Milan Reljič, Ph.D., Veljko Vlaisavljević, Ph.D., Vida Gavrić, M.Sc., Borut Kovačič, Ph.D.,

FERTILITY AND STERILITY VOL. 75, NO. 3, MARCH 2001 Copyright 2001 American Society for Reproductive Medicine Published by Elsevier Science Inc. Printed on acid-free paper in U.S.A. Value of the serum estradiol level on the day of human chorionic gonadotropin injection and on the day after in predicting the outcome in natural in vitro fertilization/ intracytoplasmic sperm injection cycles Milan Reljič, Ph.D., Veljko Vlaisavljević, Ph.D., Vida Gavrić, M.Sc., Borut Kovačič, Ph.D., and Mojca Čižek-Sajko, B.Sc. Department of Reproductive Medicine and Gynecologic Endocrinology, Maribor Teaching Hospital, Maribor, Slovenia Received May 31, 2000; revised and accepted September 18, 2000. The study was sponsored by grant no. J3-8764 from the Ministry of Science and Technology of the Republic of Slovenia. Reprint requests: Milan Reljič, M.Sc., Department of Reproductive Medicine and Gynecologic Endocrinology, Maribor Teaching Hospital, Ljubljanska 5, SI-2000 Maribor, Slovenia (FAX: 386-2312-393; E-mail: milan.reljic@sb-mb.si). 0015-0282/01/$20.00 PII S0015-0282(00)01735-0 Objective: To predict the risk of premature ovulation and possibility of pregnancy in natural IVF/ICSI cycles on the basis of the estradiol (E 2 ) level on the day of hcg injection and on the day after. Design: A prospective study. Setting: Hospital research program. Patient(s): One hundred sixty-four women undergoing 305 IVF/ICSI natural cycles. Intervention(s): Serum E 2 levels were measured 12 h before (day 0) and 12 17 h after hcg administration (day 1). Main Outcome Measure(s): E 2 levels on day 0 and day 1, the ratio of the day 1 to day 0 levels, and cycle outcome. Result(s): In cycles with premature ovulation and in conception cycles, the average E 2 level on day 0 was statistically significantly higher than in other cycles, whereas the E 2 ratio was statistically significantly lower. Multiple logistic regression was used to determine the connection of the E 2 level on day 0 (B0 0.742, B 2.147, P.01) and the E 2 ratio (B0 0.742, B 3.135, P.001) with premature ovulation. Only the E 2 ratio (B0 0.659, B 2.209, P.0068) was significantly connected with conception. Conclusion(s): In predicting the outcome of natural IVF/ICSI cycles, the importance lies not in the E 2 level on the day of hcg administration or on the day after, but rather in the E 2 ratio. (Fertil Steril 2001; 75:000 00. by American Society for Reproductive Medicine.) (Fertil Steril 2001;75:539 43. 2001 by American Society for Reproductive Medicine.) Key Words: Natural IVF/ICSI cycles, serum E 2 levels, cycle outcome prediction The first successful pregnancy resulting from in vitro fertilization (IVF) occurred during an unstimulated normal menstrual cycle (1, 2). This purely spontaneous cycle was supplemented primarily with a midcycle dose of gonadotropins to trigger ovulation; thus, this type of cycle has been termed natural or unstimulated IVF (3). This relatively small but important addition to the spontaneous cycle ensures a predictable duration of luteinizing hormone (LH) like stimulation of the preovulatory follicle and allows follicle aspiration to be scheduled for a convenient time of day (4). Accurate timing of hcg administration requires careful cycle monitoring, as the success of the cycle depends on it. When the dominant follicle has reached suitable maturity, which is immediately before the spontaneous LH peak (5), hcg administration is on time. The most significant criterion for evaluating follicular maturity is the serum estradiol (E 2 ) level, so the relationship between the E 2 level on the day of hcg injection and the cycle outcome is not surprising. In reviewing the characteristics of successful and unsuccessful cycles in natural IVF programs, Paulson et al. noted that cycles resulting in pregnancy tended to have a higher E 2 level on the day of hcg 539

