Science · Health and disease areas
Type 2 diabetes is the disease for which the road before diagnosis is best described, and one of the few where prevention has been proven in trials. Here are the studies, and the distance between the two.
Type 2 diabetes does not start on the day of the diagnosis. For years before, the body handles sugar a little worse, the pancreas works harder to keep up, and some molecules in the blood drift. Three studies, verified here, describe this from three angles: five amino acids measured once, twelve years before (a group of 2,422 American adults); blood sugar and insulin measured repeatedly over thirteen years (6,538 British civil servants); and small panels of blood proteins tested against 218 diseases in 41,931 people.
What they have in common: in each, the signal was there years before anyone was ill, in groups of people. What none of them does: give you a test, a threshold, or a prediction for one person. And what sits between a signal and prevention is a separate kind of proof, which for diabetes exists, but comes from other studies: programmes of diet and physical activity have reduced the onset of diabetes in people at risk, in randomised trials.
Each card opens the full study page: first the plain-language reading, then the technical part and the abstract.
Nature Medicine · 2011
A study of 2,422 adults followed for 12 years: higher levels of five amino acids were linked to a higher risk of type 2 diabetes, independently of weight and blood sugar. An association in a group, replicated in a second group.
The Lancet · 2009
A study of 6,538 British civil servants measured repeatedly over 13 years: in the 505 who developed diabetes, blood sugar and insulin had been drifting for years and changed sharply three to six years before diagnosis. A description of a group, not a test.
Nature Medicine · 2024
A study of 41,931 adults in the UK Biobank: for 67 of 218 diseases, a small set of blood proteins measured once improved the prediction of who would fall ill within ten years, compared with age, sex and lifestyle. For most diseases it did not. Replicated for six diseases in a second cohort.
The authors of the amino-acid study write that "proven, preventive therapies exist" for diabetes, which is why identifying people early matters to them; the authors of the trajectory study hypothesise that prevention might work better before the curve accelerates, and say this needs testing. Those preventive trials (programmes of diet, physical activity and weight) are the next block of this library: the study record that the current Digital Twin page cites for them (the Diabetes Prevention Program) has not yet passed this section's verification protocol, so it is not presented here as verified. When it has, it will be linked from this paragraph.
Keep your routine results (fasting glucose, insulin if measured, glycated haemoglobin, lipids) with their dates, in order, so that the next visit starts from a series and not from a single number. That is all, and it is what the trajectory study says is worth having. The Twin does not compute a diabetes risk, does not read amino acids or protein panels, and does not say where you are on a curve: no model validated for that exists on this platform. If one is ever built and validated, this page will name it, with its population and its limits.
Whether any of these signals works in people who are not white Europeans: all three cohorts are mostly so, and all three say it. Whether measuring more often helps or only produces more false alarms. Whether a protein signature adds anything, for diabetes specifically, to glycated haemoglobin: another study reports that it does, with internal validation only, and we have not yet verified it. And whether knowing a risk years ahead changes what people do and how they fare: the question this whole section exists to keep asking.