Does Lifting Weights Prevent Type 2 Diabetes? A 143,715-Person Study

Quick answer: In a Harvard-led cohort study of 143,715 US health professionals published in JAMA Network Open in June 2026, Zhang and colleagues found that people who lifted consistently through midlife had a 42% lower rate of type 2 diabetes. People whose training came and went showed no reduction at all. This is observational data, so it shows association, not proof.

Every few years a headline says lifting is good for your blood sugar and nobody changes anything, because the advice is always the same: do some resistance training.

This one is more specific, and more uncomfortable. It looked at what people did for twenty years, not what they did in one questionnaire, and the pattern that mattered was not how much they lifted. It was whether they kept lifting.

What the researchers actually did

The team pooled three long-running US cohorts: the Nurses' Health Study, the Nurses' Health Study II, and the Health Professionals Follow-up Study. Participants filled in a questionnaire every two to four years, for up to fourteen rounds each, reporting how many hours a week they spent on resistance training.

After excluding anyone who already had diabetes, heart disease or cancer, 143,715 people remained. Over a mean 19.2 years, 10,038 of them developed type 2 diabetes.

The key move is that resistance training was not treated as one number. It was averaged across every questionnaire, so somebody who lifted hard for two years and quit does not get counted the same as somebody who lifted moderately for twenty. In the Nurses' Health Study II, the researchers went further and sorted women into five twenty-year patterns between the ages of 40 and 60.

What they found

More lifting went with less diabetes, and the curve flattened early.

Diabetes risk fell with more weekly lifting, then levelled off

00.511.5

1

0.83

0.83

0.73

0.73

None

Under 0.5 h/wk

0.5 to 1 h/wk

1 to 2 h/wk

2+ h/wk

View the data

Group	None	Under 0.5 h/wk	0.5 to 1 h/wk	1 to 2 h/wk	2+ h/wk
Hazard ratio vs no lifting	1	0.83	0.83	0.73	0.73

Half an hour a week already went with a 17% lower rate. Going from one hour to four hours added nothing on top. That is a blunt but useful message: the gap between nothing and something is much bigger than the gap between something and a lot.

Then comes the part that makes this study worth reading twice.

Twenty years of consistency, not peak effort, tracked with lower risk

00.511.5

1

1.02

0.82

0.79

0.58

Consistently low

Fluctuating

High to low

Low to high

Consistently high

View the data

Group	Consistently low	Fluctuating	High to low	Low to high	Consistently high
Hazard ratio vs consistently low	1	1.02	0.82	0.79	0.58

The consistently high group had a 42% lower rate. Women who started late and kept going still landed 21% lower. Women who trained hard and stopped kept some of it, at 18% lower.

The fluctuating group came in at 1.02, which is statistically indistinguishable from doing almost nothing for twenty years.

In other words, on-and-off lifting tracked with the same diabetes rate as barely lifting at all, once diet, smoking and aerobic activity were accounted for.

Why does consistency matter more than volume?

Because muscle is a storage tank for blood sugar, and the tank shrinks when you stop.

Skeletal muscle is where most of the glucose in your bloodstream ends up after a meal. Training makes the muscle bigger and makes it pull glucose in more readily. Both effects fade within weeks of stopping. A person who trains in six-month bursts spends much of each decade back at their starting insulin sensitivity.

The researchers are careful here, and so should we be: this mechanism is the accepted explanation in the literature, not something this study measured. Nobody took a muscle biopsy. What they measured was questionnaires and diagnoses.

There is a second, less flattering explanation worth naming. People who report lifting for twenty straight years are different from people who do not, in ways no statistical model fully removes. The team adjusted for diet quality, smoking, alcohol, calories, family history and total aerobic activity, and the association held. It still is not the same as proof.

Did aerobic work and sitting change the picture?

Yes, and the combination beat any single behaviour.

People who met the aerobic guideline (15 or more MET-hours a week), met the resistance guideline (1 or more hours a week) and watched under two hours of TV a day had a rate 62% lower than people who met none of the three.

One more finding for anyone who trains around an injury: participants who did upper-body work only still showed a lower rate, at 25% below those who did nothing.

