Breast MRI screening eligibility
The criteria for supplemental screening MRI, and the model question that trips people up: the 20% lifetime-risk threshold has to come from a model that handles family history in depth. Gail does not, and using it here underestimates exactly the patients the threshold exists to find.
Gail versus Tyrer-Cuzick at a glance
| Gail (NCI BCRAT) | Tyrer-Cuzick / BOADICEA | |
|---|---|---|
| Family history | First-degree relatives only; no paternal side, no age at diagnosis | Extended pedigree, both sides, ages at diagnosis, bilateral and ovarian cancer |
| Also models | Menarche, age at first live birth, biopsies, atypical hyperplasia | The above plus BMI, HRT, breast density, genetic testing results |
| Valid use | Chemoprevention eligibility (5-year risk ≥1.67%) | Lifetime risk for MRI screening eligibility (≥20%) |
| For the 20% MRI threshold | Not appropriate | Appropriate |
Notes that change what you write
- Dense breast tissue ALONE is not an indication for screening MRI under these criteria. Density raises risk modestly and lowers mammographic sensitivity, which is an argument for supplemental screening in general, but the MRI thresholds above are the ones that carry guideline support.
- Risk models are not interchangeable and will disagree, sometimes by a wide margin, on the same patient. State which model produced the number you are acting on.
- These are screening criteria for asymptomatic patients. A palpable abnormality or a suspicious mammographic finding is a diagnostic problem and follows a different pathway regardless of risk category.
Which model, and where to run it
| Model | Gives you | Use it for | What it does and does not see |
|---|---|---|---|
| Gail / BCRAT | 5-year and lifetime invasive breast cancer risk | Chemoprevention eligibility (5-year risk 1.67% or more) | First-degree relatives only. No paternal lineage, no age at diagnosis, no BRCA status, no density. Not valid for MRI screening eligibility. |
| Tyrer-Cuzick (IBIS) v8 | 10-year and lifetime risk | MRI screening eligibility (lifetime 20% or more) | Extended pedigree both sides, ages at diagnosis, BMI, HRT, breast density, BRCA results. The usual choice for the 20% threshold. |
| BOADICEA / CanRisk | Breast and ovarian risk, plus carrier probability | Detailed family history; genetic counselling referral | Models pedigree and known pathogenic variants most thoroughly. Also estimates the chance of carrying a variant, which the others do not. |
Why there is no risk calculator on this page
We do not reimplement these models here, and that is deliberate. Each is a coefficient-table model: race and ethnicity stratified regression coefficients plus SEER-derived baseline hazard and competing-mortality tables, several hundred published numbers in total. A transcription slip anywhere in that produces a percentage that looks authoritative and is wrong, and a wrong risk figure changes whether someone gets an MRI or starts tamoxifen. The official implementations are maintained and validated by the groups that built the models. Use those, then bring the number back to the eligibility criteria above.