Why this is a women's disease

Graves' disease is an autoimmune disorder in which antibodies (TRAK / TRAb / TSI) bind to the TSH receptor and drive the thyroid into overdrive — too much T3 and T4. It is the most common cause of hyperthyroidism, and it overwhelmingly affects women. A woman's lifetime risk is roughly 3%, versus about 0.5% for men — a 5–8× imbalance driven by estrogen effects on the immune system, X-chromosome genetics, and reproductive transitions (pregnancy, postpartum, menopause).

Want the prevalence numbers, age-of-onset curves, and women-vs-men comparisons? Head over to the Statistics tab.

Graves' by the numbers

Who gets Graves', when, and how often — the headline epidemiology.

Compare your labs to the dataset

Enter your thyroid panel and see where it lands against ~9,100 patients in the UCI reference dataset. Saved locally in your browser — nothing uploaded.

TRAK statistics

How well TRAK performs as a diagnostic, and how its level predicts relapse after antithyroid drugs.

Pregnancy statistics

Risks of untreated maternal Graves' and the postpartum relapse curve.

Real data — women only

Filtered to female records from the UCI thyroid dataset (n=1,830 women, 64 with hyperthyroid-spectrum diagnoses).

Reading the datasets

Column schema and copy-paste shell / Python recipes for the UCI Garavan thyroid files.

Files in ./datasets/ are comma-separated, no header row, with ? for missing values and the diagnosis appended as label.|recordID.

head -5 datasets/allhyper.data

awk -F',' '{print $NF}' datasets/allhyper.data \
  | awk -F'.' '{print $1}' | sort | uniq -c

awk -F',' '$2=="F" && $NF !~ /negative/' \
  datasets/allhyper.data

import pandas as pd
cols = ['age','sex','on_thyroxine','query_thyroxine',
        'on_antithyroid','sick','pregnant','surgery',
        'I131','q_hypo','q_hyper','lithium','goitre',
        'tumor','hypopit','psych',
        'TSH_m','TSH','T3_m','T3','TT4_m','TT4',
        'T4U_m','T4U','FTI_m','FTI','TBG_m','TBG',
        'referral','diagnosis']
df = pd.read_csv('datasets/allhyper.data',
                 names=cols, na_values='?')
df['diagnosis'] = df['diagnosis'].str.split('.').str[0]
df[df.sex=='F'].diagnosis.value_counts()

Causes — why me?

Graves' isn't caused by any single thing. It's a perfect storm: a genetic predisposition that primes the immune system, plus a trigger that tips it into producing the TRAK antibody. Most of the predisposition is unchangeable. Some of the triggers are.

The female life course

Graves' interacts with every major reproductive transition. The shape below is the relative incidence across a woman's life — a peak in the reproductive years, a second peak around menopause. Hover or tap any pin to see what changes at that stage.

Symptoms women notice first

Graves' touches almost every system in the body. Symptoms below are ranked by how common they are in women with active disease (approximate, from clinical literature — varies by age and severity). Hover any bar to see more detail.

How it's diagnosed

Diagnosis follows a defined sequence: confirm hyperthyroidism with thyroid hormone labs, then confirm Graves' as the cause (rather than a toxic nodule or thyroiditis) with antibodies and/or imaging.

Differential diagnosis — Graves' vs other causes of hyperthyroidism

Course over time — what to expect

Graves' is a chronic condition with several possible long-term paths. The trajectory depends mostly on which treatment is chosen and how the immune system behaves after — relapse is common in the first 5 years, and a subset of people eventually drift into hypothyroidism even without RAI or surgery.

Common long-term patterns

Pregnancy & Graves'

For drug choices and trimester-by-trimester management, see the Treatment in pregnancy section on the Treatment tab. This page focuses on what to watch for — fetal/neonatal effects, the thyroid-storm emergency, and the postpartum relapse window.

Risk timeline — what to watch for, when

Fetal & neonatal effects

Four possible problems — two driven by maternal TRAK crossing the placenta, two driven by maternal antithyroid drugs crossing the placenta.

Thyroid storm in pregnancy

Postpartum — the relapse window

New hyperthyroid symptoms in the first 12 months postpartum are common and split into two very different conditions — they look similar but the treatment and prognosis diverge sharply.

Medical treatment

Three disease-modifying options. None is best for everyone — the choice depends on age, pregnancy plans, severity, eye involvement, and your own preference.

Beta-blockers control symptoms but don't treat the disease.

Success rates at a glance

TRAK — your lab result

The single most useful number for predicting how the disease will behave and when you can stop antithyroid drugs.

TRAK (German TSH-Rezeptor-Antikörper; English TRAb / TSI) is the autoantibody that drives Graves' — it activates the TSH receptor on the thyroid and causes the disease.

What your band predicts

Approximate relapse risk after a standard 12–18 month antithyroid drug course, based on TRAK level at the end of treatment.

Treatment in pregnancy & breastfeeding

Pregnancy changes almost every treatment decision. Methimazole becomes risky in the first trimester; radioactive iodine is off the table entirely; some supplements need to be re-thought.

