Lu, Z.

Accessing the relationship of chemotherapy agents and acute ovarian failure in female childhood cancer survivors
Zhe, L., Maira, P.G., & Yan, Y.

Chemotherapy has been shown to be associated with acute ovarian failure (AOF) in pediatric cancer survivors. Understanding how chemotherapy agents affects ovary function is important in developing a reliable predictive model of AOF risk.

Treatment exposure and menstrual history information from Childhood Cancer Survivor Study (CCSS) were used to develop AOF risk prediction model (n=5589). We relax the linearity assumption on the relationship of the exposure dose of chemotherapy agents and the outcome AOF status by employing the cubic spline functions in the logistic regression model, adjusting for other important risk factors. Model performance was evaluated using cross-validated AUC, average positive predictive value (AP), and Brier score.

From a group of 22 agents, we identified 6 top toxic chemotherapy agents, which are consistent with reports from other investigations. Some chemotherapy agents were found to have threshold effects. When compared with models assuming linear effect of 10 chemotherapy agent dosage, our revised model with fewer agents and non-linear dose relationship fit the data better and had better predictive performance.

Chemotherapy agents have different toxicity profiles for different organs. For risk prediction purposes, it is imperative to examine their effects individually without imposing a linearity assumption wherever sample size allows.