UH Seidman Team First to Validate PREDICT Survival Prognostication Tool in American Women with Breast Cancer
May 31, 2022
Online tool can help facilitate complex treatment decision-making in women with early-stage disease
Innovations in Cancer | Summer 2022
How will different treatments after surgery for early-stage breast cancer affect a woman’s overall chances of survival? Patients and their physicians wrestle with these questions every day. Now, American women having these conversations with their doctors have new access to validated statistical predictions about the likely impact of different treatment options on their overall survival, thanks to new research from University Hospitals Seidman Cancer Center.
Lifen Cao, MD, PhD, Clinical Research Fellow, and her mentor, Alberto Montero, MD, MBA, Co-Director of the Breast Cancer Program at UH Seidman, and Professor of Medicine at Case Western Reserve University, School of Medicine, led a UH research team that analyzed predictions made by the online tool PREDICT UK against actual American outcomes data for hundreds of thousands of early-stage breast cancer patients in the large National Cancer Database (NCDB). The UH team found that PREDICT UK did indeed accurately predict both five- and 10-year overall survival in American women with early-stage breast cancer. Dr. Cao and Dr. Montero presented their findings recently at the virtual National Comprehensive Cancer Network 2022 Annual Conference, where Dr. Cao won an award for top scoring abstracts.
PREDICT UK uses a patient’s individual clinical characteristics and the pathology characteristics of the tumor to give a percentage estimate how different therapies after surgery (adjuvant therapies) might affect overall survival, both five and ten years out. Adjuvant therapies estimated with PREDICT UK include: chemotherapy, endocrine therapy, trastuzumab (Herceptin), and bisphosphonates. PREDICT UK is validated and used in the United Kingdom, but it had never been validated in the U.S., which has a more ethnically diverse population.
“We wanted to validate the tool here using the largest available database in the U.S. that has outcome information,” Dr. Montero says. “In breast cancer, there is a need to have accurate clinical calculators that can estimate the risk of recurrence and impact adjuvant therapies have on reducing breast cancer mortality. Having a clinical tool like PREDICT UK is very useful when oncologists have discussions with the patients on adjuvant treatment decisions.”
The large NCDB database was ideally suited for this project, Dr. Cao says. Jointly sponsored by the American College of Surgeons and the American Cancer Society, NCDB is a clinical oncology database sourced from hospital registry data representing more than 70% of all newly diagnosed cancer cases in the U.S. From 2004-2017, there were approximately 3 million breast cancer cases collected in this database.
“It is a very large and comprehensive database with long follow-up,” Dr. Cao says. “The high-quality data available in the NCDB over a long time period provides an ideal database for assessing the validity of the PREDICT tool in a much larger and more diverse cohort of breast cancer patients,” Dr. Cao says.
The UH Seidman team’s analysis included women between the ages of 25 and 85 diagnosed with primary invasive breast cancer in one breast who underwent breast surgery between 2004 and 2012. This represented 708,652 eligible patients with complete five-year clinical and follow-up data and 233,455 patients with complete 10-year follow-up.
Results show that the correlation between PREDICT UK’S predicted outcomes and actual outcomes in the U.S. population was quite close, at 0.77 for five-year overall survival and 0.78 for 10-year overall survival. To put this in a context, Dr. Montero says, a good clinical predictive test typically has a correlation value between 0.65 and 0.7.
“Calibration results showed relatively close agreement between predicted and observed deaths,” Dr. Cao says.
“Dr. Cao’s research shows that PREDICT UK is a very good tool when we validated it in the breast cancer data set here in the U.S. in NCDB,” Dr. Montero adds.
Dr. Cao and Dr. Montero say they hope the results of this study will encourage more oncologists and surgeons treating women with early-stage breast cancer to use the PREDICT tool as part of their discussions with patients.
“Given our successful validation of the PREDICT tool using NCDB patient data, we conclude that PREDICT is able to accurately predict five- and 10-year overall survival in a contemporary population of U.S. patients with early-stage breast cancer, and therefore should be a tool that clinicians could use as part of their adjuvant therapy decision-making,” Dr. Cao says.
“A lot of people just don’t know about this tool,” Dr. Montero says. “This validation of PREDICT UK in a large data set of patients in the U.S. will give people more confidence to use it within the clinic because the NCDB data is reflective of the breast cancer patients we’re taking care of here in the U.S. When we are having discussions with patients about whether they should receive adjuvant chemotherapy or endocrine therapy, the most common question is: ‘What is the benefit?’ The ability to quantify that for an individual patient, chemotherapy will reduce their risk of dying of breast cancer mortality by 5 percentage points vs. 1 percentage point really informs the discussion. Instead of adjuvant therapy in breast cancer being a binary ‘yes’ or ‘no,’ decision, having accurate clinical calculators helps providers to have a more nuanced discussion with their patients, about whether chemo, for example, will provide them with a modest or a significant benefit.”