administration than cycles that failed (4). Other investigators could not confirm these results, finding a similar pregnancy rate in cycles with lower E 2 levels on the day of hcg injection (5 8). Our own preliminary results also showed that in predicting implantation, the relationship between the E 2 levels immediately before and after hcg injection is more important than the E 2 level on the day of hcg injection (9). The significance of the E 2 pattern after hcg administration in predicting follicular maturity and cycle outcome was also pointed out by Paulson, who noted that E 2 levels on the day after hcg helped to confirm premature ovulation if the level decreased more than 20% from the previous day (3). The aim of our study was to establish whether we might predict the risk of premature ovulation and the possibility of pregnancy on the basis of E 2 levels on the day of hcg injection and the E 2 dynamics immediately before and after hcg administration and to assess which factor has the greater predictive value. MATERIALS AND METHODS All spontaneously ovulating women who were on the waiting list for stimulated IVF/ICSI cycles at our hospital were given the option of undergoing natural cycles. All patients who agreed to participate in the study and fulfilled the criteria for hcg administration were included in the study. Between August 1998 and December 1999, 164 women underwent 305 natural IVF/ICSI cycles (198 IVF and 107 ICSI). The age range of the women was 25 43 years (mean, 32.7; SD, 4.16). The indications for IVF/ICSI procedure were female factor infertility only (n 66), male factor infertility only (n 46), female and male factor infertility (n 18), and unexplained infertility (n 34). With female factor infertility, tubal infertility prevailed (n 68, 81%), whereas with male factor infertility, oligozoospermia prevailed (n 28, 43.7%). A pelvic ultrasound examination was performed on day 2 of the menstrual cycle to evaluate the ovaries for the presence of cysts, and on the same day serum E 2 was measured to confirm that the cycle has just begun. The cycle was canceled if baseline E 2 was 0.3 nmol/l. Cycles with more than one growing follicle 12 mm were not included in the study. The cycles were monitored as described in the literature (9). When the average follicular diameter was 15 mm and E 2 was 0.5 nmol/l, hcg (5,000 IE) was administered. Blood was drawn 12 h before hcg administration (day 0) and 12 17 h after hcg administration (day 1) to estimate E 2. Follicular puncture was performed in patients with negative LH before hcg administration 35 37 h after hcg injection. Serum E 2 concentrations were determined with the enzyme-immune technique (AXSYM; Abbott, Germany). Premature LH was assessed with the urine test (RAPITEST; Morwell Diagnostic, Zurich, Switzerland). The oocytes were cultivated and inseminated, and the embryos were transferred as described elsewhere (10). The luteal phase was supported with hcg (1,500 IU) on the day of embryo transfer and 4 and 9 days later. Serological documentation of pregnancy was scheduled 16 days after embryo transfer. A pregnancy was confirmed only if there was ultrasonic evidence of a gestation sac. Using the t-test, E 2 levels on day 0 and day 1 and the ratio of the level on day 1 to day 0 (the E 2 ratio) were compared between cycles with and without premature ovulation (with oocyte recovery) and between nonconception (without premature ovulation) and conception cycles. The association between E 2 levels on day 0, day 1, the E 2 ratio, and cycle outcome (premature ovulation and pregnancy) was analyzed with univariate logistic regression. In the first case, the dependent variable was coded by marking cycles with premature ovulation as 1 and those without premature ovulation 0. In the second case, the dependent variable was coded by marking nonconception cycles (without premature ovulation) 0 and conception cycles 1. All independent variables were continuous variables and were not coded. Variables proved statistically significant by univariate analysis were analyzed with multiple logistic regression. If only one variable was connected with cycle outcome, the best cut-off limit was established with the receiver operator characteristic (ROC) curve, and on the basis of logistic regression the odds ratio was calculated for the chance of pregnancy in relation to this cutoff limit. The cycles were divided into two groups according to this cutoff limit, and the difference in premature ovulation, oocyte recovery, fertilization, implantation rate, and pregnancy rate between the two groups was evaluated by Pearson s 2 test. The data were processed with the Statistica program (StatSoft, Tulsa). Statistical significance