In a significant development towards improving prognostic information for breast cancer patients, the PREDICT breast model has recently been upgraded to version 3.0. The new version utilizes a more comprehensive dataset that includes patients diagnosed between 2000 and 2017. The primary goal of this update is to refine the hazard ratio estimates for existing variables and incorporate the impact of radiotherapy and chemotherapy on mortality.
PREDICT v3.0: Advancing Prognostic Accuracy
The updated model now takes into consideration the diagnosis year, recognizing the advancements in treatment techniques over time and the subsequent impact on prognosis. A multivariable fractional polynomial model was implemented to fit the data, with a good resemblance found between tumor size and mortality hazard functions, as mirrored by the Eastern Cancer Registry.
Compared to its previous iteration, PREDICT v2.2, the new model exhibited improved calibration. In both the development and validation datasets, the predictions for breast cancer-specific mortality were within 10% of the observed values. However, the model slightly underestimated non-breast cancer mortality in the validation sets.
Implications for Patient Care
Reclassification under the updated model indicated that around 13% of cases would be classified into different risk groups compared to the older version. This enhanced discrimination by PREDICT v3.0 indicates its potential to provide more accurate prognostic information for breast cancer patients. Consequently, it can assist in tailoring treatment strategies based on the expected benefits of adjuvant chemotherapy.
Further Research
In parallel research, scientists identified 921 rodent mammary carcinogens that may increase the risk of developing breast cancer in humans. These carcinogens were identified through databases such as the International Agency for Research on Cancer (IARC) and ToxCast from the U.S. Environmental Protection Agency.
Another study discussed the use of deep learning techniques to analyze histopathological images for breast cancer detection. A proposed ensemble deep transfer convolutional neural network achieved an impressive accuracy of 97.72% for multiclass classification and 99.2% for binary classification using the benchmark BreaKHis dataset.
Furthermore, other researchers examined a mathematical model that combines an oral estrogen receptor inhibitor with a selective inhibitor of cyclin-dependent kinases for the treatment of breast cancer. The study confirmed the efficacy of this combination therapy.
