A new thought-provoking article: Metformin and Sulfonylurea Use and Risk of Incident Dementia. No one asked Alzheimer Gadfly for a guest editorial, but since one did not accompany the article, in August’s Mayo Clinic Proceedings, I’m going to provide an amateur perspective here.
People with Type 2 diabetes mellitus (with high blood sugar) are at higher risk for eventually developing dementia, along with a lot of other things. This kind of diabetes usually starts in adulthood, and many times can be controlled, at least initially, with an oral medication, without insulin. Metformin or sulfonylurea meds (like glyburide or glipizide) are commonly used first. Could these medicines then influence the development of dementia?
The authors here, Scherrer et al, are not the first to ask this interesting question; they cite a number of studies giving conflicting results on metformin’s effect on eventual dementia (in humans and rodents). So the investigators thought they’d look at a retrospective “natural experiment,” examining the medical records of Veterans Administration (VA) patients, assuming that the diabetes prescribing habits of VA clinicians was not mandated. It also assumes that clinicians would be good about including diagnoses like dementia in the medical record.
Starting with over 2 million VA outpatient records, they winnowed the files down to 75,000 patients who were given a new diabetes prescription with no mention of dementia in their records, and other criteria. They followed the records and found that ~7% did develop dementia in their timeframe. They found that those started on metformin had fewer cases of dementia than those started on sulfonylureas, with a reduction of 8%. But at least one-third of the patients did switch drugs at some point, as happens in clinical practice. They also mention that sicker patients tended to start on sulfonylureas, but they used a statistical weighting system to balance the groups.
The authors tried to “replicate” their findings by using a similar (not identical) selection method to look at a different group of patients, the Kaiser Permanente (KP) patient data base in Washington State (disclosure: I was a doctor there over 15 years ago, when it was called GHC, and I’m still a patient there!). About the same number, ~ 6% , of the KP patients given a new diabetes prescription without dementia eventually developed it. Although the KP metformin group vs. the sulfonylurea group also had proportionately fewer cases of dementia, it was evidently not statistically significant, even when the authors looked at several different models for calculations (Table 3 and text).
The most fascinating thing to me was looking at the details of the co-morbid conditions of the VA group compared to the KP group. The VA group definitely seems sicker, with more obesity, high blood pressure, higher lipids, and much more heart disease! The VA group also had more strokes, traumatic brain injury, and B12 deficiency, not to mention more depression and PTSD. There was also more smoking, more alcohol use and more medication prescription use, listed in Table 2 for the VA group. As mentioned in the text, the sicker heart patients tended to be started on sulfonylureas. With so many risk factors, I was a bit surprised that their rate of dementia incidence was not substantially higher.
The World Health Organization produced guidelines to prevent dementia. To quote the press release: ” People can reduce their risk of dementia by getting regular exercise, not smoking, avoiding harmful use of alcohol, controlling their weight, eating a healthy diet, and maintaining healthy blood pressure, cholesterol and blood sugar levels…”
So, the authors, in their words, found metformin provided “…a modest reduction in the risk of dementia,” at least in the sicker VA cohort, using a weighted variable methodology (age 50-74, including a “glycemic burden” measure to estimate blood sugar, etc). It was somewhat disappointing that same methodology did not show a significant result in the KP cohort. But maybe the effect can only be seen in high-risk lifestyle persons?? Might be interesting to give metformin to high-risk genetics and lifestyle folks (smoking, drinking, sedentary, etc) in a placebo controlled trial, but the UK study suggested that fewer then 2% will develop dementia in a study timeframe, so it might be hard to get statistical power.
One thing that surprised me is that the authors bring up donepezil in their penultimate paragraph: “…it warrants mentioning that the Food and Drug Administration–approved treatment for dementia, donepezil, has only a modest effect on dementia progression.” Donepezil is FDA approved to treat Alzheimer’s disease; its effect may be modest, but it was approved through a series of clinical trials. The authors I guess are talking about a possible prevention strategy here, based on the less stringent clinician-coded diagnoses in retrospective data, with nothing about cognitive improvement. An effective prevention might meet the FDA’s guidance criteria for accelerated approval from last year, if a trial used biomarkers, (see my commentary on that in the Federal Register, or here, comparing it to oncology accelerated approvals).
I’m no biostatistician, but l have a simpleton’s way of thinking about the “number needed to treat” to prevent cases of dementia, in the VA population presented. Out of 75,187 diabetes patients, 5249 developed dementia (~7%). If the metformin result is real, there could be an 8% reduction of the 5249, or 419 cases of preventing (or perhaps delaying) dementia, which would be fantastic!! Of course, 4,830 cases would not be prevented. And then you would need to give all 75,000 patients metformin, generic and relatively cheap but not without side effects, so you’d be unnecessarily treating, for prevention purposes…at least 70,000 people? (OK maybe only 30,000 if sulfonylureas might really be the better diabetes agent for half). If one could predict who responds to metformin, that would improve the ratio.
Switching gears, have you ever heard the radio commercial calling Alzheimer’s “diabetes of the brain” or heard it called “Type 3 diabetes?” I’m not gonna go there, go search for it until your brain’s sweet spot is satisfied; there are papers listed by PubMed.Gov. There are research groups looking at the complexity of brain metabolism, or “metabolomics.” Ketone diet anyone? I did comment on the trial of nasally inhaled insulin mentioned at the recent AAIC2019 meeting, on this blog. BTW, a smaller set of GHC/KP folks volunteered for a long term study of dementia; Crane et al wrote about their blood sugar levels and risk of dementia in 2013.
Taken altogether, it seems intriguing, but I’m not into sugarcoating: it’s all so far from a therapeutic application for current patients it’s enough to turn an Alzheimer Gadfly into a sourpuss!