A case study of the applicability of a prediction model for the selection of patients undergoing in vitro fertilization for single embryo transfer in another center

Erasmus Universiteit Rotterdam, Rotterdam, South Holland, Netherlands
Fertility and sterility (Impact Factor: 4.59). 07/2007; 87(6):1314-21. DOI: 10.1016/j.fertnstert.2006.11.052
Source: PubMed


To evaluate the application in a different fertility clinic of a prediction model for selecting IVF patients for elective single embryo transfer.
Retrospective analysis of a large database obtained from a tertiary infertility center.
University medical center.
The model, derived at the "development center" was applied in 494 consecutive first IVF cycles carried out at the "application center."
After adjustment of embryo scoring system to be compatible with that used by the prediction model, it was applied to the development center data. A score chart for predicting the probability of singleton or twin pregnancy was constructed.
The area under the receiver operator curve (ROC) was determined to measure the ability of the model to discriminate between ongoing pregnancy and twin pregnancy. Calibration plots were made to assess agreement between predicted and observed pregnancy rates (PR).
The areas under the ROC for predicting ongoing pregnancy and twin pregnancy were 0.63 and 0.66, respectively. Insertion of a correction factor equivalent to the difference in odds ratios for ongoing PR between the two centers was required to improve the calibration of the model.
After adaptation, the model performed well in the application center.

