Primary ovarian insufficiency prediction in adult survivors of childhood cancer: model concerns – Authors' reply

Considering Wallace and colleagues’3 evaluation of 19 patients with cumulative ovarian radiation dose exposures of 30 Gy, our ovarian dose model estimates high predicted primary ovarian insufficiency risks (71–92%) by 21 years for all patients (table), consistent with the sample primary ovarian insu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:The lancet oncology 2024-02, Vol.25 (2), p.e59-e60
Hauptverfasser: Im, Cindy, Lu, Zhe, Yasui, Yutaka, Hudson, Melissa M, Ness, Kirsten K, Armstrong, Gregory T, Mostoufi-Moab, Sogol, Yuan, Yan
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Considering Wallace and colleagues’3 evaluation of 19 patients with cumulative ovarian radiation dose exposures of 30 Gy, our ovarian dose model estimates high predicted primary ovarian insufficiency risks (71–92%) by 21 years for all patients (table), consistent with the sample primary ovarian insufficiency prevalence in the study by Wallace and colleagues.3 We recommend applying the ovarian dose model when such information is available. Regarding comments from Fangjieyi Zheng and colleagues, our age-specific XGBoost4 model was tuned by optimising model performance for hyperparameters including tree depth, step size shrinkage, and maximum number of boosting rounds under the nested cross-validation framework.1 The model's excellent external validation performance (AUROCs: 0·92–0·95, across ages 21–40 years) suggests the tuning process was adequate. Given the performance similarity across different algorithms in both this study and a previous study5 of predicting acute ovarian failure using random forests and support vector machines, including high-quality treatment predictors with large effect sizes could primarily drive the robustness of results.
ISSN:1470-2045
1474-5488
DOI:10.1016/S1470-2045(24)00027-5