Australia's melanoma risk calculators mostly underestimate your true risk

Australia and New Zealand have the highest melanoma rates in the world. That is not a statistic that changes with better counting — it is a function of UV exposure, skin type, and culture. Around 16,000 Australians are diagnosed with melanoma each year, and roughly 1,800 die from it.

Into this context, a new Australian study has examined something that matters to anyone relying on a risk calculator to gauge whether they need a skin check: how accurate are these tools, actually?

The short answer, according to research published in JAMA Dermatology in October 2026 and reported this week in the Medical Republic, is that most melanoma risk prediction models are moderately useful for ranking risk — but most underestimate how likely a person actually is to develop melanoma. And that gap matters.

What the researchers did

Researchers from the University of Sydney, Cancer Council NSW and Melanoma Institute Australia externally validated six published melanoma risk models using data from the Sax Institute’s 45 and Up Study — a large-scale longitudinal study of people aged 45 and over in New South Wales. The analysis included more than 43,000 participants, 946 invasive melanomas and 1,406 melanomas in situ, with over a decade of cancer registry follow-up. This is one of the largest external validation efforts for melanoma risk tools published to date.

They assessed two things: discrimination (how well a model separates people who go on to develop melanoma from those who do not) and calibration (how accurately a model’s predicted probability matches the actual observed rate).

On discrimination, five of the six models performed moderately well — the best, an Australian model by Vuong and colleagues, and the Olsen model covering invasive and in situ melanoma, each achieved an area under the curve of 0.70. For context, 0.5 is chance and 1.0 is perfect. These models do better than chance at identifying higher-risk individuals. The sixth model — the French Quéreux model — scored 0.58, performing poorly in this Australian cohort after achieving 0.73 in its original study.

On calibration, the picture was weaker. Four of the five models for which calibration data were available underestimated the number of melanomas diagnosed in the cohort. The best-calibrated model was the Olsen invasive-and-in-situ model, which slightly underestimated risk in lower-risk participants and slightly overestimated it in the highest-risk group.

Why tools underperform — and why this cohort may be different

The researchers identified several possible explanations for the underestimation. The 45 and Up Study participants had more frequent clinical skin checks than the broader national survey population (44% versus 37% among people aged 45 to 69). Some risk factors were missing from the data — including hair colour, sunbed use and sunburn history. And some variables were collected in a 2020 follow-up survey, after some diagnoses had already occurred, which may introduce bias into the analysis.

The researchers were also careful to note that findings may not generalise to all population subgroups. Validating tools specifically for people from culturally and linguistically diverse backgrounds, Indigenous Australians, and younger age groups remains a gap.

The broader conclusion the team reached is sobering for the direction this field is heading: there is “limited evidence” on which tools are most accurate or most clinically useful, and standardised risk thresholds or agreed protocols for acting on a predicted risk score do not yet exist. Some recalibration of existing models may be needed before they are suitable for routine clinical use.

Both and

It would be easy to read this study as bad news for risk tools and stop there. But there is a more useful frame.

The moderate discrimination results are real — these models are better than nothing at identifying people who warrant more intensive surveillance. A person in the highest-risk decile of the Vuong or Olsen model genuinely does face a higher probability of melanoma than someone in the lowest-risk decile. That ranking utility has value in clinical triage when resources for surveillance are limited.

The problem is not the concept of risk stratification. It is the precision gap between ranking and predicting. A tool that tells you “you are higher risk than the median” is a clinical aid. A tool that tells you “your five-year risk of melanoma is 2.3%” when the real figure might be 4.1% is a tool that may inadvertently create false reassurance.

Australia is currently considering whether a national targeted melanoma screening program is feasible. That conversation requires knowing which risk tools are reliable enough to identify who should be screened. This study is a useful input into that debate — it shows the tools need further work before being embedded in a national program, and it names exactly what that work looks like: recalibration with more complete risk factor data and validation in diverse subpopulations.

2 cents

The practical implication here is not to stop paying attention to melanoma risk — the opposite. A moderate or low score on an online risk calculator is not clearance. These tools underestimate. Australia’s UV burden does not care about a calculator score.

The standard of care in Australia remains a clinical skin examination by a trained clinician — GP or dermatologist. If you have fair skin, a history of sunburn, a family history of melanoma, more than 50 moles, or any changing or unusual spot, your GP can assess your personal risk profile and recommend an appropriate check frequency. The Cancer Council Australia recommends being sun smart and attending regular skin checks based on your personal risk.

If you are using a risk calculator as part of that conversation with your GP, that is reasonable — but use it as a conversation-starter, not a verdict.

Verdict: yes — a significant Australian study that recalibrates how much weight GPs and patients should place on melanoma risk calculator scores, relevant to anyone navigating skin check decisions.

---

Sources cited

1. Melanoma risk tools under the microscope — Medical Republic, 5 November 2026
2. JAMA Dermatology — Journal of the American Medical Association
3. Cancer Council Australia — melanoma information