Breast cancer is the most common cancer in the United States, which occurs when breast cells grow uncontrollably. There are many different types of breast cancer, but triple-negative breast cancer is considered more aggressive than other types.
Three proteins are absent in patients with triple-negative breast cancer – estrogen, progesterone, and epidermal growth factor. Estrogen and progesterone are sex hormones that regulate reproduction, whereas epidermal growth factor proteins are involved in cell growth and division. Doctors use these proteins to help determine treatment options for patients with triple-negative breast cancer. Doctors also use the stages of cancer to describe where it is located and how far it has spread. These stages are numbered from 0 to IV, with IV being the most advanced stage, when the cancer has spread to other parts of the body.
Doctors most commonly prescribe chemotherapy as the best treatment for patients with stage IV triple-negative breast cancer. This treatment uses drugs to kill breast cancer cells. Elderly patients with triple-negative breast cancer receive less chemotherapy than young patients because they experience more side effects. This difference in the treatment between the young and elderly patients could affect survival of triple-negative breast cancer patients.
Scientists in the past investigated how chemotherapy affected elderly patients with triple-negative breast cancer by researching how long the patients survived after their treatment. But they didn’t compare these effects in patients from different age groups and cancer stages, so they couldn’t resolve which patients needed additional rounds of chemotherapy.
Kaiyan Huang and his team at the Fujian Medical University Union Hospital wanted to better understand how chemotherapy can treat elderly patients with triple-negative breast cancer. They aimed to fill the gaps in previous research by testing machine learning algorithms to investigate how chemotherapy affects the survival of triple-negative breast cancer patients aged 70 and over.
The scientists used one of the largest databases on cancer statistics, called the Surveillance Epidemiology and End Results (SEER), to collect information on elderly triple-negative breast cancer patients from 2010 to 2016. They divided 4,696 patients into two groups, one group of patients who received chemotherapy, and one group who did not. They studied how long the patient survived from the time of diagnosis until they died from any cause, referred to as overall survival, and how long the patient survived from the time of cancer diagnosis or treatment until the patient died specifically from breast cancer, or breast cancer-specific survival.
First, the researchers performed statistical tests on the patients’ personal data, including their marital status, age, race, cancer stage and whether or not they had surgery. They used a type of statistical analysis, called the chi-square test, to see how the chemotherapy and non-chemotherapy groups of patients differed in terms of marital status, age, race and cancer stage.
They used a method that predicts survival time, called the Kaplan-Meier method, to assess differences in how long the chemotherapy and non-chemotherapy patient groups survived. In addition, they performed a pair-matching test between the two groups to check if they differed in receiving treatment and undergoing surgery.
The researchers found 2,122 patients received chemotherapy, while a little more than 2,500 did not. They found 86% of patients who received additional chemotherapy were younger, whereas only 49% were elderly patients. In addition, they showed the chemotherapy group had better survival rates than the non-chemotherapy group, with 25% to 35% less deaths.
Next, the team developed a machine learning model based on the patients’ cancer stage and whether they had received chemotherapy. They selected nine machine-learning algorithms to predict five-year breast cancer-specific survival and overall survival of the patients. For this purpose, they used two subsets of the data from patients who underwent chemotherapy to train the model and test its performance.
They reported the average accuracy of all nine algorithms to be about 89% for five-year breast cancer-specific survival and 86% for overall survival. Among the nine algorithms, the researchers found the LightGBM algorithm performed best in predicting five-year breast cancer-specific survival and overall survival, with accuracies of 88% and 81%, respectively.
The scientists concluded elderly patients treated with chemotherapy had better breast cancer-specific survival and overall survival. They also determined all the machine learning models accurately predicted the effect of chemotherapy on the survival of elderly patients, but suggested they be tested further to confirm their reliability.
The researchers pointed out their study was limited by the SEER database, which lacks information on the type of drug patients are given and their dosage. Regardless, the team strongly recommended chemotherapy for elderly patients with advanced triple-negative breast cancer.
