O-197 Predicting individualised risk of developing amenorrhoea in young breast cancer patients: an international multi-cohort study

Abstract

Abstract Study question What is the likelihood of developing amenorrhoea after breast cancer treatment in young women with early breast cancer? Summary answer We developed risk prediction model for a web-based tool to provide an individualized risk of developing amenorrhoea of young breast cancer patients at diagnosis. What is known already Young breast cancer survivors have an increased risk of developing premature ovarian insufficiency as a long-term adverse effect of chemotherapy. The potential for developing amenorrhoea is a high priority for these young patients, with some choosing less optimal cancer treatments to minimise the impact on fertility and/or unnecessarily taking on the physical and financial burden of fertility preservation. A key component in good-quality decision-making is understanding the likelihood of outcomes. Yet, current tools to predict individualised ovarian function after breast cancer treatments are imprecise with minimal applicability for clinical practice. Study design, size, duration The FoRECAsT (Fertility after Cancer Predictor) study is a multi-institutional study of 5-year survivors of breast cancer. A literature review addressing the impact of chemotherapy regimens for breast cancer on developing amenorrhoea was conducted. Authors of identified articles and known data registries were contacted and invited to contribute their data to the FoRECAsT database. The FoRECAsT database contains 7473 individuals sourced from Australia, UK, USA, Hong Kong, France, Denmark, Italy, Belgium and International Trial Groups. Participants/materials, setting, methods FoRECAsT cohort includes premenopausal women diagnosed with primary breast cancer, reported history of two or more predictors of developing amenorrhoea and menstrual history information after chemotherapy. Patients with bilateral salpingo-oophorectomy were excluded. Primary outcomes were likelihood of developing amenorrhea at 12 and 24 months. Cross imputation was used to manage missing values. Logistic regression was used to develop risk prediction models and prediction performance was evaluated internally using area under the receiver operating curve (AUC). Main results and the role of chance Out of 7473 individual records from FoRECAsT database, 2833 participants (37.91%) reported menstrual history information at 12 months and 2118 participants (28.34%) reported menstrual history information at 24 months and, thus were included in the analysis. In multivariate logistic regression analyses, common predictors for amenorrhea at 12 and 24 months were older age at diagnosis, lower body mass index, chemotherapy regimens, endocrine therapy and pre-treatment follicle stimulating hormone. In addition, history of smoking and alcohol consumption were also predictors for developing amenorrhea at 12-months. Receiver–operator characteristic analysis produced an estimated AUC of 0.88, sensitivity of 95.75% and specificity of 58.52% for amenorrhoea at 12 months and AUC of 0.92, sensitivity of 94.09% and specificity of 53.58% for amenorrhoea at 24 months. Internal validation with 1000 bootstrap resampling showed good discrimination for both models, C-index of 0·88 (95% CI 0·84–0·91) for amenorrhoea at 12 month and C-index of 0·93 (95% CI 0.90 − 0.94) amenorrhoea at 24 months. Based on these models, a web-based calculator that predicts individualised ovarian function is under development for implementing in clinical practice worldwide. Limitations, reasons for caution The limitations including reliance on self-reported menstrual history, heterogeneity in collecting data in different countries could result in bias and insufficiency. These models require external validation. However, excellent prediction performance of the models and good precision in the estimate of accuracy in internal validation provides reassurance. Wider implications of the findings The model and the associated web application could help clinical discussions and decision-making regarding fertility preservation before chemotherapy when appropriate and to optimise adjuvant endocrine therapy in young women with early breast cancer. Additionally, this tool could be adapted to newer breast cancer treatments and for other cancer treatments. Trial registration number N/A