 10%, retinopathy 5%

     -- comorbidities: cardiovasc disease 26%, heart failure 15%, hyperlipidemia 64%, depression 12%, anxiety 13%, hypertension 75%, CKD 15%, obesity 30%, anemia 8%, sleeping disorder 26%, thyroid disease 17%

     -- medications: ACE-I 26%, b-blocker 28%, diuretics 34%, ARBs 27%, statins 53%, NSAIDS 20%, PPIs 22%, antidepressants 13%, benzos 15%, oral steroids 28%, opioids 24%, aspirin 17%, insulin 34%, metformin 42%

 

--Main outcomes: 

    --comparisons of incident dementia in patients initiating GLP1 receptor agonists and SGLT2 inhibitors vs the other second-tier glucose lowering drugs

    --comparison between initiating GLP1s vs SGLT2s

 

Results:

-- incidence rate of ADRD, with ITPW-weighted cohorts:  

    -- GLP1 initiators versus other GLD initiators:

        -- GLP1 initiators: 75 ADRD cases with mean followup of 2.22 years

        -- GLD initiators: 639 cases with mean followup of 3.74 years

            --incidence difference, favoring GLP1: -2.26 per 1000 person-years (-2.88 to -1.64)

            -- 33% decreased risk, HR 0.67 (0.47-0.96)

    -- SGLT2 versus GLD:

        -- SGLT2 initiators: 101 ADRD cases with mean followup of 1.95 years

        -- GLD initiators: 642 cases with mean followup of 3.76 years

            -- incidence difference, favoring SGLT2:  -3.05 per 1000 person-years (-3.68 to -2.42)

            -- 43% decreased risk, HR 0.57 (0.43-0.75)

   -- GLP1 versus SGLT2 cohort: 

        -- GLP1 initiators: 90 ADRD cases with mean followup of 2.39 years

        -- SGLT2 initiators: 130 cases with mean followup of 2.07 years

            -- incidence difference:  0.09 per 1000 person-years (-0.80 to 0.63)

            -- nonsignificant 3% difference

-- no significant differences in subgroup analyses by age <65 vs older, sex, race/ethnicity, obesity, taking metformin, taking insulin, or using different types of GLP1s or SGLT2s.

-- no significant differences in sensitivity analyses using different statistical methods to equalize groups (ie, propensity score matching), excluding those with mild cognitive impairment at baseline, excluding those with Parkinson's at baseline, excluding those with ADRD within 0.5 years after index date, by addressing competing risk of death, or by per-protocol analysis

-- further analysis suggested that an unknown confounder would need for have an HR of at least 2.35 to affect the GLP1 superiority over GLD, and at least 2.9 to affect the SGLT2 superiority 

 

 Commentary:

-- as we know, ADRD is unfortunately common, with an estimated 6.9 million older adults living with ADRD in 2023, and projections that this will double by 2060

-- ADRD is the fifth leading cause of death among older US residents, associated with an estimated cost of $360 billion related to ADRD in 2023

-- the efficacy of our medications is somewhat minimal and controversial 

 

-- there are fundamental differences between the medications we used to treat type II diabetes, a remarkably common disease (and increasingly so) associated with ADRD

    -- GLP1 and SGLT2 medications have important additional benefits including for cardiovascular disease, kidney disease, as well as weight loss (the latter being more so for GLP1s). GLP1s also have studies suggesting decreases in substance use, schizophrenia, respiratory failure and pneumonia, thromboembolic disorders (see https://www.nature.com/articles/s41591-024-03412-w )

   -- on the other hand, there are several lines of evidence that insulin may well be associated with increased dementia risk:

        -- there are increasing data suggesting that insulin crosses the blood brain barrier and has potent direct effects on the brain, where there are abundant insulin receptors on both astrocytes and neurons, especially in the olfactory bulb, cerebral cortex, hippocampus, hypothalamus, amygdala, and septum (see dm insulin cognition and dementia EurJPharm2013 in dropbox,or https://pmc.ncbi.nlm.nih.gov/articles/PMC5405627/pdf/nihms527106.pdf )

