Predicting in vitro fertilization success in the Brazilian public health system: a machine learning approach.

Infertility has become a global health problem, increasing the number of couples looking for in vitro fertilization (IVF). Despite advances and technical improvements, some couples remain childless due to the high complexity of the technique. The use of machine learning (ML) in the prediction of pregnancy, computing factors that could interfere in the effectiveness of the treatment, is an important tool to optimize these factors and reach the success of pregnancy. The aim of this study was to apply ML models to determine variables related to pregnancy after IVF in a public health service, including pre-implantation variables. This study included 771 women who underwent IVF treatment at Hospital das Clínicas, Federal University of Minas Gerais, between 2013 and 2019. We used the following Machine Learning algorithms: Logistic Regression, Random Forest, XG Boost and Support Vector Machines. The Random Forest algorithm achieved the best performance, with better accuracy, sensitivity and area under the ROC curve to predict the success of IVF evaluated by pregnancy frequency. We also trained a specific model only for women older than 35 years old. Variables in the Random Forest model related to pregnancy after in vitro fertilization.

Activin betaA subunit, follistatin and follistatin-like 3 are expressed in the endometrium of ovariectomized rats and regulated by estrogen replacement.

Activin A is a growth factor expressed in the endometrium, where it modulates tissue remodeling and enhances decidualization. The effects of activin A are counteracted by two binding proteins, namely follistatin and follistatin-like 3 (FSTL3). We have evaluated the effects of estrogen and progestin on the endometrial expression of activin betaA subunit, follistatin and FSTL3 in ovariectomized rats. Adult female Wistar rats (n = 21) were ovariectomized and received one week later a single dose of estradiol benzoate (1.5 mg/kg body weight, i.m. injection), either alone (n = 7) or associated with depot medroxyprogesterone acetate (3 mg/kg body weight, i.m. injection, n = 7), or oil vehicle (control group, n = 7). One week later, activin betaA subunit mRNA levels had increased significantly in the uteri of rats treated with estradiol alone (7.4 fold increase over controls, P < 0.05) and to the same extent in rats receiving estradiol plus medroxyprogesterone (6.1 fold increase over controls, P < 0.05). This was accompanied by increase of betaA subunit immunostaining in estradiol and estroprogestin treated rats, which was noted only in the surface endometrial epithelium. Follistatin mRNA expression, conversely, showed a significant decrease in the groups treated with estrogen alone and estrogen plus progestin (P < 0.05), and follistatin immunostaining in the glandular epithelium was weaker in estradiol and estroprogestin-treated rats compared to controls. FSTL3 expression was similar in the 3 groups. In conclusion, the expression of activin betaA subunit increases and that of follistatin decreases following estrogen replacement in the endometrium of ovariectomized rats, and these effects are not further altered by the addition of progestin.

Tissue specific localization of angiotensin-(1-7) and its receptor Mas in the uterus of ovariectomized rats.

The vasoactive peptide angiotensin (Ang)-(1-7) has vasodilator, antifibrotic and antihypertrophic properties, but little is known about its regulation in the uterus. The aim of this study was to evaluate Ang-(1-7) and its receptor Mas expression throughout rat uterine tissues, in ovariectomized animals treated with estrogen alone or combined with progestin. Adult Wistar rats (n = 19) were ovariectomized and randomly assigned into three different groups 1 week later. One group received a single dose of estradiol benzoate (1.5 mg/kg, i.m. injection, n = 6). Another group received estradiol associated with depot medroxyprogesterone acetate (3 mg/kg, i.m. injection, n = 6). Control group (n = 7) received oil injection. One week later, the rats were euthanized and their uteri were fixed and stained by immunohistochemistry, using a polyclonal antibody specific to Ang-(1-7) and its receptor Mas. Ang-(1-7) was detected in all uterine tissues, but it was weak or absent in the circular myometrium of treated animals. The intensity of the immunostaining decreased in the glandular epithelium of hormonally treated animals when compared to controls. In estrogen treated rats, Ang-(1-7) labeling was scattered and sometimes included the nuclei of glandular cells. We also detected Ang-(1-7) expression in longitudinal myometrium and uterine serosa. Mas receptor was present in all tissues with similar intensity among the tissue types in the control and estrogen plus progestin groups. In the estrogen group, Mas staining was stronger in the luminal and glandular epithelium when compared with stroma or circular myometrium. In conclusion, ovarian steroids are not required to allow endometrial expression of Ang-(1-7) and its receptor Mas in rats, as it remains abundant in ovariectomized animals. However, estrogen and progestin may modulate the distribution pattern of this peptide in the endometrium, especially in the glandular compartment.