Background:Predicting the number of oocytes retrieved (NOR) following controlled ovarian stimulation (COS) is the only way to ensure effective and safe treatment in assisted reproductive technology (ART). To date, there have been limited studies about predicting specific NOR, which hinders the development of individualized treatment in ART.
Objective:To establish an online tool for predicting NOR.
Materials and methods:In total, 621 prospective routine gonadotropin releasing hormone (GnRH) antagonist COS cycles were studied. Independent variables included age, body mass index, antral follicle counts, basal FSH, basal and increment of anti-mullerian hormone, Luteinizing hormon, estradiol, testosterone, androstenedione, and inhibin B. The outcome variable was NOR. The independent variables underwent appropriate transformation to achieve a better fit for a linear relationship with NOR. Pruned forward selection with holdback validation was then used to establish predictive models. Corrected Akaike's information criterion, Schwarz-Bayesian information criterion, scaled -log[likelihood], and the generalized coefficient of determination (R2) were used for model evaluation.
Results:A multiple negative binomial regression model was used for predicting NOR because it fitted a negative binomial distribution. We established Model 1, using basal ovarian reserve markers, and Model 2, using both basal and early dynamic markers for predicting NOR following COS. The generalized R2 values were 0.54 and 0.51 for Model 1 and 0.64 and 0.62 for Model 2 in the training and validation sets, respectively.
Conclusion:Models 1 and 2 could be applied to different scenarios. For directing the starting dose of recombinant follicle stimulation hormone (rFSH), Model 1 using basic predictors could be used prior to COS. Model 2 could be used for directing the adjustment of rFSH dosages during COS. An online tool (http://121.43.113.123:8002/) based on these two models is also developed. We anticipate that the clinical application of this tool could help the ART clinics to reduce iatrogenic ovarian under- or over-responses, and could reduce costs during COS for ART.
Keywords:NORs; negative binomial regression; online tool; predicting model; pruned forward selection with holdback validation.

---

 

乐备实（上海优宁维生物科技股份有限公司旗下全资子公司），是国内专注于提供高质量蛋白检测以及组学分析服务的实验服务专家，自2018年成立以来，乐备实不断寻求突破，公司的服务技术平台已扩展到单细胞测序、空间多组学、流式检测、超敏电化学发光、Luminex多因子检测、抗体芯片、PCR Array、ELISA、Elispot、PLA蛋白互作、多色免疫组化、DSP空间多组学等30多个，建立起了一套涵盖基因、蛋白、细胞以及组织水平实验的完整检测体系。

 
我们可提供从样本运输、储存管理、样本制备、样本检测到检测数据分析的全流程服务。凭借严格的实验室管理流程、标准化实验室操作、原始数据储存体系以及实验项目管理系统，已经为超过3000家客户单位提供服务，年检测样本超过100万，受到了广大客户的信任与支持。