Six kinds of pre-diabetes: the one that diet and exercise can't stop

Eight years after completing a 24-month diet and exercise program, a small group of participants in a German diabetes prevention study had done everything the standard advice asks. They had restructured their meals, lost a mean 8% of their body weight, and kept it off. Their fasting blood sugar still rose. The TULIP study, published this month in the journal Diabetes, reports the reason. Their weight loss was real. The biology it was supposed to slow was not what their biology was running on.

TULIP, the Tübingen Lifestyle Intervention Program, enrolled 190 people at elevated risk for type 2 diabetes and tracked them for 8.7 years on average after the two-year program ended. The headline number is reassuring: across the full cohort, sustained weight loss of at least 3% tracked with better long-term blood-sugar control. Sixty participants met that bar. Lifestyle intervention, the kind that public-health campaigns are built around, is doing what it is supposed to do for most of them.

Earlier work by the same German Center for Diabetes Research (DZD) team at University Hospital Tübingen, and colleagues at Helmholtz Munich, sorted people with pre-diabetes into six clusters, defined by insulin resistance, insulin secretion, body fat distribution, blood lipids, and other metabolic markers. Clusters 3 and 5 carried the highest baseline risk of progressing to type 2 diabetes, but for different reasons. Cluster 3 is dominated by poor insulin secretion, with the β-cells (the insulin-producing cells in the pancreas) under-secreting from the start. Cluster 5 is dominated by severe insulin resistance, with β-cells forced to overwork until they wear out.

In TULIP's 60-person weight-loss group, 17 participants fell into cluster 5. Their adjusted fasting blood sugar rose faster than in any other cluster after weight loss, and their two-hour post-meal glucose rose faster than the cluster 3 group. P-values in the paper are reported as significant; the trend lines on the figures do not need statistical translation. Diet and exercise bought them some metabolic flexibility, but the β-cell trajectory kept moving in the same direction.

Cluster 5's pathology is not the kind that weight loss alone is built to interrupt. The standard lifestyle-intervention model assumes the dominant problem is excess weight driving insulin resistance, and that losing weight reverses enough of that resistance to spare the β-cells. In cluster 5, insulin resistance is severe from the start and the β-cell workload is correspondingly high. The TULIP data suggest the cells are already on a decline trajectory that an 8% body-weight drop does not reset. They keep declining.

For cluster-5 patients, identified by their metabolic profile rather than by their bathroom scale, the next clinical step is not a stricter diet. It is earlier identification through screening that goes beyond BMI, and earlier additional intervention, including pharmacologic or β-cell-targeted strategies, layered on top of the lifestyle work the patient is already doing.

A common alternative reading has done real damage in diabetes care: that the patient must have failed the program. The TULIP cohort is too small to redefine clinical practice on its own, and the cluster framework is a research classification, not a routine diagnosis. But the signal is strong enough that the field now has a more honest answer to a familiar question. Weight loss protects most people at risk for type 2 diabetes. For the one group whose β-cells are running on a different engine, protection has to start somewhere else.