All studies
DNA·November 20, 2026·7 min read

Why Am I Prediabetic or Diabetic When I Am Thin and Can My Raw DNA Explain It

You can be prediabetic while thin. A chip-based raw DNA file shows only a small common-variant nudge and mostly misses rare forms like MODY. Here is what to do now and what to test next.

Yes. You can be prediabetic or develop type 2 diabetes while thin, at a normal weight, and genes that affect how the pancreas releases insulin are one of the main reasons. A 23andMe or AncestryDNA raw file can show a common variant such as TCF7L2 that nudges risk up a little, but it cannot rule rare inherited forms of diabetes in or out. If your picture does not fit ordinary type 2, the answer is in a few blood tests, not in the file.

Why am I prediabetic when I am thin?

Body weight is a proxy, and not a perfect one. Blood glucose stays normal as long as two things keep up. Tissues, mostly muscle plus the liver and fat, have to respond to insulin. And the beta cells of the pancreas have to release enough of it, at the right moment. Extra weight stresses the first. But the second is inherited too, and many of the several hundred common variants linked to type 2 diabetes seem to act on how the pancreas releases insulin, not on body weight. A lean person whose beta cells release insulin a bit late or a bit short can drift into high glucose with a flat stomach.

Three other routes show up in lean people:

  • Fat in the wrong place. Some people have a low BMI but store fat in the liver and around the organs, with little under the skin. The scale cannot see it.
  • Little muscle. Muscle is one of the main places glucose goes after a meal. A smaller muscle mass is a smaller sink, so the same meal can land higher.
  • Ancestry. Type 2 appears at lower BMI in South Asian and East Asian populations. The American Diabetes Association lowers its screening threshold to a BMI of 23 for Asian Americans, against 25 for everyone else.

What it looks like on a lab report

The cutoffs below are the ones to read your own result against. They are the American Diabetes Association (ADA) values. The World Health Organization defines impaired fasting glucose from 110 mg/dL (6.1 mmol/L), so a German or other European lab may flag you at a different point.

5.7 to 6.4% (39 to 47 mmol/mol)
ADA cutoffs: HbA1c range for prediabetes (diabetes starts at 6.5%, 48 mmol/mol)
100 to 125 mg/dL (5.6 to 6.9 mmol/L)
ADA cutoffs: fasting glucose range for prediabetes (diabetes starts at 126 mg/dL, 7.0 mmol/L)
140 to 199 mg/dL (7.8 to 11.0 mmol/L)
ADA cutoffs: glucose 2 hours after a 75 g drink, prediabetes (diabetes starts at 200 mg/dL, 11.1 mmol/L)

Whether fasting or after-meal glucose is high hints at the mechanism, but not cleanly. Isolated high fasting glucose is tied to liver insulin resistance and a weak early insulin response. Isolated high 2-hour glucose is tied to muscle insulin resistance plus late insulin secretion. Fasting insulin or HOMA-IR can be ordered to gauge insulin resistance, but neither is standardised for diagnosis, so a doctor reads them as supporting information only. GCK-MODY shows a different pattern: mildly raised fasting glucose and only a small rise after a glucose drink.

One trap. HbA1c is an indirect measure, and anaemia, thalassaemia trait and some haemoglobin variants can push it up or down without any change in your glucose. If your A1c and a fasting glucose disagree, that disagreement is itself a finding.

What to do now if you are thin and prediabetic

The basics work in lean people too, and a few matter more for you:

  • Lift weights. Resistance training builds muscle, which enlarges the sink that takes glucose out of the blood after a meal.
  • Cut refined carbs and alcohol. Sugary drinks, white bread, pastries and regular alcohol raise glucose and, for alcohol, load the liver.
  • Recheck in 3 to 6 months. Repeat fasting glucose and HbA1c. If only A1c was done so far, ask for a 75 g oral glucose tolerance test (OGTT), which also shows the after-meal response.

If the numbers climb despite this, or the picture does not fit type 2, move to the tests further down.

Can my raw DNA explain it?

Partly, and weakly. TCF7L2 (the variant usually reported is rs7903146) is among the strongest common genetic signals for type 2 diabetes in European-ancestry studies. In East Asian studies KCNQ1 and other loci have effects of a comparable size, so the ranking depends on ancestry. Each TCF7L2 risk copy raises the odds by roughly 40 percent, and the effect seems to run through insulin release rather than weight, which is why it can matter in a thin person. But the variant is common. Plenty of carriers never develop diabetes and plenty of non-carriers do. It is a nudge, not a cause.

The bigger limit is the chip. A consumer array reads several hundred thousand chosen positions, mostly common ones. The variants behind monogenic diabetes are rare and spread across many genes, so the array mostly does not see them. A clean chip file is not reassurance here.