was set at P.05. The study was approved by our Institutional Review Board. RESULTS Of 305 cycles, 28 (9.2%) were premature ovulation cycles. Among the remaining 277 cycles in which follicle puncture was done, clinical pregnancy was established in 35 (12.6%) cycles. In cycles with premature ovulation, the average E 2 level on day 0 was statistically significantly higher and the E 2 ratio was statistically significantly lower compared with average levels in cycles without premature ovulation. Average E 2 levels on day 1 did not differ between the two groups. The comparison of conception and nonconception cycles gave similar results (Table 1). Using univariate logistic regression we found an association between the serum E 2 levels on day 0 (B0 4.908, B 3.119, P.001) and the E 2 ratio (B0 1.872, B 3.895, P.001) with premature ovulation. Also, in the multiple logistic model both variables remained statistically significantly associated with ovulation, the E 2 ratio being 540 Reljič et al. E 2 pattern predicts natural IVF cycle outcome Vol. 75, No. 3, March 2001

TABLE 1 A comparison of serum E 2 levels on day 0 and day 1 and the ratio between cycle groups. E 2 level (nmol/l) Cycles with premature ovulation Cycles without premature ovulation P value Conception cycles Nonconception cycles P value Day 0 0.97 0.25 0.73 0.23.001 0.82 0.23 0.72 0.23 0.018 Day 1 0.87 0.27 0.90 0.31 NS 0.85 0.28 0.91 0.31 NS Day 1/day 0 0.91 0.27 1.27 0.37.001 1.06 0.35 1.30 0.36.001 Note: Values are means SD. NS not significant. associated more closely (for E 2 on day 0, B 2.147, P.011; for the E 2 ratio, B 3.135, P.001). The risk of premature ovulation can thus be calculated on the basis of the multiple logistic regression equation z exp( 0.742 (2.147) E 2 ( 3.135) E 2 ratio)/(1 exp( 0.742 (2.147) E 2 ( 3.135) E 2 ratio)). With univariate logistic regression we verify the connection of serum E 2 levels on day 0 (B0 3.224, B 1.615, P.0215) and the E 2 ratio (B0 0.659, B 2.209, P.0068) with conception. In the multiple logistic regression model, only the E 2 ratio remains statistically significantly connected, while E 2 on day 0 is no longer statistically significant. The association between the E 2 ratio and pregnancy is shown in Figure 1. The best cutoff value of this ratio calculated with the ROC curve was 1.1, with 68.6 sensitivity and 68.2 specificity (Fig. 2). The odds ratio for the chance of pregnancy in relation to this cutoff value was 3.77 (95% confidence interval, 1.69 8.02). If we divide the cycles into two groups according to a 1.1 cut-off ratio value, we find that the pregnancy rate per cycle was statistically significantly higher in cycles with E 2 ratio 1.1 because of the higher implantation rate, despite a higher premature ovulation rate (Table 2). DISCUSSION It is difficult to foresee the right time for hcg administration in natural cycles on the basis of E 2 levels since the E 2 levels at which the follicle reaches suitable maturity differ. This is also confirmed by findings in the normal menstrual cycle as E 2 levels at the onset of the spontaneous LH surge move between 0.451 nmol/l and 3.071 nmol/l (11). The E 2 levels at which various investigators recommend hcg administration are also different, varying between 0.4 nmol/l and 1.1 nmol/l (8, 12). Some of these differences may be the result of different tests for E 2 determination and the different time intervals between E 2 measurement and hcg adminis- FIGURE 1 Probability of pregnancy in relation to the E 2 ratio in cycles without premature ovulation. FIGURE 2 Receiver operator characteristic curve for determining pregnancy on the basis of the E 2 ratio. A cutoff 0.9, B cutoff 1, C cutoff 1.1, D cutoff 1.2, E cutoff 1.3. FERTILITY & STERILITY 541

