Is Type 1 Diabetes Hereditary?

The conversation sits behind almost every search for "is Type 1 diabetes hereditary." A clinician said the words "higher risk." No baseline, no scale, no sense of whether "higher" means a couple of percentage points or a near-certainty. The math is doable. It just didn't get done in the room.

For a future parent with T1D in the picture (yours, your partner's, or a sibling's), the honest answer is short, and most pages give it badly. Type 1 diabetes is one of the most heritable common diseases we know of. Most children of T1D parents never develop the disease. Both halves of that sentence are true at the same time, and the gap between them is exactly where a family-planning decision lives.

The American Diabetes Association puts numbers on the family-history math the way a careful clinician should have. If a father has T1D, his child's risk is roughly 1 in 17. If a mother has T1D and the child was born before she turned 25, about 1 in 25; born after, about 1 in 100. If both parents have T1D, the child's risk falls between 1 in 10 and 1 in 4. The general-population baseline in the United States is about 4 per 1,000, or roughly 1 in 250. So "higher risk" for a T1D parent is real, and the absolute number, in most family configurations, lives in the low single digits.

And here's the part most explainers miss. About 85% of new T1D diagnoses occur in people who have no family history of the disease. T1D's genetic architecture runs silently for generations and surfaces in a child whose parents and grandparents are unaffected. A diagnosis with no family history isn't an anomaly. It's the typical shape of a T1D case.

The cleanest single proof that genetic identity doesn't equal disease identity is the identical-twin number. When one identical twin develops T1D, the other reaches roughly 50% concordance at diagnosis, rising to about 65% by age 60. Same genome. Several decades of additional life. Only about fifteen extra percentage points of disease. Half of the genome's load is doing something. The other half of the load is waiting for something the genome can't supply on its own.

That something is the autoimmune trigger. T1D isn't a metabolic disease the way Type 2 diabetes is. The immune system, in genetically susceptible people, attacks and destroys the pancreatic beta cells that make insulin. Once enough beta cells are gone, blood sugar climbs and the disease becomes clinically obvious. The genetics tell you who can develop the autoimmune attack. They don't tell you whether the attack will fire. The TEDDY consortium's 2024 review lays out the field's best guesses for what fires it: enteroviruses, especially Coxsackie B viruses, are the strongest single suspect, but the field hasn't converged on a single cause.

What the genes are actually doing

For most common diseases that have a polygenic story, the genetic loading is spread across hundreds or thousands of small variants. Type 2 diabetes works that way. So does Alzheimer's. So does ADHD. T1D is structurally different. Roughly half of T1D's genetic risk is concentrated in one block of immune-system genes: the part of the genome that controls how the immune system tells self from non-self. Geneticists call that block the HLA region. About 95% of people with T1D carry one of two specific HLA variants, called HLA-DR3 and HLA-DR4, against roughly 40% of the unaffected general white US population.

That gap is what makes HLA both important and insufficient. The variants matter so much that the immune-system architecture they encode is the gating factor for whether T1D is even possible in a given person. They don't matter enough that carrying them tells you you'll get the disease. The unaffected population carrying the same variants is large. HLA is a necessary part of the answer for almost every T1D patient. It's a sufficient part of the answer for almost none.

That distinction is exactly why a generic polygenic score does badly on T1D. A score built for a flatter polygenic disease underweights the HLA region; a score built only on HLA misses the rest of the genome that decides whether HLA loading actually expresses. T1D needs a model that handles both at once.

What modern screening can read

That's the gap Herasight's cross-ancestry T1D preprint was built to close. The method models HLA haplotypes directly, then layers a Bayesian regression across roughly 7.4 million variants for the rest of the genome. In the All of Us research cohort, an ancestrally diverse NIH-funded sample, the score reaches an AUROC above 0.91 in European individuals and above 0.89 in non-European groups, outperforming three existing T1D scores in every ancestry group tested. AUROC is a discrimination measure: it tells you whether the score consistently ranks people who go on to develop T1D above those who don't. It's not the same number as "how much your child's risk drops."

Embryo screening adds a second question on top of cohort accuracy: does the score still work between siblings? Embryos in the same IVF cycle share parents, ancestry, and roughly half their DNA. If a score is just picking up population-stratification artifacts, it falls apart on that test. Herasight tested that directly. In a 2025 within-family validation across seventeen disease scores, sixteen of the seventeen showed no decrease in predictive performance compared with population-level analysis. The signal is real, even between siblings.

What this means in IVF

For families already in IVF, polygenic embryo screening for T1D rides the same biopsy already used for chromosomal screening. On Day 5 or Day 6, a small sample is taken from each embryo. That sample, originally collected for preimplantation genetic testing for aneuploidy (PGT-A), can also support preimplantation genetic testing for polygenic conditions (PGT-P). No second biopsy. Family history feeds back into the baseline so a couple with strong T1D history sees a different starting point than a couple with none, and Herasight's counselors handle the result. Very strong family history can sometimes point at a high-effect rare variant rather than a polygenic one, which is a different kind of testing question entirely.

What the screen does is rank the family's embryos by estimated T1D risk; what it can't do is prevent T1D, because the autoimmune trigger window remains for any embryo, and a low-risk genome is still a genome the trigger can in principle find. That's the honest version of the answer. Ranking changes the odds your child draws inside the genetic component. It doesn't close the door the immune system might still walk through.

For the prevention question specifically, the only FDA-approved drug in this space is Tzield (teplizumab), which delays the onset of stage 3 T1D by about two years on average in patients whose autoimmune attack has already begun (stage 2, meaning two or more T1D-related autoantibodies and dysglycemia). It's a different intervention from anything that can happen during IVF. Tzield delays the disease in a child whose pancreas is already under attack. Embryo screening, before pregnancy, can move the dial on which embryo a family transfers. Neither one is a cure.

If you want to see, in numbers, how embryo ranking could change estimated risk inside a family like yours, the calculator is the fastest way in. If you want a real conversation about what these results mean for you, reach out and we'll set up a call with one of our counselors.