The full epidemiology, fetal/neonatal considerations, and postpartum relapse window live on the Pregnancy tab — this section focuses on what to do.

Pre-conception planning

Ideally, achieve a stable euthyroid state before conception. Choices:

• On antithyroid drugs, controlled? Two accepted paths — decide with the endocrinologist before stopping contraception:
  • Pre-conception switch to PTU. Safest if conception may happen quickly — no window where an unrecognized early pregnancy (weeks 4–6) gets methimazole.
  • Stay on methimazole, switch at first positive test. Limits PTU liver exposure, but only works if she tests early and the switch happens within days — the teratogenic window (weeks 6–10) comes up fast.
• In remission, off drugs for ≥ 1 year? No PTU needed — but remission ≠ cure. TRAK antibodies can persist and cross the placenta regardless of maternal thyroid status. Required:
  • TRAK measured pre-conception and again at 18–22 weeks. If > 3× ULN, fetal monitoring (heart rate, growth, goiter on US) is indicated even if mom is euthyroid.
  • TSH / FT4 through pregnancy — relapse risk is elevated, especially postpartum (up to ~50% in the first year).
• TRAK very high (>5× ULN) or large goiter? Consider definitive therapy (surgery preferred over RAI if planning pregnancy soon) 6+ months before trying — removes the drug question entirely.
• If RAI: wait at least 6 months and confirm euthyroid before trying.
• Already pregnant on methimazole? Switch to PTU urgently in 1st trimester (lower teratogenic risk).

Trimester-by-trimester management

Off the table during pregnancy

What to ask your doctor & follow-ups

Questions for each visit, the labs that matter, and the symptoms that mean "call today, don't wait."

TSH lags weeks behind your symptoms, antibodies wax and wane, and antithyroid drugs have their own risks. Going in with a list and tracking your own numbers between visits makes a real difference.

Labs to track — the core thyroid panel

Call your doctor the same day if

Natural remedies & lifestyle

Adjuncts only — these don't replace antithyroid drugs, RAI, or surgery. But three of them have strong enough evidence to matter.

Bottom line — the three highest-evidence moves

Download all datasets

Eight sources, mixed access. Summary first; expand any row for the command-line recipe.

Dataset summary

Command-line recipes

# bash
mkdir -p datasets && cd datasets
BASE=https://archive.ics.uci.edu/ml/machine-learning-databases/thyroid-disease
for f in allhyper allbp sick-euthyroid hypothyroid sick allhypo; do
  curl -sSLO $BASE/$f.data
  curl -sSLO $BASE/$f.names 2>/dev/null
done
# index of all files
curl -s $BASE/ | grep -oE 'href="[^"]+"' | head -40

# 2007-2008 thyroid module (TSH, T4, antibodies)
mkdir -p datasets/nhanes && cd datasets/nhanes
curl -sSLO https://wwwn.cdc.gov/Nchs/Nhanes/2007-2008/THYROD_E.XPT
curl -sSLO https://wwwn.cdc.gov/Nchs/Nhanes/2007-2008/DEMO_E.XPT

# Python read
python3 -c "
import pyreadstat
df,_ = pyreadstat.read_xport('THYROD_E.XPT')
print(df.columns.tolist()[:15])
print(df.head())"

# 1. install CLI
pip install kaggle

# 2. get API token at https://www.kaggle.com/settings → Create API Token
#    saves kaggle.json — move it:
mkdir -p ~/.kaggle && mv ~/Downloads/kaggle.json ~/.kaggle/
chmod 600 ~/.kaggle/kaggle.json

# 3. download
cd datasets
kaggle datasets download -d emmanuelfwerr/thyroid-disease-data
kaggle datasets download -d yasserhessein/thyroid-disease-data-set
kaggle datasets download -d nguyenvy/nhanes-19882018
unzip -o "*.zip"

# Example: download a GEO series matrix
mkdir -p datasets/geo && cd datasets/geo
ACC=GSE9340
curl -sSLO "https://ftp.ncbi.nlm.nih.gov/geo/series/${ACC:0:5}nnn/$ACC/matrix/${ACC}_series_matrix.txt.gz"
gunzip -f ${ACC}_series_matrix.txt.gz

# Python alternative with GEOparse
pip install GEOparse
python3 -c "
import GEOparse
g = GEOparse.get_GEO('GSE9340', destdir='./')
print(g.metadata.get('summary'))"

# JSON API — no login
curl -sSL "https://clinicaltrials.gov/api/v2/studies?query.cond=Graves+Disease&pageSize=100" \
  -o datasets/clinicaltrials_graves.json
python3 -c "
import json
d = json.load(open('datasets/clinicaltrials_graves.json'))
print('Studies:', len(d['studies']))
for s in d['studies'][:5]:
    print('-', s['protocolSection']['identificationModule']['briefTitle'])"

Sources

Patient resources, clinical guidelines, and the research papers behind the numbers on this page.