TABLE 2 Outcome of natural IVF/ICSI cycles according to the ratio of 1.1 E 2 onday1toe 2 on day 0. Parameter 1.1 E 2 on day 1/E 2 on day 0 1.1 No. of cycles 182 123 305 Premature ovulation rate 3.3 (6/182) 17.9 (22/123) a 9.2 (28/305) Oocyte recovery rate 74.4 (131/176) 83.2 (84/101) 77.6 (215/277) Fertilization rate 70.2 (92/131) 70.2 (59/84) 70.2 (151/215) Implantation rate 12.8 (11/86) 44.4 (24/54) a 25.0 (35/140) Pregnancy rate per cycle 7.0 (11/182) 19.5 (24/123) a 11.3 (33/291) a P.001 vs E 2 on day 1/E 2 on day 0 1.1. tration. It is clear, however, that the serum E 2 level is not the optimum criterion for assessing the right time of hcg administration or follicular maturity, since regardless of the criteria taken into consideration in a certain percentage of cycles, the timing of hcg administration is not optimal. If we decide on hcg administration at lower E 2 levels, there is a greater probability of follicular immaturity, but there is also a smaller probability of the spontaneous LH peak taking us by surprise and being followed by premature ovulation and vice versa. It seems understandable, however, that with higher E 2 levels on the day of hcg injection there is a greater possibility of follicular maturity and therefore a greater possibility of conception. Our results confirm this; like Paulson et al. (4), we also found higher E 2 levels on the day of hcg injection in conception cycles as compared with nonconception cycles. Waiting for higher E 2 levels would not necessarily increase the pregnancy rate per cycle, as this would increase the percentage of premature LH surges and premature ovulation. Our own study shows this, because the E 2 levels on day 0 were higher in cycles with premature ovulation and the risk of premature ovulation increased with the E 2 level. In our opinion, the dynamics of the E 2 level immediately before and after hcg administration has an even greater predictive value for cycle outcome. Our results show that in predicting ovulation the most significant factor is the E 2 ratio. This is in accordance with the findings of Paulson et al. (3), who found that E 2 levels on the day after hcg help to confirm premature ovulation if the level decreases more than 20% from the previous day. In predicting premature ovulation, we did not determine the best cutoff limit of the E 2 ratio but rather expressed the risk by the logistic regression equation, since besides the E 2 ratio, the E 2 level on day 0 is also a significant predictor of premature ovulation. In cycles without premature ovulation, E 2 level on day 0 was significantly associated with conception only for univariate analysis. When the E 2 ratio was also included in the T model, it became insignificant and only the E 2 ratio remained associated with cycle outcome. This points to a correlation between these two factors and shows that the E 2 ratio has a greater value in predicting conception. Similarly, as the risk of ovulation increased with the decrease in E 2 ratio, the possibility of conception also increased if follicle puncture occurred. These results show that there is a possibility that the E 2 dynamics on day 0 and day 1 indicate follicular maturity. The answer may be found in the physiology of the menstrual cycle. In a normal menstrual cycle it is assumed that the endogenous LH release is synchronized with optimal follicular and oocyte maturity, so serum E 2 levels decrease after the initial rise in LH level because of a decrease in androgen production by theca cells and a similar fall in the aromatase activity of granulosa cells (13). We assume that follicles of different maturity respond differently to hcg administration. With follicular maturation, namely, the number of hcg receptors on granulosa cells increases, which may affect the rate of luteinization of these cells and E 2 production (14). In conception cycles and cycles with premature ovulation we suppose that hcg administration was performed on more mature follicles, and because of the higher hcg receptor content after hcg administration, luteinization was faster and more expressed, leading to a drop or merely minimum increase in E 2. In cycles without premature ovulation and in nonconception cycles, luteinization would be less expressed because of lesser follicle maturity, so E 2 synthesis would not materially decrease and serum E 2 levels would increase significantly even after hcg administration. However, these are mere speculations, because our results do not make it possible to determine the precise dynamics because we did not assess E 2 at hcg administration but rather 12 h before and 12 17 h after. Thus, we cannot know whether, e.g., the E 2 decrease on day 1 is the result of the smaller E 2 increase 12 h before hcg injection or of the more rapid decrease 12 17 h after. It is remarkable that in dividing cycles on the basis of the best cutoff limit of the E 2 ratio, we established a difference