American College of Surgeons Develops Cancer Survival Prediction Tool

According to research presented at the American College of Surgeons (ACS) Clinical Congress 2023, a machine learning (ML) tool can accurately estimate patient-specific prognoses for thyroid, pancreatic, and breast cancer.

Current methods to predict cancer patients’ survival rates rely mostly on cancer stage, which the researchers indicated may not account for additional nuances that influence survival.

“There is a multitude of other factors that may influence a patient’s survival beyond just their staging criteria,” explained lead study author Lauren Janczewski, MD, a clinical scholar with ACS Cancer Programs and a general surgical resident at Northwestern University McGaw Medical Center, Chicago, in a press release. “We sought to develop this Cancer Survival Calculator to provide a more personalized estimate of what patients can expect regarding their cancer prognosis.”

The research focused on identifying patient, tumor, and treatment characteristics that would have the greatest impact on five-year patient survival across thyroid, pancreatic, and breast cancers.

The research team began by taking expert recommendations for which characteristics to investigate and gathering data from the National Cancer Database (NCDB) of patients diagnosed with any of the three cancer types in 2015 and 2017. In total, records from 76,624 thyroid cancer patients, 84,514 pancreatic cancer patients, and 259,485 breast cancer patients were included.

Approximately 75 percent of those data were used to train the ML tool to recognize associations between characteristics at diagnosis and patients’ survival at five years. From there, the model ranked the factors that had the most significant impact on survival.

This analysis revealed that various characteristics specific to patients, tumors, and treatments greatly influenced survival across each type of cancer.

The top four factors associated with survival outcomes were also identified for each cancer type.

For thyroid cancer, these were age at diagnosis, tumor size, time to treatment, and lymph node involvement.

Factors relevant for pancreatic cancer survival were cancer surgery, histology, tumor size, and age at diagnosis.

Breast cancer outcomes depended on whether the patient had cancer surgery, the patient’s age at diagnosis, tumor size, and time from diagnosis to treatment. Additionally, the presence of the biomarker Ki-67 and hormone receptor status were also relevant for estimating breast cancer survival rates.

The researchers noted that while some of these factors are part of cancer staging, their findings demonstrate that many factors beyond those related to disease stage influence a cancer patient’s survival.

Overall, the tool was highly accurate at estimating cancer survival rates for all three types of cancer, with predictions falling within nine to ten months of actual survival.

Moving forward, the researchers aim to finalize the tool’s interface to make it useable in clinical practice, in addition to pilot testing the model at selected cancer centers.

Further, the research team aims to expand the tool by adding the other cancer sites included in the NCDB.

This research highlights a growing interest in how artificial intelligence (AI) applications may potentially improve cancer care.

Last year, a team from Case Western Reserve University leveraged an AI tool to identify which head and neck cancer patients would benefit from reducing the intensity of their treatments, including radiotherapy and chemotherapy.

While many patients with human papillomavirus (HPV)-driven head and neck cancer benefit from aggressive forms of these treatments, the model revealed that a significant portion of patients may be receiving more aggressive treatment than necessary.