In American Indian communities across the Dakotas, Oklahoma and Arizona, researchers have followed thousands of people for more than thirty years to answer one question: what actually helps? Here is what they found — the tests that catch harm early, the things worth treating sooner than you'd think, and the two biggest causes, neither of which is inside anybody's kitchen.
It does its damage quietly, over years, in four places at once: the small blood vessels at the back of the eyes, the nerves in the feet, the filters inside the kidneys, and the arteries around the heart.
By the time any of it is something a person can actually feel, much of it is permanent.
So the useful question isn't really what causes diabetes. It's this: if someone walks into a clinic today, what finds the damage while it can still be stopped?
Diabetes damages the tiny blood vessels in the retina. It causes no pain and no blurring until late — and by then the loss is permanent. Caught early, it can be treated.
When researchers brought the camera into the clinic itself, rather than referring people to a specialist hours away, 45% of people with diabetes already had damage. One in twenty had the advanced form, the kind that leads to blindness.
Almost none of them knew. And 85% of the people offered the photograph agreed to it — which matters as much as the number, because a test nobody takes finds nothing.
Berinstein 1997 · 45.3% with retinopathy, 5.1% proliferative, 413 people across two Dakota communities. Photos were graded on site so treatment could start that day.
The tool is called a monofilament. It is a single strand of nylon, about as thick as a stiff fishing line, mounted on a small handle.
A nurse presses it against the sole of the bare foot until the strand buckles into a C shape. It's manufactured to bend at exactly ten grams of pressure — light, but impossible to miss if the nerves are working. You close your eyes and say when you feel it. It takes a minute and costs almost nothing.
If you can't feel it, you've lost protective sensation. You could step on something sharp, or wear a blister through, and never notice. That's how a small wound becomes an amputation.
14% of people with diabetes couldn't feel it, against 5% of people without.
But the number that actually tells you something is this one. Among people who already have diabetes, what predicts losing sensation isn't how high their blood sugar runs — it's how long they've had it. Nerve damage accumulates.
Compared with someone diagnosed in the last five years, the odds of having lost feeling were about double at five to ten years, and around seven times higher past the ten-year mark.
So the test is close to useless on someone newly diagnosed, and it is the single most valuable minute you can spend on someone fifteen years in. That's not a reason to skip it — it's a reason to know who to chase.
Sosenko 1999 · 5–10 years vs under 5: odds ratio 2.12 (1.25–3.59). Over 10 years vs under 5: 7.04 (4.68–10.88). Both comparisons are among people with diabetes. Prevalence reached 22% in the Arizona community. This counted how common the problem was at one moment; it did not follow people forward to see who later developed an ulcer.
Protein leaking into the urine is normally read as a kidney problem. In people with diabetes, it turned out to predict something else.
A positive result meant 64% higher risk of heart attack or stroke, and 79% higher risk of dying.
The leak is a sign that small blood vessels are failing all over the body — the kidney is just the one place where it shows up in a cup. So a cheap test already being run is quietly flagging who needs their heart looked at.
One limit worth keeping: this held in people with diabetes. In people without it, the signal wasn't there.
Kizer 2009 · 1.64 for cardiovascular events, 1.79 for death.
Cholesterol. Every 10-point rise in LDL brought about 12% more cardiovascular risk — climbing in a straight line, right down into what most people would call a normal range. There was no point where it flattened out and became safe.
Guidelines draw a line at 100. The biology doesn't have a line. Lower is simply better.
Triglycerides and HDL together. High triglycerides paired with low HDL predicted coronary disease and stroke — but only in people with diabetes. In everyone else, the same pattern meant nothing.
Howard 2000 · risk rising continuously from a group average of 70. Lee 2017 · 1.54 for coronary disease, 2.13 for stroke.
There's a band below the diagnostic threshold — the range where a doctor says "let's watch it." For someone with diabetes, that band is not a waiting room.