6 Reads
  • Source
    • "In order to facilitate clinical decisionmaking , numerous investigators have attempted to identify a specific factor or a combination of factors that are associated with IVF outcomes . Clinical parameters that positively correlated with IVF success include the total number of oocytes retrieved (Bouckaert et al., 1994; Hunault et al., 2002; Sharma et al., 2002; Hunault et al., 2008), endometrial thickness (McWilliams and Frattarelli, 2007; Richter et al., 2007; Traub et al., 2009 ), number of embryos transferred (Sharma et al., 2002; Pandian et al., 2009) and good quality embryos available for transfer (Minaretzis et al., 1998; Burke et al., 2000; Strandell et al., 2000; Terriou et al., 2001; Sharma et al., 2002; Thomas et al., 2010). In contrast, factors associated with a reduced likelihood of IVF success include advanced maternal age (Creus et al., 2000; Chuang et al., 2003; Ottosen et al., 2007; Hunault et al., 2008; Nelson and Lawlor, 2011), elevated basal FSH levels (Creus et al., 2000; Ottosen et al., 2007; van Loendersloot et al., 2010), duration of infertility (Templeton et al., 1996; Lintsen et al., 2005; Nelson and Lawlor, 2011), low antral follicle number (Hendriks et al., 2005; Khairy et al., 2008), premature luteinization (Bosch et al., 2003; Ou et al., 2008) and low serum E2 levels during controlled ovarian stimulation (Phelps et al., 1999; Khalaf et al., 2000). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Although close observation of serum estradiol (E2) levels remains a mainstay of assessing clinical response to controlled ovarian stimulation, the prognostic value of any change in E2 levels after administration of hCG remains unclear. The objective of this study is to evaluate the relationship between serum E2 response after hCG administration and the clinical pregnancy and live birth rates in fresh IVF cycles. We conducted a retrospective cohort study of women aged 21-45 years undergoing their first IVF cycle from 1999 to 2008 at a single practice. We compared the post-hCG serum E2 level with values on the day of hCG trigger. IVF cycles were stratified by post-hCG E2 response and appropriate parametric and non-parametric statistics were performed. Clinical intrauterine pregnancy and live births were the primary outcomes of interest. Multivariable logistic regression models were created to identify predictive factors associated with outcomes while adjusting for potential confounders. Among the 1712 IVF cycles, 1065 exhibited a >10% increase (Group A), 525 had a plateau (± 10%, Group B) and 122 showed a >10% decrease (Group C) in post-hCG E2 levels. While the E2 levels on the day of hCG were similar across groups, Group C had more patients with diminished ovarian reserve, required higher gonadotrophin doses and had the lowest implantation rates. After adjusting for age, total gonadotrophin dose, infertility diagnosis, number of oocytes and number of transferred embryos, the associations between post-hCG E2 decline (Group C) and clinical pregnancy [adjusted odds ratio (aOR): 0.53; 95% confidence interval (CI): 0.33-0.84, P= 0.007] and live birth (aOR: 0.40; 95% CI: 0.22-0.71, P= 0.002) were significant. We also found significant associations between E2 plateau (Group B) and clinical pregnancy (aOR: 0.73; 95% CI: 0.57-0.94, P= 0.013) and live birth (aOR: 0.74; 95% CI: 0.56-0.97, P= 0.032) when adjusting for the same factors. In our study, >10% decrease in E2 levels after hCG administration was associated with 40-50% reduction in clinical pregnancy and live birth rates. Similarly, post-hCG E2 plateau (± 10%) lowered the clinical pregnancy and live birth rates by >25%. Our study suggests that the change in the post-hCG E2 level is another parameter that can be used by clinicians to counsel patients regarding their likelihood of success with assisted reproductive technologies prior to oocyte retrieval.
    Preview · Article · Jun 2012 · Human Reproduction
  • Source
    • "Further, the variation without adjustment was only slightly larger than the variation with adjustment, and a modest 17% of the differences between centres can be explained by patient mix. Several validation studies have concluded that differences in pregnancy outcome rate between clinics have important limitations for the reproducibility of prediction models (Stolwijk et al., 1998; Smeenk et al., 2000; Hunault et al., 2007). By chance alone, pregnancy rates may vary from year to year and natural variation causes fluctuations in results (Kremer et al., 2008). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Pregnancy rates cannot be used reliably for comparison of IVF clinic performance because of differences in patients between clinics. We investigate if differences in pregnancy chance between IVF centres remain after adjustment for patient mix. We prospectively collected IVF and ICSI treatment data from 11 out of 13 IVF centres in the Netherlands, between 2002 and 2004. Adjustment for sampling variation was made using a random effects model. A prognostic index for subfertility-related factors was used to adjust for differences in patient mix. The remaining variability between centres was split into random variation and true differences. The crude 1-year ongoing pregnancy chance per centre differed by nearly a factor 3 between centres, with hazard ratios (HRs) of 0.48 (95% CI: 0.34-0.69) to 1.34 (95% CI: 1.18-1.51) compared with the mean 1-year ongoing pregnancy chance of all centres. After accounting for sampling variation, the difference shrank since HRs became 0.66 (95% CI: 0.51-0.85) to 1.28 (95% CI: 1.13-1.44). After adjustment for patient mix, the difference narrowed somewhat further to HRs of 0.74 (95% CI: 0.57-0.94) to 1.33 (95% CI: 1.20-1.48) and 17% of the variation between centres could be explained by patient mix. The 1-year cumulative ongoing pregnancy rate in the two most extreme centres was 36% and 55%. Only a minor part of the differences in pregnancy chance between IVF centres is explained by patient mix. Further research is needed to elucidate the causes of the remaining differences.
    Preview · Article · Oct 2009 · Human Reproduction
  • Source
    • "An improved prognostic model to enable professionals to select appropriate couples for eSET would be helpful and may minimize the lower success rate reported with eSET in an unselected population (van Montfoort et al., 2006). New prediction models have recently been published (Hunault et al., 2007; van der Steeg et al., 2007), but none of them are as yet translated to daily practice. The same situation applies to embryo selection. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Elective single embryo transfer (eSET) enables the prevention of multiple pregnancies after in vitro fertilization (IVF). However, in Europe, the multiple pregnancy rate after IVF remains stable at approximately 23%, with SET occurring in 15% of all IVF cycles. In most European clinics, the decision for the number of embryos transferred is established through a form of shared decision-making between patients and professionals. The aim of this study is to explore factors influencing this decision, in particular factors preventing eSET use. We performed explorative, semi-structured, in-depth interviews, based on two theoretical models. The interviews were performed among 19 Dutch IVF professionals and 20 patients who had just undergone IVF or were on the waiting list for IVF. The interviews were fully transcribed and two researchers independently scored the factors according to the models. We identified a wide variety of factors, potentially influencing eSET use: 37 with the professionals and 26 among the patients. Examples of factors mentioned by both patients and professionals were: uncertainty about the eSET technique, couples' lack of knowledge about essential eSET aspects, absence of a reimbursement system which favours eSET, inadequate options to select couples suitable for eSET and inferior cryopreservation success rates. This study demonstrates that both IVF professionals and patients identify numerous factors preventing eSET use in clinical practice. To estimate the impact of these factors identified, a quantitative confirmation and assessment of the magnitude of the effect is necessary.
    Full-text · Article · Jul 2008 · Human Reproduction
Show more