Development and Internal Validation of a Lifestyle-Based Risk Prediction Model for 10-Year Incident Kidney Cancer in the 45 and Up Study

Kidney cancer is a disease that often hides in plain sight. It is the sixth most common cancer diagnosed in men and the tenth in women worldwide, and the number of cases is rising. This increase is partly due to better detection of small tumors that cause no symptoms, but also because more people are developing the risk factors that lead to the disease, such as obesity and high blood pressure. Because the disease is relatively rare in the general population, doctors have long agreed that checking everyone for it is not a practical idea. Scanning the entire population would catch very few cases while exposing millions of healthy people to unnecessary tests, anxiety, and the risk of finding harmless growths that would never cause harm. Instead, the medical community has turned its attention to a more targeted approach: identifying the specific people who are at a higher risk and focusing screening efforts on them. To do this effectively, researchers need a reliable way to sort people into risk groups using information that is already known or easily gathered.

A team of researchers in Australia has taken a significant step toward solving this problem by creating a new tool to predict who is likely to develop kidney cancer in the next ten years. They built this tool using data from the "45 and Up Study," a massive project that has followed more than 265,000 adults aged 45 and older in New South Wales since 2006. The researchers linked the personal health questionnaires these volunteers filled out at the start of the study with official cancer records up to the end of 2019. This allowed them to see exactly who developed kidney cancer after joining the study and to look back at what their lives were like before the disease appeared. They found that out of the 265,712 people they studied, 1,010 were diagnosed with kidney cancer during the average follow-up period of nearly eleven years.

The team tested a model that uses eight simple pieces of information to estimate risk: a person's age, sex, body mass index, whether they have high blood pressure or diabetes, how much they exercise, their alcohol consumption, and their smoking history. The results showed that older age, being male, having a higher body mass index, high blood pressure, and diabetes were all linked to a higher chance of developing the disease. Conversely, engaging in vigorous physical activity was linked to a lower risk. The model performed well at distinguishing between those who would develop cancer and those who would not, and it was very good at predicting the actual number of cases that occurred in different groups.

The most important finding from this work is that the model can successfully identify a small group of people who carry a much higher risk than the average person. When the researchers used the model to look for people with a predicted ten-year risk of at least 1.0 percent, they found that this group made up only about 4 to 5 percent of the total population. However, within this small group, the rate of new kidney cancer cases was more than double, and in some versions of the model, nearly four times higher, than in the rest of the population. This means that by focusing on this small slice of the community, doctors could potentially find a significant portion of the cancers that would otherwise go undetected until they were more advanced.

The researchers selected a final version of their model that included heavy smoking—defined as a history of smoking at least 30 pack-years—as a key factor. This choice was strategic because it aligns with existing guidelines for lung cancer screening, suggesting that kidney cancer screening could be added to lung cancer programs without needing entirely new systems. The model does not rely on complex genetic tests or rare biological markers; it uses everyday health data that is routinely collected in general practice. While the study was limited to an Australian population and requires further testing in other groups to confirm its accuracy everywhere, it offers a practical, scalable way to move away from guessing and toward a system where prevention and early detection are directed at the people who need them most.