(In type 1 diabetes, blood sugar soars because the body makes no insulin, the hormone that acts like a key to allow sugar into cells. In the more common type 2 diabetes, blood sugar soars because insulin no longer works properly—that is, people are insulin resistant.)

“There’s no question that diabetes damages small blood vessels,” says David Knopman, professor of neurology at the Mayo Clinic in Minnesota.

It may also shrink parts of the brain. A recent study found more brain atrophy in 350 people with diabetes than in 363 people without the disease.

It’s not just those with diabetes who are at risk. People who have what doctors call “metabolic syndrome” also have a higher risk of cognitive decline. That’s roughly one out of three U.S. adults.

Their blood sugar levels are higher than normal, but not high enough to be diabetes. That gives them an increased risk of dementia.

And they may have high blood insulin levels because obesity—especially an oversized waist—has made them insulin resistant. (When insulin doesn’t work well, the pancreas responds by pumping out more.)

That may also spell trouble for the brain. Men with high blood insulin levels declined more on cognitive tests over three years than those with lower levels.

Why would high levels of insulin in the blood matter?

“High insulin in the body means lower insulin in the brain,” says Angela Hanson, a physician and senior fellow at the University of Washington School of Medicine.

That’s because, over time, high levels of insulin in the blood may shrink the number of receptors for insulin in the blood-brain barrier, allowing less to enter the brain, says Hanson. And insulin may help keep the brain healthy.

“Insulin helps clear toxic beta-amyloid out of the brain,” Hanson explains. “So if you put someone on a diet that increases brain insulin, you might have less of the toxic amyloid around.”

The key to lowering sugar and insulin in the blood—and presumably raising insulin in the brain—is to lose excess weight and exercise more.

But one pilot study suggests that it’s not just how much, but what you eat that matters.

Hanson and her colleagues assigned 20 older adults without mild cognitive impairment and 27 older adults with MCI to eat one of two diets. The LOW diet was low in saturated fat, and its carbs had a low glycemic index—that is, they didn’t cause a bump in blood sugar. The HIGH diet was high in saturated fat, and its carbs had a high glycemic index.

The HIGH diet was unusually high in saturated fat and sugar, but it wasn’t off the charts. “If you look at a fast-food combo meal, it’s got a sugary soda and a high-fat burger,” notes Hanson.

After four weeks, people who got the HIGH diet had higher levels of unbound beta-amyloid fragments in their cerebrospinal fluid (which bathes the brain and spinal cord), while people who ate the LOW diet had lower levels.

“The theory is that the beta-amyloid that’s not bound to fats or other lipids is free, and it’s free to wreak havoc, if you will,” says Hanson. “We believe it’s a more toxic form of beta-amyloid because it’s less likely to be cleared. But that’s hard to test in humans.”

The results seemed to fit with a finding from a similar, earlier study: the LOW diet raised insulin levels in cerebrospinal fluid (and presumably the brain), while the HIGH diet lowered insulin levels.

“A Western diet or obesity or other things that cause high blood insulin may decrease brain insulin,” says Hanson. “If you make someone less insulin resistant with weight loss or a diet, they may have more brain insulin.”

Until more studies are done, it’s too early to know if a diet lower in saturated fat and sugars can protect the brain. But the research is encouraging.

“The most striking finding from these studies was that you could change the brain chemistry of people who have mild cognitive impairment,” says Hanson.

“When I’m in my clinic, I can tell patients with MCI that if they eat a healthier diet and exercise, things might get better. That’s the message that keeps me going.”

Sources: J. Am. Geriatr. Soc. 56: 1028, 2008; Exp. Gerontol. 47: 858, 2012; Diab. Care 36: 4036, 2013; JAMA 292: 2237, 2004; N. Engl. J. Med. 369: 540, 2013; Eur. J. Pharmacol. 719: 170, 2013; Neuroepidemiol. 34: 200, 2010; JAMA Neurol. 70: 967, 972, 2013; Arch. Neurol. 68: 743, 2011.