A year with a community health worker, and no fewer diabetes diagnoses: 5.9% against 4.3% in 2,509 New Yorkers with prediabetes. The coaching averaged about five contacts

We've known for more than 20 years that lifestyle change can hold off type 2 diabetes. The US Diabetes Prevention Program cut new cases by 58% against placebo, with a goal of at least 7% weight loss and 150 minutes of activity a week. The hard part is delivering that at a busy city clinic. Community health workers are one proposed answer: people hired from the patients' own communities and trained in coaching. A review of 22 studies cited by this trial's authors found they improved blood sugar and weight and were cost-effective.

A trial published in PLOS One on 30 October tested the idea in New York City, at the Bellevue safety-net hospital and the Veterans Affairs clinics in Manhattan, between 2018 and 2022. Fourteen primary care teams were randomised, seven to each side. Adults aged 18 to 75 with an HbA1c of 5.7 to 6.4% and no diabetes diagnosis were eligible. On one side, patients were offered a health worker who drew up a personal plan on weight, diet and activity with them, ran education sessions, talked through their doctor's visits, and pointed them to weight clinics, farmers markets and social services. On the other, patients carried on with usual care and didn't know they were in a study.

By the end, up to 18 months in, 43 of 728 patients with a health worker (5.9%) had a diabetes diagnosis, against 76 of 1,781 with usual care (4.3%). The gap runs the wrong way, but it isn't statistically significant (p = 0.10, or 0.28 after adjustment), so the fair reading is that the programme showed no benefit, not that it did harm. Everything else stayed flat as well. Average weight changed by +0.36 lb with a health worker and -0.12 lb without, and HbA1c rose by about 0.04 points in both groups.

Weight may be the simplest explanation. Weight loss was at the centre of the Diabetes Prevention Program's lifestyle goal, and here neither group's average moved. The authors say much the same: without a difference in weight loss, no difference in diabetes isn't a surprise. Contact was the other weak spot. Weekly contact for six months and monthly for six more would come to about 30 contacts, by our arithmetic, though the protocol said "up to". The paper reports a mean of about five. So this is less a test of intensive coaching than of a light version of it, which may be closer to what a stretched clinic can actually run.

The comparison itself has soft spots. Doctors on the health-worker side could remove patients from the list, and 646 of 2,280 were taken off. Then people had to say yes: of 1,374 selected, 849 enrolled and 728 finished the first survey and got the programme, and those 728 are the ones in the results. Control patients were simply picked at random from their teams' lists, without being asked anything. The two groups ended up different in ways that could matter. The health-worker group was older (56.9 against 55.0), less often female (32.7% against 41.4%) and smoked more (24.7% against 20.6%), and its average heart-risk score was higher. The authors raise selection bias themselves, and with so few diabetes cases they kept their adjustment to age, sex and clinician.

Then there's how diabetes was counted. A single diagnosis code in the medical record was enough, with no study blood test at set times. And the people with a health worker saw their doctors more: 77.6% had a primary care visit, against 62.5%. More visits mean more chances for a doctor to spot something and write it down. The authors suggest exactly this for one odd finding, a higher rate of ischaemic heart disease with a health worker (10.6% against 6.4%), which they think may reflect more diagnosis rather than more disease. They don't apply it to diabetes itself, but by our reading the same thing could have happened, and it would push the count against the programme.

The extra engagement is the result the authors want to hold on to. It's clearest for primary care, which stayed significant after adjustment, with 2.87 visits per patient against 2.33. Emergency visits per patient were higher too, 0.73 against 0.54, while the share of patients with any emergency visit (35.0% against 29.1%) lost its significance once adjusted. Whether more care eventually pays off in better health is something this trial can't show, with follow-up of 18 months at most.

Two more things temper the result. Only 4.7% of the whole sample got a diagnosis, well below the roughly 10% a year the authors cite for prediabetes, which suggests lower-risk patients and less power to detect a difference. Some may no longer have had prediabetes by the time they enrolled, since the qualifying test could be two years old. The trial also spanned the first COVID years, which the paper doesn't discuss, though the pandemic could well have affected clinic visits and home meetings. None of this overturns what's known about lifestyle change. If you have prediabetes, the Diabetes Prevention Program's result still stands. What this trial suggests is that a light version of coaching, on its own, may not be enough to get you there.