        -- several of these medial temporal lobe structures are associated with the learning process

        -- insulin, sulfonylureas, and meglitinide can be associated with hyperinsulinemia and insulin resistance, decreasing insulin-mediated glucose clearance from the blood, and potentially damaging muscle, liver, adipose tissue, endothelium, and the brain

        -- autopsy studies have found that there was brain insulin resistance, especially in the hippocampus, in patients with Alzheimer’s disease, and has been found to be independent of type II diabetes diagnoses or apolipoprotein E genotypes

           -- this latter finding brings up the issue that there are many other causes of insulin resistance (eg visceral obesity, physical activity, diet high in carbohydrates and saturated fats) as well as “pre-diabetes”. Insulin resistance is also associated with hypertension, dyslipidemia, hyperuricemia, elevated inflammatory markers, endothelial dysfunction, and a pro-thrombotic state (http://ncbi.nlm.nih.gov/books/NBK507839/ )

    -- insulin also has direct effects in vitro and in animal studies on beta-amyloid clearance and degradation and also tau protein deposition that forms neurofibrillary tangles found in dementia; insulin dysregulation is also associated with oxidative stress, inflammation, and impaired neurogenesis

 

-- this study found that GLP1s were associated with a 33% lower risk of ADRD and the SGLT2s with a 43% lower risk, though a direct comparison between these meds did not find any significant difference

    -- these conclusions, though based on an emulation study and not a direct assessment, was reinforced by the consistency of the results in subgroup and sensitivity analyses, and the calculation that there was unlikely to be unidentified confounders that have a high level of confounding (they would need to have a hazard ratio of at least 2.3)

    -- the results of the study are consistent with a few observational studies and meta-analyses

-- the mechanism of action for these meds in decreasing ADRD is unknown, though there is reasonably speculation since:

    -- GLP1 receptors are identified in the brain

    -- GLP1s have been shown to reduce neuroinflammation and oxidative stress, improve insulin signaling in the brain (and reducing the downregulation associated with insulin resistance), decrease amyloid-beta and tau hyperphosphorylation in animal models, and promote neurogenesis: https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2022.1033479/full

    -- SGLT2s may be neuroprotective by improving endothelial function, having anti-inflammatory and anti-atherosclerotic effects, being neuroprotective (including for pericytes, astrocytes, microglia and oligodendrocytes), increasing cerebral blood flow, reducing oxidative stress, and improving mitochondrial function: https://www.mdpi.com/1420-3049/26/23/7213 

-- the study authors comment that their subgroup analysis (tucked away in the supplement) does have some mixed results regarding the benefits of metformin and obesity, though, as they point out, these "inconsistencies" don't even come close to being statistically significant

    -- they reference 2 studies finding, per their quote "weight loss may increase risk" of ADRD. But on my assessing these studies, both are observational studies (including a long-term assessment of the Framingham study) that have found that in older (but not younger) people with low BMIs, there is an increased risk of ADRD. these studies do NOT assess the role of weight loss. nor do they have any data on comorbidities (including diabetes, or diet, or exercise, or cancers, or meds, etc). so, a bit of a stretch....

 

-- the authors do mention that there are 2 ongoing placebo controlled phase 3 studies (EVOKE AND EVOKE Plus) assessing the efficacy of oral semaglutide 14mg in patients with early Alzheimer's

 

Limitations:

-- the GLDs include meds acting by very different mechanisms of action.

    -- sulfonylureas and meglitinides can create insulin resistance and, to the extent that insulin resistance is associated with dementia, it would be important to separate this group in the GLD bundle from other second-line medications that do not (eg, alpha-glucosidase inhibitors or thiazolidinediones), which really muddies the results a bit [unfortunately they did not include numbers of patients on these meds, nor provide a breakdown of dementia from the individual meds; so they are likely adding apples and oranges of differing quantities into their averages and decreasing our ability to figure out which meds are better than others]

