For a condition that often seems to appear out of nowhere, in childhood, in adolescence, sometimes even in adulthood, Type 1 diabetes may actually have much earlier roots than we thought. A new study from the INITIALISE consortium, published in Nature Communications, suggests that the earliest signs of risk may already be present at the moment a child is born.

The study’s authors are Angelica P. Ahrens, Raquel Dias, Tuulia Hyötyläinen, Matej Orešič, Eric W. Triplett, and Johnny Ludvigsson, representing the University of Florida, Örebro University, the University of Turku, and Linköping University.

Most research on Type 1 diabetes has focused on children with known high genetic risk, but most children who develop the disease don’t fall into that category. The researchers instead used data from the All Babies in Southeast Sweden (ABIS) cohort, which enrolled over 16,000 newborns between 1997 and 1999 and has tracked their health since birth. Umbilical Cord blood was collected and stored for nearly every child. Some later developed Type 1 diabetes, diagnosed anywhere from age 2 to their mid-twenties; most did not. This long follow-up let researchers compare blood samples collected years, sometimes decades, before diagnosis.

Using Olink proteomic panels, the researchers screened cord blood for inflammatory and immune-related proteins, comparing samples from 146 children who went on to develop Type 1 diabetes with 286 who did not. They identified 32 proteins that differed significantly between the two groups, present at birth, years, sometimes over a decade, before diagnosis. Several of these differences held up even after accounting for prenatal and perinatal factors, including family history of diabetes.

Key findings

• Two proteins in particular, HLA-DRA and IDS, were consistently higher at birth in children who later developed T1D, regardless of their genetic (HLA) risk status.
• Other proteins, including TIMP3, ADA, and CD40LG, were consistently lower in children who did not go on to develop the disease, suggesting a possible protective role.
• Using machine learning, the team built a model based on a subset of these proteins that predicted future T1D with high accuracy (AUC = 0.89), performing just as well whether or not genetic (HLA) risk information was included.
• Some of the identified markers were linked to differences in specific metabolites and to persistent environmental chemicals, including PFOS and other PFAS compounds, that are known to be widespread in the environment.
• The strongest proteomic differences were seen in children diagnosed with T1D before age five, pointing to particularly early biological disruption in this group.

Why it matters

Current approaches to identifying T1D risk rely on genetic testing and on detecting autoantibodies, immune markers that only appear once the immune system has already begun attacking insulin-producing cells. By that point, the biological processes driving disease may already be underway. This study suggests that some of the relevant biology may be detectable much earlier, at birth, through a non-invasive, low-cost source: cord blood that is otherwise discarded.

The authors note that these findings are preliminary and based on one Swedish cohort. Larger, more diverse populations will need to be studied to confirm the findings and to better understand the biological mechanisms behind them. Still, the results open a possible path toward earlier, more proactive screening and, eventually, earlier opportunities to intervene before the immune attack on insulin-producing cells begins.

Read the full paper: Ahrens, A.P., Dias, R., Hyötyläinen, T., Orešič, M., Triplett, E.W. & Ludvigsson, J. (2026). The inflammatory path toward type 1 diabetes begins during pregnancy. Nature Communications 17, 979.