UMass Amherst biostatistician develops a statistical tool for prediction

Breast cancer is a complex disease, and its development is difficult, but important, to predict.

While many elements can contribute to breast cancer prognosis, University of Massachusetts Amherst biostatistician Chi Hyun Lee has focused attention on one emerging risk factor for its potential to predict disease progression.

Lee will use a $154,791 two-year grant from the National Institutes of Health (NIH) in an effort to develop statistical tools that will better predict breast cancer survival rates and survival time after breast cancer recurrence.

While the project focuses on breast cancer research, the proposed statistical method will have wide application to other chronic diseases, he noted.

Lee’s research involves the androgen receptor (AR), a biomarker that plays a role at the cellular level in regulating hormones, including in female sexual, somatic, and behavioral functions. However, in excess, it is associated with an increased risk of breast cancer.

“Our goal is that this new statistical approach will help us determine the prognostic value of AR and potentially lead to better targeted therapy for patients and advances in breast cancer survival,” said Lee.

For the project, Lee will work with data from the Nurses’ Health Study (NHS), one of the world’s largest prospective cohort studies investigating risk factors for major chronic diseases in women, including breast cancer. Data from this study, established in 1976, contains invaluable information for breast cancer research such as lifestyle, hormonal and genetic risk factors, including RA, and clinical outcomes such as breast cancer diagnosis, recurrence and mortality.

“In many epidemiological studies of breast cancer survival,” Lee explains, “researchers rely on hazard ratios, or the likelihood of adverse events such as death or disease progression compared to controls. This ratio is determined using a statistical method called the proportional hazard model.

“However, we have found in the NHS data that the model assumptions about the relationship between AR expression and breast cancer survival are wrong. This means that the hazard ratio results are often misleading in assessing the prognostic value of AR.”

Lee notes that another statistical method, based on limited mean survival time (RMST), has much better prognostic value. RMST is a summary metric defined as life expectancy up to a point in time, eliminating any assumptions of proportional harm that might prove wrong. RMST has many advantageous features, such as its direct interpretation and robustness.

“Specifically, we can assess the effect of prognostic factors in relation to the absolute effect, which is more clinically interpretable,” said Lee.

Lee’s funding will allow him to develop new statistical methods based on RMST to fully utilize the rich data from the Nurses’ Health Study.

As a result, she hopes to gain a better understanding of the complex effects of AR on breast cancer development and survival. Ultimately, the funding will support two goals: to develop a flexible regression method based on RMST that will be used to explain the clinical significance of AR in survival with various breast cancer subtypes; and to develop a model-independent approach to compare survival rates after breast cancer recurrence between groups with different AR statuses.