-- the data in the emulation is very short-term (1.95-3.76 years), which undercuts the interpretation of their results, especially in a disease process like ADRD, which is a long-term, progressive disease. they comment that many people had mild  cognitive impairment (which accelerates the progression to ADRD), that decreasing diabetes as is well done by GLP1s especially but also SGLT2s, and the rather profound neuroprotective effects of these meds might have a shorter time course to benefit

-- there still could be important and unrecognized confounders that would change the results of this study, as noted above

    -- they did try to include some of these tangentially, such as socioeconomic factors (family income,  private vs public health insurance, and educational level as surrogates for very complex social relations), though there was no information on the important effects of diet, exercise, family history, etc

-- as with any of these data-mining studies, they rely on the accuracy of the diagnostic codes in the databases. these codes are based on what clinicians put in on billing. and very busy clinicians often focus more on examining and treating patients than on the accuracy of their coding and billing data.....

-- the database, being based on clinical encounters, is therefore not following strict protocols. so, the medication choices made by clinicians when seeing their patients may be pretty biased, both by the patients underlying conditions (there were quite significant differences in baseline comorbidities between each of the three studies of GLP1 vs GLD, SGLT2 vs GLD and GLP1 vs SGLT2) and by clinician comfort with prescribing their diabetes drugs. it is notable that several of the prescribed medications are quite different from the guidelines (eg, only about 50% of people with diabetes were on statins)

    -- and there might even be outcome differences between the GLP1 or SGLT2 medications selected

 

so, an intriguing study that expands the potential benefits of the GLP1s and SGLT2s to ADRD, especially in their comparison to the older and potentially harmful meds used (in particular, the ones that increase insulin resistance)

-- of course, there are an array of nonpharmacologic approaches that help and are not part of this study (or even mentioned in the article):

    -- increased cognitive stimulation and social connections/support seem to decrease dementia risk: https://gmodestmedblogs.blogspot.com/2019/07/mental-and-social-activities-dec.html

    -- a healthy lifestyle with a Mediterranean diet, decreased sugary beverages, etc may help, even if there is a genetic predisposition to dementia: https://gmodestmedblogs.blogspot.com/2019/07/dementia-genetics-and-lifestyle-both.html

    -- a diet high in extra-virgin olive oil seems to help: https://gmodestmedblogs.blogspot.com/2022/01/olive-oil-and-decreased-mortality.html 

    -- exercise is important: https://gmodestmedblogs.blogspot.com/2020/01/exercise-and-cognitive-health.html

-- this argument to expand the benefits (and perhaps indications) for GLP1s or SGLT2s unfortunately flies in the face of health insurers cutting back on their approvals for these meds because of their excessive costs (up to 1000+ times that in other countries), however:

    -- the underlying issue is that the US, unlike most other resource-rich countries, has a fundamentally dysfunctional health care system that is not inclusive of the population, is largely employment-based, is not cohesive, has substantial obstacles to care for many people, and does not allow for the overall system to negotiate med prices

    -- this became openly discussed by the drug companies with the advent of the really important hepatitis C meds, where the drug companies openly stated that their med pricing was determined not by their (overstated) Research and Development expenses but by the health care system's anticipated savings on the expensive sequelae of untreated hepatitis C (liver failure/transplants, hepatocellular cancer)

        -- ie, drug companies can charge whatever they want. and highly important drugs (eg GLP1s and SGLT2s) are a fertile base for exorbitant prices and profits

        -- for more specific information on the differences in prices of diabetes meds over the world: 

 

https://gmodestmedblogs.blogspot.com/2024/04/drug-co-shenanigans-dm-meds.html

geoff

If you would like to be on the regular email list for upcoming blogs, please contact me at gmodest@bidmc.harvard.edu

  

to get access to all of the blogs:

 

 go to http://gmodestmedblogs.blogspot.com/ to see the blogs in reverse chronological order

  -- click on 3 parallel lines top left, if you want to see blogs by category, then click on "labels" and choose a category​

  -- or you can just click on the magnifying glass on top right, then type in a name in the search box and get all the blogs with that name in them

 

if you would like to see the articles in this blog, please email me. 

 

please feel free to circulate this to others. also, if you send me their emails (gmodest@bidmc.harvard.edu), i can add them to the list