1. Prognostic models can help predict the chances of conceiving naturally, conceiving with interventions like IUI or IVF, and the chances of live birth from IVF. 2. Key factors in these models include female age, duration of infertility, cause of infertility, ovarian reserve tests, and data from previous treatment cycles. 3. Existing prediction models have been shown to have good discrimination and calibration for outcomes like IUI or IVF success, though individual patient preferences must also be considered in treatment decisions.

1. Prognostic models in Infertility

2. Basic fertility work up referral gyn History Physical examination Cycle evaluation Ovulation Semen analysis ? PCT Tubal p atency : CAT HSG DLS FSH, E2 AFC

3. Causes of infertility Azoospermia Anovulation Double sided tubal occlusion Sexual dysfunction

4. Causes of subfertility Unexplained subfertility One-sided tubal pathology Cervical factor subfertility Endometriosis Decreased semen quality Decreased intercourse frequency

5. Evers JL, Lancet 2002 Infertility or subfertility?

6. Clinical problem Distinction between couples who need treatment and couples who are likely to conceive spontaneously

7. Clinical Problem II You scheduled a couple to do ICSI and the woman asked you : What is my chance to get a baby after doing ICSI???

8. Gynaecologists differ widely in estimating pregnancy chances of subfertile couples Van der Steeg et al.,HR, 2006

9. Why Models!! Prediction models are intended to help gynaecologists in patient communication and decision making about treatment

10. How to Choose: Expectant management or intervention Prediction models for Chance to concieve naturally (home conception) (treatment independent pregnancy) Prediction models for pregnancy after IVF Prediction models for pregnancy after IUI

11. Hunault et al. HR 2004 Prediction models for spontaneous pregnancy Eimers Collins Snick Hunault Female age + + - + Duration subfertility + + + + F.A. man Urethritis vg. man + - - - - - - - prim/ sec subfertility + + - + Anovulation - - + - Tubal pathology - + + - Semen-analysis + + - + Endometriosis - + - - PCT Referral status + - + +/- +

12. Calculation Prognosis P = 1-0,0166 EXP(-0,053* age -0,152* duration -0,447* prim/sec +0.0035* prog.mot -0,949* PCT -0,321* referral )

13. External validation the agreement between predicted probabilities and the outcome event rates Calibration

14. Calibration Synthesis model 10 groups of N~260 Van der Steeg HR 2007

15. http:// www.amc.nl/prognosticmodel http:// www.amc.nl/prognosticmodel

16. Clinical consequences Couples with prognosis <30% = IVF Couples with prognosis > 40% = expectant management Couples with prognosis 30-40% = IUI

17. Expectant management or intervention Prediction models for Chance to concieve naturally (home conception) (treatment independent pregnancy) Prediction models for pregnancy after IVF Prediction models for pregnancy after IUI

18. Protocols for IVF GnRH Antagonist Protocols GnRH Agonist Protocols 225 IU per day (150 IU Europe) Individualized Dosing of FSH/HMG 250  g per day antagonist Individualized Dosing of FSH/HMG GnRHa 1.0 mg per day up to 21 days 0.5 mg per day of GnRHa 225 IU per day (150 IU Europe) Day 6 of FSH/HMG Day of hCG Day 1 of FSH/HMG Day 6 of FSH/HMG Day of hCG 7 – 8 days after estimated ovulation Down regulation Day 2 or 3 of menses Day 1 FSH/HMG

19. Which day!!! Day of start of cycle Day of start of stimulation Day of OPU Day of ET the time of embryo transfer will be more accurate but limited since the couple has already gone through the whole process of IVF.

20. Ideal model the probability of live birth in an IVF cycle prior to start of ovarian stimulation.

21. Day of start: Baseline factors female age, duration of infertility, primary cause of infertility, duration of GnRH agonist use, Hormonal level the number of previous IVF cycle

22. The age of the woman is still considered to be the most important predictor of IVF success (Broekmans and Klinkert, 2004).

23. increasing duration of infertility has also been shown to be negative impact , even after adjustment for age, whereas previous pregnancy increases the likelihood of success (Collins et al., 1995; Templeton et al, 1996).

24. couples with different infertility diagnoses will likely have different probabilities of achieving a live birth

25. Ovarian reserve tests Basal FSH, inhibin B, and anti-Müllerian hormone concentrations, as well as antral follicles count can be used to measure the ovarian reserve ( Broekmans et al., 2006; Kwee et al., 2008).

26. AMH If kits are available, AMH measurement could be the most useful in the prediction of ovarian response in anovulatory women. It is done at any day of cycle It is too expensive Exact normal levels not yet well agreed upon

27. ? Pregnancy correlation with the degree of response to COH, but identifying poor responders by means of these tests has low prognostic value in relation to the chance of live birth after IVF Broekmans et al. (2006)

28. How to build a model! Multivariate logistic regression analysis for previous prognostic variables to create prediction models of ovarian response and/or ongoing pregnancy has been used to a lesser extent (e.g., Bancsi et al., 2002).

29. Existing Models Most statistical models for prediction of IVF outcome use both prestimulation parameters and data obtained during the treatment, such as data on embryos

30. IVF prediction models Prediction models Outcome Discrimination Calibration Templeton (1996) IVF 0.63 good

32. Calculation The predicted probability ( P ) of achieving a live birth after IVF was calculated using the Templeton the model: Where y was defined as y = –2.028 + [0.00551x( age – 16)2] – [0.00028x(age – 16)3] + [i – (0.0690x no. of unsuccessful IVF attempts )] – (0.0711xtubal subfertility) + (0.7587xlive birth after IVF) + (0.2986 x previous pregnancy after IVF which did not result in a live birth) + (0.2277x live birth which was not a result of IVF) + (0.1117x previous pregnancy , not after IVF and which did not result in a live birth).

33. IVF prediction models Prediction models Outcome Discrimination Calibration Templeton (1996) IVF 0.63 good

34. Lintsen, A.M.E. et al. Hum. Reprod. 2007

35. classified for each woman into one of three groups, i.e., (i) predictor of good prognosis (ii) intermediate prognosis (iii) predictor of poor prognosis.

36. Expectant management or intervention Prediction models for Chance to concieve naturally (home conception) (treatment independent pregnancy) Prediction models for pregnancy after IUI Prediction models for pregnancy after IVF

37. Prognostic factors of pregnancy in intrauterine insemination Women with intermediate prognosis

38. IUI prediction model prediction models Outcome Discrimination Calibration Steures (2004) IUI 0.59 good

39. PICO Patient woman, 34 years, 2ys 1ry unexplained inf. Intervention IUI Comparison wait Outcome Pregnancy

42. -- delayed treatment -- early treatment RR: 1, 0 (CI: 0,86-1,2) N= 90 (71%) N= 90 (71%)

43. Take Home Message Prediction models are now available and ready for use Female age is the overwhelming factor affecting prediction models The prognosis should be discussed clearly with the patients based on scientific evidence and existing models.

44. However Patient preferences Private vs medical insurance Patient values

45. http:// www.amc.nl/prognosticmodel http:// www.amc.nl/prognosticmodel

46. Clinical consequences Couples with prognosis <30% = IVF Couples with prognosis > 40% = expectant management Couples with prognosis 30-40% = IUI

47. Lintsen, A.M.E. et al. Hum. Reprod. 2007

48. Basics Clinical Expertise Prediction Model Patient Preferences

49. THANK You