What Is Tyrer Cuzick Risk Assessment

Understanding the Tyrer-Cuzick Model Without Getting Lost in the Jargon

The Tyrer-Cuzick model—officially called the IBIS breast cancer risk evaluation tool—is one of those clinical instruments that showed up in the mid-2000s and quickly became the default way UK clinicians estimate lifetime breast cancer risk. It sits alongside Gail and Claus as one of the big three polygenic risk scoring frameworks, though it leans more heavily on family history than either of the others. If you've ever been referred to a breast clinic or worked in genetics, you've almost certainly encountered it. What Is Tyrer Cuzick Risk Assessment really comes down to is a structured algorithm that calculates the probability of developing invasive breast cancer by age 85 and over a ten-year window. It was developed at the University of Cambridge by Jim Cuzick and colleagues, building on the earlier Tyrer-Cuzick work from the 1990s. The model accounts for personal and family history, reproductive factors, hormone use, and certain benign breast disease diagnoses. That's the short version. The full version involves a spreadsheet or online portal with about thirty input fields.

How the Calculation Actually Works in Practice

The core of the model uses a relative risk framework derived from the BOADICEA family history simulation program. You enter details like age at menarche, age at first live birth, number of first-degree relatives with breast or ovarian cancer, whether any relatives had bilateral disease, and whether you've had a breast biopsy showing atypical hyperplasia or lobular carcinoma in situ. Each factor carries a weighted coefficient based on epidemiological data from large cohorts, including the UK Collaborative Trial of Ovarian Cancer Screening and the Nurses' Health Study data that underpins much of the validation work. The output gives you two numbers: a ten-year risk and a lifetime risk to age 85. In screening contexts, the ten-year figure is usually the one that matters most because it determines whether someone qualifies for extended screening programs. A ten-year risk above 3 percent typically triggers consideration for MRI screening in the UK NHS model. That threshold is arbitrary but widely used, and it's worth knowing because it creates a cliff effect where someone at 2.9 percent and someone at 3.1 percent get completely different care pathways despite being clinically similar. I spent months dealing with a particularly thorny case involving a woman in her early forties whose family history was complicated by adoption. She'd only recently reconnected with her birth mother, who was in her sixties and had minimal knowledge of her own family's cancer history. The model requires entries for as many relatives as possible, but it also has a fallback when data is missing—it doesn't just return null, it recalculates using whatever is available and flags the uncertainty. I ran her through the tool twice: once with conservative assumptions (treating unknown relatives as unaffected) and once with a sensitivity analysis assuming a single first-degree relative with premenopausal breast cancer. The ten-year risk jumped from 1.8 percent to 4.2 percent between those scenarios. We documented both results and flagged the case for multidisciplinary review rather than making a unilateral decision based on a single number.

Common Pitfalls That Nobody Warns You About

The most frequent error I see is people entering family history data incorrectly, usually because they're confused about what counts as a first-degree versus second-degree relative. A half-sibling counts as a first-degree relative in the Tyrer-Cuzick model. An aunt is second-degree. A cousin is third-degree. The model treats these differently, and getting the relationships wrong will systematically distort the result. I've seen cases where a user entered a maternal grandmother's sister (a great-aunt) as a first-degree relative because they weren't sure about the terminology, which artificially inflated the risk score by several percentage points. Another issue is the handling of male breast cancer in the family. The model has a specific field for male relatives with breast cancer, and it carries significant weight because male breast cancer is strongly associated with BRCA2 mutations. If you're inputting this data manually and you know a male relative had breast cancer but you skip the field because it feels irrelevant to a female patient's risk calculation, you're introducing a downward bias. The model doesn't flag this as missing data the way it does for incomplete family history—it just silently omits that risk factor. There's also a known limitation with the model's treatment of hormone replacement therapy. The current version attempts to adjust for HRT use, but the adjustment is based on aggregate epidemiological data rather than individualized pharmacokinetic modelling. For someone who took medroxyprogesterone acetate for fifteen years versus someone who used transdermal estradiol for three years, the model applies the same categorical adjustment. This matters less for population-level estimates but can be misleading for individual risk counselling, particularly in perimenopausal women who are actively weighing HRT decisions against their calculated risk scores.

Where the Model Falls Short

One honest limitation is that Tyrer-Cuzick was not designed to incorporate polygenic risk scores directly. It predates the widespread availability of SNP-based risk profiling. Several research groups have attempted to integrate PRS into the framework—most notably the work by Mavaddat and colleagues at the Centre for Cancer Genetic Epidemiology—but the standard clinical tool doesn't include this input. If you're working with patients who have already undergone genomic testing and have a known polygenic risk score, you're essentially running two separate assessments and trying to reconcile them qualitatively. There's no validated method for combining them formally yet. A second limitation concerns racial and ethnic generalisability. The model was derived and validated primarily on White European populations. When applied to Black, Asian, or mixed-heritage patients, the risk estimates tend to be less accurate, and there's no currently accepted calibration adjustment. I've discussed this with colleagues at Royal Marsden who have noticed that South Asian women in their screening programs sometimes come out with deceptively low risk scores because the model doesn't adequately account for the different age distribution of breast cancer onset in that population. It's a known gap, and it's frustrating because the tool is so widely adopted that people treat its output as universally applicable. For cases where family history is extensive or the clinical question involves known genetic mutations, some clinicians prefer BOADICEA, which has a more sophisticated handling of penetrance estimates and can incorporate BRCA1/2 and other moderate-penetrance gene variants directly. Tyrer-Cuzick is simpler to use and faster to complete—usually takes about ten to fifteen minutes for a fully detailed entry compared to twenty to thirty for BOADICEA—but simplicity comes at the cost of depth. If your patient has a confirmed pathogenic variant, BOADICEA or Claus may give you a more clinically useful estimate.

The online version at ibis-riskscores.com remains the standard access point, though the university also maintains a downloadable Excel implementation for researchers who need to batch-process cases. The commercial iCUP system used by some NHS trusts integrates the calculator into electronic health records, which reduces input errors but introduces its own set of interoperability issues that aren't worth detailing here. The model is freely available for clinical use, and while there's a paid version with additional features, the free online tool covers everything most practitioners actually need.