Abstract
Study question
How is the cumulative pregnancy probability of individual patients after IVF-ET,could we develop a visualized clinical model to predict it based on patient’s characteristics?
Summary answer
The visualized clinical mode incorporates five items of female age, number of oocytes, antral follicle count, endometrium thickness and basal FSH level.
What is known already
Many factors can result in infertility, prognosis prediction is clinically relevant for making the right therapeutic strategy while avoiding overtreatment. It is also helpful in counselling, making the patient aware of possible treatment duration and estimated expense and managing patient’s expectation. Visualized clinical mode and accurate prediction would also be helpful in designing clinical trials to evaluate new treatments.
Study design, size, duration
We conducted a retrospective analysis of a single-center database using prospectively collected data from women who underwent IVF/ICSI treatment from January 2013 to December 2015, All the participants were followed up for at least 2 years, 3538 IVF-ET cycles were included in the study.A total of 3538 IVF/ICSI cycles were included in the study.
Participants/materials, setting, methods
Data from a total of 2312 IVF/ICSI cycles from January 2013 to December 2014 were randomly split into training dataset (1550, 67%) and internal validation dataset (762, 33%). A total of 1226 IVF/ICSI cycles in 2015 was applied to external validation dataset (temporal validation)
Main results and the role of chance
Multivariable logistic regression model combined with restricted cubic splines function was used to test independent prognostic factors and estimate their effects on treatment outcome for patients treated with IVF/ICSI. Female age, number of oocytes retrieved, AFC, endometrium thickness and basal FSH were included the final model. The above model was used to calculate prediction scores for all women in the training and validation datasets. The C-index was 0.693 (95% CI: 0.692∼0.695) in training sets, 0.689 in internal validation sets and 0.710 in external validation sets, which denotes a good performance. Calibration curves suggest excellent model calibration, with an ideal agreement between the prediction and actual observation . The DCA showed that if the threshold probability is between 0 and 0.7, using the nomogram derived in the present study to predict cumulative pregnancy provided a greater benefit than either thetreat-all or the treat-none strategy.
Limitations, reasons for caution
it was a retrospective, single-center study.In the future, prospective, randomized controlled, multicenter clinical studies will be designed.
Wider implications of the findings: The visualized nomogram model provides great predictive value for infertility patients in their first IVF/ICSI cycle, and predicts the pregnancy probability of individuals ,and could help clinicians improving clinical counselling.
Trial registration number
Not applicable