Diabetes on its own raised cardiovascular risk substantially. Slightly-raised blood pressure on its own raised it a little. Together they were worse than either.
The two don't multiply each other; the effects stack. But stacked is still stacked — and the second one is treatable long before it's officially a diagnosis.
Zhang 2006 · 3.70 for both, versus 2.90 for diabetes alone and 1.80 for slightly-raised blood pressure alone.
Not "get fit." Not thirty minutes a day. The finding is narrower, and much more usable.
People who did any physical activity at all had about a third lower risk of developing diabetes than people who did none. Measured properly with pedometers rather than by asking, the same thing appeared: above roughly 3,500 steps a day, risk dropped.
And then it stopped dropping. People walking far more were no better off than people walking a little. Nearly the whole benefit is in getting from nothing to something.
That changes the instruction. You're not trying to turn anyone into an athlete. You're trying to move one specific group — people currently doing nothing — a short distance.
Fretts 2009 and 2012 · 33% lower odds for any activity versus none; 29% lower above 3,500 steps a day, and flat above that.
Researchers asked whether processed meat was linked to diabetes. The answer came back unusually specific.
People eating the most Spam had roughly double the odds of developing diabetes.
Plain red meat — steak, roast, ground beef — showed nothing at all.
So this isn't a story about meat, and it isn't a story about people making poor choices in a grocery aisle. It's about one specific product.
It's distributed free through the federal commodity food program — the system meant to prevent hunger on reservations. It's shelf-stable, it survives without refrigeration, and for decades it was among the most dependable sources of protein that actually got delivered.
Which means the strongest dietary finding in thirty years of this research isn't about willpower. It's about procurement. Somebody decided what went in the box.
You can't counsel your way out of that in an exam room. But you can stop telling people to cut back on red meat, because red meat was never the problem.
Fretts 2012 · Spam roughly doubled the odds; unprocessed red meat showed no effect.
Arsenic occurs naturally in groundwater across much of the American West, and many households draw from private wells that no public system ever tests.
Among people whose blood sugar was still completely normal, higher arsenic exposure meant 57% higher risk of developing diabetes later.
Across everybody together, that signal faded — which is what you'd expect, since most of the group was already further along. The risk sits with the people who look perfectly healthy today. Which is exactly why waiting for symptoms would never catch it.
Testing a well costs very little, and nobody has to change a single habit for it to work.
Grau-Pérez 2017 · the researchers committed to examining this group before seeing the results, and it's their own headline finding.
People with more social support had measurably better blood-sugar control. Across the range seen in the study, the gap came to about 0.6% of A1C.
That number deserves a pause. It's roughly what a diabetes medication is expected to deliver.
It was measured at a single moment, so it can't prove which way the arrow runs — support may improve control, or people doing better may attract more support. But the size of it makes treating isolation as a medical problem defensible rather than sentimental.
Sawyer 2025 · measured across a single survey, not followed over time.
Don't treat diet soda as the safe swap. When it was tested here, every result pointed toward more diabetes rather than less — people drinking any diet soda at all had 74% higher risk. The study was too small to call it settled, and people often switch to diet soda because they're already worried, which muddies it. But there's nothing here that makes it a solution.
Don't hang the whole plan on exercise. It works, and it plateaus quickly. It doesn't replace looking at the eyes, the feet, the kidneys and the heart.
One question genuinely came back empty: alcohol showed no link to diabetes across four years of follow-up. That one can rest.
Jensen 2020 · 74% higher risk for any diet soda versus none. Lu 2003 · alcohol flat in every group examined.
Look in four places while the damage is still silent. Treat cholesterol and blood pressure earlier than the thresholds suggest. Move the people who are doing nothing a short distance.
And then be honest that the two strongest findings here — what was in the food boxes, and what's in the well water — were never going to be solved by anything said to a patient across a desk.
Every line traces to a specific study of these communities. Where a finding couldn't carry the weight, it isn't here.