Use of the models could lead to more personalized surgical treatment.
Photo/Thumb have been modified. Courtesy of Giovanni Cancemi - stock.adobe.com.
Up to one in every three lymph node operations on patients with breast cancer could be avoided, according to research published in the journals Clinical Cancer Research and BMC Cancer.
A prediction model identified 6 to 7% more women with healthy lymph nodes than other models, researchers found. If the model was used to predict the spread of breast cancer to lymph nodes, it is possible to reduce the number of lymph node operations by up to 30%, the study authors noted.
“The results indicate that we may be a step closer to more personalized surgical treatment by using the prediction models as a decision support tool,” said Lisa Rydén, M.D., Ph.D., professor of surgery at Lund University in Sweden.
Researchers studied gene expressions from approximately 3,000 breast and other tumors and patient-related factors linking the spread of disease to the lymph nodes. Tumor size and the invasion of cancer cells into vessels were significant factors in predicting the spread of disease, the researchers found. The findings were published in the journal Clinical Cancer Research.
The research team developed a prediction model for patients with hormone-sensitive breast tumors based on the tumor’s genetic profile and routinely collected data on tumor characteristics. The model identified 6 to 7% more women with healthy lymph nodes than other models.
Researchers then developed three prediction models through artificial neural networks. One identified healthy lymph nodes, one identified limited disease in the lymph nodes, and one was for widespread lymph node disease indicating more extensive surgery or primary oncological treatment with chemotherapy.
The models were used on 800 patients with primary breast cancer. Researchers extracted clinical and radiological data from the patients’ files.
The prediction models showed better discriminatory performance than multivariable logistic regression models and showed promise as decision support tools for estimating nodal disease, researchers concluded.
Further studies are needed on other patient data to confirm the reliability and precision of the models and independently evaluate the results, Rydén said.