Before chemotherapy even begins, information from a breast biopsy may help estimate whether the treatment is likely to erase all visible cancer—and whether that result will matter for the person’s future. After chemotherapy given before surgery, a complete disappearance of detectable cancer is an established marker of prognosis in people with breast cancer.

The prediction uses information available from a biopsy, especially markers measured in the tumor, to help identify people who could benefit from chemotherapy before surgery. The central idea is to combine several pieces of information instead of dividing people by one marker alone.

It is like judging whether a garden will grow by looking at the soil, sunlight, and water together, rather than checking only the soil. The tools estimate each person’s likelihood of complete response from clinical information and three routine tumor markers, either as measured in ordinary care or grouped using thresholds.

The most precise tool was then tested in several independent groups, rather than only in the group used to build it. The final model used the version of those tumor markers that performed best when tested on unseen parts of the data. Its prediction was turned into an online calculator and a chart that gives a person’s estimated chance of complete response.

The same information also contributes to a score that provides prognostic information about the risk of cancer spreading to distant parts of the body. In the study group, the reported results showed that two hundred seventy-three patients, representing twenty-three point four percent of all patients, achieved a complete response.

In separate external validation groups, the model’s performance, measured by area under the curve, ranged from zero point seven five to zero point eight three. Combining the estimated chance of complete response with the separate recurrence-risk score made the prediction of disease-free survival more precise.

The more likely chemotherapy is to fully clear the tumour, the more strongly that result predicts longer disease-free survival. At a predicted chance of fifty percent, having that complete response was linked with about one-fifth the risk seen without it. Across the three analyzed treatment arms, the model’s area under the curve ranged from zero point seven five four to zero point seven nine five.

However, its estimates were too high for some patients and too low for others in the GeparSepto treatment arms, suggesting that the model needed recalibration there. For some groups, the model predicted chances that were systematically too high or too low, so it should be recalibrated before being used with similar patients.

An online calculator and chart provide updated estimates for those patients. The differences between the original group and the outside groups were expected, and more testing across regions, time periods, and healthcare institutions could clarify and improve the model.

Adding other tumor markers might improve accuracy, while people with uncommon tumor types may need prediction tools built from larger groups specific to those cancers. A prediction tool built from routine clinical and biopsy information estimated the chance of complete response and became more useful when paired with a separate estimate of recurrence risk.

That could help tailor treatment discussions, though more testing is needed.