Abstract


OBJECTIVE: Live birth predictive models for IUI are based on independent retrospective data, resulting in poor precision and inconclusive external validations. Recent developments in predictive models highlight the need of hypotheses based on prior evidence synthesized by a meta-model synthesizing conducted on previous models. We relate our results based on this approach applied on a new large data base.
DESIGN: Single-center retrospective study view more

OBJECTIVE: Live birth predictive models for IUI are based on independent retrospective data, resulting in poor precision and inconclusive external validations. Recent developments in predictive models highlight the need of hypotheses based on prior evidence synthesized by a meta-model synthesizing conducted on previous models. We relate our results based on this approach applied on a new large data base.
DESIGN: Single-center retrospective study
MATERIALS AND METHODS: Single-center retrospective data including all couples (n=4337), and 13668 cycles (on Intent to treat basis) between January 2000 and March 2014. Inclusion criteria were at least one open tube and total motile sperm count after preparation (NTMS)≥106/ml. All cycles were hyper-stimulated with FSH/HMG +/- clomifene citrate. Ovulation was triggered by hCG, and IUI performed 36-40 hours later. The logistic model was built by a pre-determined series of hierarchically related models, each subsequent model defined as more general in assessing the significance of a new subset of variables as determined by prior evidence synthesized by the meta-analysis, in avoiding biases due to exploratory analysis. At a last stage, our resulting model is compared with the basic meta-model, and a cumulative meta-analytical calculation terminates in updating the meta-model.
RESULTS: Unadjusted Live-birth rate (LBR) per cycle was 12.4%, and cumulative LBR per patient 32.5%.

OR, 95%CI and p of the model fitted on data, and aggregate resuts accounting for prior evidence
Model fitted on new data Cumulative Meta-analysis
Effect OR 95% CI p OR 95% CI p I2
Age (quadratic Age) .99 0.99, 0.99 <.001 .99 99, .998 .015 8
Log(NTMS) 2.32 2.04, 2.65 <.001 1.95 1.65, 2.367 <.001 16
N of dominant Folicles 1.46 1.42, 1.51 <.001 1.29 1.17, 1.356 <.001 24
Infertility Duration 0.94 0.92, 0.97 <.001 0.90 0.88, 0.940 <.001 11
N of previous attempts 0.85 0.82, 0.89 <.001 0.84 0.79, 0.893 <.001 13
Basal FSH >10 IU/L 0.74 0.61, 0.89 0.002 0.28 0.19, 0.367 <.001 23
Infertility cause 0.52 0.43, 0.63 <.001 0.36 0.26, 0.427 <.001 27
Previous failed IVF cycle 0.45 0.23, 0.89 0.023 0.09 0.01, 0.707 <.001 -
Previous abortions 0.73 0.63, 0.84 <.001 0.69 0.56, 0.864 <.001 -
N of alive children 1.38 1.31, 1.47 <.001 1.20 1.09, 1.326 <.001 -

First, fitting the new data provided results (left part of the table). Compared with the historical meta-model, we confirm strong prediction of Age, NTMS, Number of dominant Follicles, infertility duration, number of previous attempts and basal FSH>10 IU/L.
In addition we provide new significant predictors: a non linear effect of age, infertility causes (tubal factor, endometriosis and uterus abnormalities), previous live birth (OR≅1.4), previous abortion and past failure to IVF cycle strongly influence IUI success. The resulting model shows a highly significant improvement compared with other models (area under ROC curve Θ =76) and perfectly calibrated.
CONCLUSIONS: Our model based on prior evidence summarizing historical results provides confirmation of the historical meta-model. In addition, a significant improvement of the precision of the estimate is found with a correction of I2 heterogeneity scores, and new predictors are proposed. New forthcoming researches based on new data might revisit this model in re-discussing these findings, and continuously improving the meta-model.

show less

Share this PosterTalk

About PosterTalks

PosterTalks allows meeting attendees the ability to view these presentations, download or bookmark their favorite presentations, download PDF versions of the posters, ask questions, leave comments, and share presentations with their colleagues – all from the convenience of a smart phone.

Contact Us

Have a question? Click here to contact us. Need technical support? Click here to email support.

© 2017 PosterTalks and Connect BioMed. All other content and data, including data entered into this website are copyrighted by their respective owners.