LAUSR

Type 2 diabetes (T2D) is an important cause of mortality, disability and health care expenditures. Prevention is crucial to counter global trends, but individual variability of different risk factors leading to T2D has made prevention difficult to achieve. We aimed to identify clusters of individuals based on individual level characteristics of different risk factors using a data driven approach. We used data from the Stockholm Diabetes Prevention Program. A population-based, prospective study. Healthy participants who were born in Sweden and aged 35 to 55 years old were recruited between 1992 and 1998 and followed up after 10 and 20 years. At each visit, participants answered extensive questionnaires, anthropometric measures, laboratory examinations and an oral glucose tolerance test. We used age, sex, family history of diabetes, fasting and two hours glucose and insulin, body mass index (BMI), systolic and diastolic blood pressure, and level of education were at baseline to group participants using the k-prototype algorithm. We then examined the risk of diabetes between clusters using survival analysis. A total of 7,173 participants representing 138,942 person years of follow-up were included in this study. Among them, 998 (14%) developed T2D. We identified six stable clusters: the group with the lowest cumulative incidence of T2D (2.3%) was used as reference (n = 1,265). In the group with the highest risk, 47% developed T2D (n = 772, HR 26.3, 95%CI 17.9-38.5) followed by 17.1% (n = 1,146, HR 7.6, 95%CI 5.1-11.22), 15.2% (n = 1,453, HR 6.7, 95%CI 4.5-9.8), 9.5% (n = 1,384, HR 4.1, 95%CI 2.8-6.2) and 5.2% (n = 1,157, HR 2.0, 95%CI 1.3-3.1) in the remaining groups. There is important variability between individuals regarding the effect of different risk factors on the incidence of T2D. Data driven categorization strategies can be useful epidemiological tools to pave the way for more precise T2D intervention programs. Prevention strategies for T2D usually follow a one-size fits all approach, ignoring the important variability of risk factors and their consequences among individuals. It is possible to identify subgroups of people with different risk of T2D based on simple clinical and phenotypical characteristics, and could help to improve prevention and treatment of T2D.