only in ovulation and implantation rate but not in fertilization or oocyte recovery success. Although other investigators also did not find a connection between E 2 pattern and fertilization in stimulated cycles (15), we admit the possibility that our results were influenced by the choice of indications for an IVF/ICSI cycle. Our study also included couples treated for male factor infertility, including ICSI cycles. Although the recovery rate was higher with a lower E 2 ratio, this difference was not statistically significant. So it is possible that Daya et al. are correct in stating that the success of follicle puncture depends more on the choice of the needle and the technique than on follicular maturity (6). In our opinion, the better implantation in cycles with a lower E 2 ratio is the result of greater follicle and oocyte maturity because we assume the embryo originating from a 542 Reljič et al. E 2 pattern predicts natural IVF cycle outcome Vol. 75, No. 3, March 2001

suitably mature oocyte has a greater chance of implantation and further development. Daya et al. also thought that follicle maturity had the greatest effect precisely on implantation, and they stated that oocytes from nondominant follicles are capable of being fertilized and undergoing cleavage, but are probably less likely to implant (6). Our results show that it is possible to predict the success of natural IVF/ICSI cycle on the basis of E 2 dynamics immediately before and after hcg administration. However, the clinical relevance of these observations is still not clear, and before clinical application, these results should be confirmed in a prospective study. Acknowledgment: The authors thank Marijana Gajšek-Marchetti, a translator from the Medical Research Department, for her contribution in preparing the manuscript. References 1. Edwards RG, Steptoe PC, Purdy JM. Establishing full-term human pregnancies using cleaving embryos grown in vitro. Br J Obstet Gynecol 1980;87:737 56. 2. Steptoe PC, Edwards RG. Birth after the reimplantation of a human embryo. Lancet 1978:366. 3. Paulson RJ. Natural cycle in vitro fertilization. Infertil Reprod Med Clin North Am1993;4:653 65. 4. Paulson RJ, Sauer MV, Francis MM, Macaso TM, Lobo RA. Factor affecting pregnancy success of human in-vitro fertilization in unstimulated cycles. Hum Reprod 1994;9:1571 5. 5. Foulot H, Ranoux C, Dubuisson JB, Rambaud D, Aubriot FX, Poirot C. In vitro fertilization without ovarian stimulation: a simplified protocol applied in 80 cycles. Fertil Steril 1989;52:617 21. 6. Daya S, Gunby J, Hughes EG, Collins JA, Sagle MA, Young Lai EV. Natural cycles for in vitro fertilization: cost-effectiveness analysis and factors influencing outcome. Hum Reprod 1995;10:1719 24. 7. Seibel MM, Kearnan M, Kiessling A. Parameters that predict success for natural cycle in vitro fertilization-embryo transfer. Fertil Steril 1996;63:1251 4. 8. Tomaževič T, Geršak K, Meden-Vrtovec H, Drobnič S, Veble A, ValenčIč B, et al. Clinical parameters predict the success of in vitro fertilization-embryo transfer in natural cycles. Assist Reprod 1999;9: 149 56. 9. Reljič M, Vlaisavljević V. The preovulatory serum estradiol pattern in natural IVF/ICSI cycles. J Assist Reprod Genet 1999;16:535 9. 10. Vlaisavljević V, Kovačič B, Gavrić V. In vitro fertilization program based on programmed cycles monitored by ultrasound only. Int J Gynecol Obstet 1992;39:227 31. 11. Cahill DJ, Wardle PG, Harlow CR, Hull MGR. Onset of the preovulatory luteinizing hormone surge: diurnal timing and critical follicular prerequisites. Fertil Steril 1998;70:56 9. 12. Paulson RJ, Sauer MV, Lobo RA. In vitro fertilization in unstimulated cycles: a new application. Fertil Steril 1989;51:1056 60. 13. Hiller SG, Reichert LE, Van Hall EV. Control of preovulatory follicular estrogen biosynthesis in the human ovary. J Clin Endocrinol Metab 1981;52:847. 14. Speroff L, Glass RH, Kase NG. Regulation of the menstrual cycle. In: Speroff L, Glass RH, Kase NG. Clinical gynecologic endocrinology and infertility. Baltimore: Williams & Wilkins, 1994:183 30. 15. Dor J, Rudak E, Mashiach S, Nebel L, Serr DM, Goldman B. Periovulatory 17 -estradiol changes and embryo morphologic features in conception and nonconception cycles after human in vitro fertilization. Fertil Steril 1986;45:63 8. FERTILITY & STERILITY 543