What this study doesn't prove

• It is observational, so causation is off the table. Nobody was assigned to lift. People who choose to lift for two decades differ from people who do not, and adjustment reduces that problem without solving it.
• Resistance training was self-reported hours, not sets. The questionnaire asked about time, so "1 to 2 hours a week" covers a serious three-day programme and a slow circuit of machines equally. There is no way to separate them here.
• The trajectory analysis is women only. Those five twenty-year patterns come from the Nurses' Health Study II. The dose-response result covers both sexes, but the headline 42% figure does not.
• The participants are not the general population. 96.7% were White and all were US health professionals, a group with above-average income, health literacy and access to care.
• The flat top of the curve may be a measurement artefact. Beyond two hours a week the numbers of cases get small, and self-reported hours get less reliable at the high end. The study cannot say whether four hours a week really adds nothing.
• No conflicts of interest were declared, and funding came from the National Institutes of Health. That is about as clean as funding gets, and it does not fix any of the above.

How to test this in your own training

The study reports twenty-year averages across 143,715 people. You are one person, and the only measurement you control is your own attendance.

1. Count sessions per month, not per week. A weekly target makes one missed session look like a 33% failure. A monthly count of 10 to 12 absorbs a bad week without triggering the quit response.
2. Set the floor absurdly low. The lowest category in this study, under half an hour a week, still tracked with a 17% lower rate. Two hard sets on a bad day beats a skipped week, and it keeps the streak in your log intact.
3. Log the sessions you almost skipped. Those are the data. Six months from now the useful number is not your best month, it is your worst one.
4. Review consistency every quarter, not every week. Open your log, count training months out of the last twelve, and look for the gaps rather than the peaks. Gaps are what this study says to fix.
5. Add the cheap extras. Meeting the aerobic guideline and cutting TV under two hours a day was where the largest risk reduction sat. Neither needs a gym.

Your log already answers this

Consistency is the one training variable you cannot feel and cannot remember accurately. Ask anyone how often they trained in March and they will guess high, which is exactly why a written record beats a recollection here.

If you write your sessions into Apple Notes or a notebook, the record already exists. What is missing is the view: sessions per month, month after month, so a three-month gap in 2025 is visible instead of forgotten.

Gym Note Plus reads plain gym notes and builds that view for you.

• Type your session the way you always have and it becomes structured, dated data.
• Session counts and per-exercise trends show up without any tagging.
• Old notes import, so the history you already wrote comes with you.
• Gaps show up as gaps, which is the point.

Curious what your own text looks like as data? Paste a recent session into the free notes to workout translator.

The lifters in this study who trained on and off for twenty years ended up where the people who barely trained ended up. Turning up beat trying hard.

Frequently Asked Questions

What did the 2026 JAMA Network Open study on resistance training and diabetes find?

Zhang and colleagues followed 143,715 US health professionals for a mean of 19.2 years and recorded 10,038 new cases of type 2 diabetes. Two or more hours a week of resistance training was associated with a 27% lower rate. Women who lifted consistently from age 40 to 60 had a 42% lower rate.

How much resistance training is needed to lower type 2 diabetes risk?

In the JAMA Network Open cohort, under half an hour a week was already associated with a 17% lower rate of type 2 diabetes than doing none. The benefit levelled off around one hour a week, with no further reduction between one hour and four hours. Consistency across years mattered more than weekly hours.

Does stopping and restarting weight training still help?

In this study, no. Women whose resistance training fluctuated across twenty years had a hazard ratio of 1.02, meaning their diabetes rate matched that of women who consistently did very little. Women who started late and then kept going showed a 21% lower rate, so sustained restarting did count.

Does cardio or lifting matter more for preventing type 2 diabetes?

The study did not rank them. It found that resistance training remained associated with lower risk even after adjusting for total aerobic activity, and that meeting both guidelines while limiting TV to under two hours a day went with a 62% lower rate than meeting none. The combination outperformed any single behaviour.

Can this study prove lifting weights prevents diabetes?

No. It is a prospective cohort study, so it reports associations between reported behaviour and later diagnoses. Participants chose their own training, and people who lift for decades differ from those who do not in ways statistical adjustment cannot fully remove. A randomised trial of that length is not feasible.

Final Takeaways

• The 42% figure belongs to consistency, not intensity. It describes women who lifted through their forties and fifties without a long gap.
• Fluctuating training scored 1.02. Twenty years of starting and stopping matched twenty years of barely training.
• The curve flattens at about an hour a week. Getting off zero is worth far more than adding a fourth session.
• Upper body only still counted, at 25% below no training, which matters if you train around an injury.
• Association, not proof. Nobody was randomised, everyone was a US health professional, and the twenty-year patterns are from women only.
• Attendance is the variable to log. Sessions per month over a year tells you more here than any single workout ever will. Our guide to tracking gym progress covers how to read that view, and why am I not making progress in the gym covers what gaps usually turn out to be.