Breast cancer risk tools fall short for women with family history — Cochrane

What just happened

A Cochrane systematic review published in June 2026 has evaluated the four tools most commonly used to estimate breast cancer risk in women with a family history of the disease. The verdict: all four show only modest accuracy. None achieves the precision needed for fully personalised clinical decisions.

This is worth sitting with — particularly for women in their 40s who have been told they have an “elevated” or “moderately increased” risk based on a family history assessment, and who are making decisions about how often to screen, whether to see a specialist, and whether to pursue genetic testing.

The Cochrane review is an international finding — the lead researchers are from Trinity College Dublin and St James’s Hospital in Ireland. But the models they evaluated are the same ones used across Australian general practice and familial cancer clinics. The finding applies here.

---

The both-and

What was reviewed

The research team analysed 45 studies examining four breast cancer risk prediction models:

• Gail (BCRAT) — one of the oldest and most-used models, originally developed from US data
• Tyrer-Cuzick (IBIS) — incorporates hormone use and a broader family network than Gail
• BOADICEA — the most comprehensive; includes paternal family history, hormonal exposure, and BRCA variant status; used through the free CanRisk web tool
• BRCAPRO — focuses specifically on BRCA1/2 mutation probability

Among these:

• BOADICEA showed the most balanced overall performance
• Gail and BOADICEA produced estimates that aligned reasonably with actual cancer occurrences
• Tyrer-Cuzick tended to overestimate risk
• BRCAPRO tended to underestimate risk

Crucially, all models showed only modest discriminatory accuracy — meaning they are reasonably good at distinguishing between groups of women (higher-risk versus lower-risk populations), but are not precise enough to give a meaningfully personalised number for any given individual.

Why this matters for patients

When a GP or familial cancer clinic tells a woman that her 10-year breast cancer risk is, say, 12%, it can sound more precise than it is. That number comes from a model built on population data. It represents the average outcome for many women who share similar characteristics — not a calibrated forecast for that specific woman’s biology.

The distinction matters when it drives decisions: whether to get mammograms every 12 months versus every 2 years, whether to pursue MRI surveillance, whether to refer for genetic counselling, or whether to consider preventive medication like tamoxifen or raloxifene. Each of those decisions carries different trade-offs, and they can hinge on a number that the Cochrane review now tells us is less reliable than the clinical setting implies.

Why models are used despite their limits

Risk assessment tools are still clinically useful — the Cochrane review does not say abandon them. They are better than no assessment. They identify women who should be screened more frequently or referred to specialist services, and they provide a structured framework for conversations that would otherwise be vague.

The issue is calibration rather than uselessness. The models do their job at the population level — correctly stratifying most women into risk tiers. What they cannot do is give a precise number that applies to each individual woman within those tiers.

In Australian practice, Cancer Australia’s family history framework classifies women into near-population risk, moderately increased risk, and high risk. Even with imprecise models, these categories hold clinical value: a woman categorised as high risk will be screened more intensively than one at near-population risk, and that differential is supported by evidence regardless of the specific number attached to it.

The gap for women with family history

The Cochrane review specifically flags women with a family history as the group for whom current models fall furthest short. This is partly because these models were mostly developed on women at average population risk. Women with a meaningful family history — especially where the history involves early-onset breast cancer, bilateral disease, or multiple relatives — may sit outside the population on which the tool was validated.

The review’s authors call for continued model refinement and better-quality primary studies. Until then, the most honest clinical position is: risk models are one input among several, not a standalone answer.

---

My 2 cents

If you have had a breast cancer risk assessment — whether through your GP or a familial cancer clinic — the Cochrane review does not mean the assessment was wrong or should be ignored. What it means is that the number is an approximation.

A few things worth knowing:

BOADICEA is the most comprehensive of the four models. If you had a risk assessment that used only first-degree relatives and age, or if you are not sure what model was applied, it is worth asking. A familial cancer clinic can run a more thorough assessment using CanRisk (the free web-based BOADICEA tool), which considers your paternal family history and hormonal exposure history in addition to the basics.

The risk category matters more than the specific number. Being classified as moderately increased or high risk drives what screening you are eligible for — more frequent mammograms, MRI screening in some cases, genetic testing referral. That classification is clinically meaningful even if the exact probability figure is imprecise.

BreastScreen Australia is free for women 40 and over — no referral needed, no co-payment. Women aged 50–74 are the target group, but 40–49 and 75+ can self-refer. If you have a family history that has not been formally assessed, a screening mammogram and a GP visit are both reasonable starting points.

---

Verdict

Verdict: yes — worth knowing about.

(International source — Cochrane review from Trinity College Dublin; applies directly to Australian clinical practice since the same four models are used here.)

The Cochrane finding matters because risk numbers can be imbued with more certainty than they carry. Women making decisions about screening intensity, genetic testing, and preventive treatment deserve to know the tools underpinning those decisions are less precise than the specificity of the number implies. That is not a reason to disengage from family history assessment — it is a reason to engage with it more critically.

---

Sources cited

1. McGarrigle S et al. Breast cancer risk prediction models for women with a family history of breast or ovarian cancer. Cochrane Database of Systematic Reviews, 2026. https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD013185.pub2
2. Cochrane. Breast cancer risk models fall short for women with family history, 1 June 2026. https://www.cochrane.org/about-us/news/breast-cancer-risk-models-fall-short-women-family-history
3. BreastScreen Australia. Free mammography screening. https://www.breastscreen.org.au
4. Cancer Australia. Breast cancer risk factors — family history. https://www.canceraustralia.gov.au/cancer-types/breast-cancer/risk-factors/family-history