A whole-genome (30x) VCF covers far more. It includes the coding variants in GCK, HNF1A and HNF4A, the main MODY genes. But standard variant calling can miss whole-gene deletions, such as in HNF1B, and some promoter or structural changes. So a clean VCF is not a full MODY exclusion either. Any variant it does show still needs interpretation and confirmation by a doctor or genetics service. The HealthOSX ClinVar panels, as listed, do not name a MODY or diabetes panel, so do not assume the tool screens GCK, HNF1A or HNF4A.

What a DNA file can and cannot show here
QuestionCan the raw file answer it?Detail
Common type 2 variants such as TCF7L2PartlyA small shift in odds
MODY (GCK, HNF1A, HNF4A) on a chip fileMostly noVariants are rare and mostly not on the chip
MODY in a whole-genome VCFYes, with confirmationCoding variants are covered, but whole-gene deletions and some promoter or structural changes can be missed, and a doctor or genetics service must confirm any finding
LADA or type 1 autoimmunityNoNeeds antibody blood tests

Scroll the table sideways for the rest →

Raw files from chip-based tests cover common positions, not most rare disease variants. A whole-genome VCF covers far more, but findings still need clinical confirmation.

When it does not look like type 2

Some features should make you and a doctor suspect something else:

  • Diagnosis young, or in someone lean with no insulin resistance signs
  • Diabetes in a parent, grandparent and sibling, across generations
  • A mildly raised fasting glucose that has stayed flat for years
  • Rapid weight loss, ketones, or needing insulin within months

Two diagnoses matter most. MODY is a single-gene form, usually passed down in a dominant pattern, in genes such as GCK, HNF1A and HNF4A. Commonly cited estimates put it at about 1 to 5 percent of diabetes, and it is often mislabelled as type 1 or type 2. LADA is a slow autoimmune diabetes in adults, which can look like type 2 for a while in a slim person.

A concrete first step is the Exeter MODY probability calculator on diabetesgenes.org. It takes age at diagnosis, BMI, family history and a few other details and gives a rough chance of MODY, which helps you and your doctor decide whether testing is worth it. A monogenic gene panel is usually ordered through a specialist, often only after islet antibodies come back negative and C-peptide is preserved.

Next-step tests if it does not fit type 2
TestWhat it answersPoints away from ordinary type 2
C-peptideHow much insulin the pancreas still makesLow despite high glucose
Islet antibodies (GAD65, IA-2, ZnT8)Whether the immune system is attacking beta cellsAny positive
Monogenic diabetes gene panelWhether a MODY gene variant is presentA pathogenic variant, usually with a family pattern

Scroll the table sideways for the rest →

Ordered and interpreted by a doctor. Islet antibodies help separate autoimmune diabetes from insulin resistance. C-peptide is best read fasting or stimulated, alongside a glucose taken at the same time.

Does the label change anything?

Yes, and this is why it is worth chasing. Treatment follows the type. GCK-MODY is a stable, mild elevation that often needs no drug at all. HNF1A-MODY often responds to a low dose of a sulfonylurea. LADA tends to need insulin earlier than type 2 does. Whether any of that applies to you is a question for the doctor who sees your numbers. Ask for the antibodies and C-peptide before assuming the type 2 label is right.

Two common mistakes

One is treating thin as protection and ignoring a high result. The other is treating a TCF7L2 result as the explanation when it is one small piece. What is still unknown is how much of lean-person diabetes comes from muscle, liver fat or beta cell function, and for any one person nobody can tell from a consumer file.

Being lean tells you about weight, not about how well your pancreas releases insulin, and a chip-based raw DNA file cannot see most of the rare causes.

Where your file fits

If you have a raw file, it is still worth seeing the common signals in your own numbers. The free tool at healthosx.com/dna reads a 23andMe, AncestryDNA, MyHeritage or whole-genome VCF file. The story is read in the browser and asks nothing of you. The full report, including a TCF7L2 finding on how your body responds to refined carbs and an HBB/HBA2 check for thalassaemia trait, is also free after a one-time email code.

That check reads only specific thalassaemia variants. A clear result does not rule out a haemoglobin variant that skews HbA1c. A haemoglobin electrophoresis or a glucose test that does not rely on A1c does.

For a polygenic view, readers who choose to store their file can use Ask your DNA and ask "What's my type 2 diabetes risk?". The answer is a tendency, placed among the 1000 Genomes people, and it is relative. Chip files are imputed first, which takes about ten minutes. Treat all of it as context to bring to the lab conversation, not a diagnosis.

New to raw DNA files? Start with the full guide: 23andMe Raw Data Interpretation Free in Your Browser

Written by

HealthOSX Research

Keep reading