Serum anti-müllerian hormone and estradiol levels as predictors of ovarian hyperstimulation syndrome in assisted reproduction technology cycles

Transcription

1 Human Reproduction Vol.23, No.1 pp , 2008 Advance Access publication on November 13, 2007 doi: /humrep/dem254 Serum anti-müllerian hormone and estradiol levels as predictors of ovarian hyperstimulation syndrome in assisted reproduction technology cycles Tsung-Hsien Lee 1,2,3, Chung-Hsien Liu 3, Chuin-Chia Huang 4,5, Yi-Ling Wu 3,4, Yang-Tse Shih 3, Hong-Nerng Ho 1, Yu-Shih Yang 1 and Maw-Shang Lee 3,4,6,7 1 Department of Obstetrics and Gynecology, College of Medicine, National Taiwan University, Taipei, Taiwan, Republic of China; 2 Graduate Institute of Clinical Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan, Republic of China; 3 Department of Obstetrics and Gynecology, Chung Shan Medical University Hospital, Taichung, Taiwan, Republic of China; 4 Division of Infertility Clinic, Lee Women s Hospital, 263, Pei-Tun Road, Taichung, Taiwan, Republic of China; 5 Department of Biotechnology, Central Taiwan University of Science and Technology, Taichung, Taiwan, Republic of China; 6 College of Medicine, Chinese Medical University, Taichung, Taiwan, Republic of China 7 Correspondence address. Fax: þ ; msleephd@giga.net.tw BACKGROUND: Anti-Müllerian hormone (AMH) is reported to be a reliable marker of the ovarian response to controlled ovarian stimulation (COS). The objective of this study is to determine whether the serum AMH level can predict ovarian hyperstimulation syndrome (OHSS) prior to selection of COS protocols. METHODS: A cohort of 262 IVF cycles was investigated prospectively, in order to evaluate the predictive value for OHSS by means of certain risk factors, including age, body mass index (BMI), serum estradiol (E 2 ) level, number of retrieved oocytes and basal serum AMH level. RESULTS: The basal serum AMH level predicted OHSS better than age and BMI with a sensitivity of 90.5% and specificity of 81.3%. Both the basal serum AMH level (odds ratio: , P ) and serum E 2 level on the day of HCG administration (odds ratio: , P ) proved to be significant predictors of OHSS by logistic regression analysis. However, age (odds ratio: , P ) was the only significant factor for prediction of clinical pregnancy. CONCLUSIONS: The basal serum AMH level could be utilized effectively to predict OHSS and thus to direct the selection of mild COS protocols. Keywords: anti-müllerian hormone; ovarian hyperstimulation syndrome; controlled ovarian stimulation; estradiol Introduction Ovarian hyperstimulation syndrome (OHSS) represents one of the most serious complications subsequent to controlled ovarian stimulation (COS) and can be life threatening. The characteristics of patients prone to develop OHSS have been previously reviewed by Navot et al. (1992). The specific risk factors for OHSS include young age (,35 years), lean habitus, signs of polycystic ovary syndrome (PCOS), the presence of multiple (.35) small and intermediate follicles, and excessively high levels of serum estradiol (E pg/ml) on the day of HCG administration (Navot et al., 1992). These features have been confirmed by other reports (Delvigne et al., 1993; Schenker, 1993) and have formed a basis upon which clinicians have been able to predict and prevent the occurrence of OHSS. Accurate prediction of OHSS in an individual IVF treatment cycle, however, remains a difficult task (Mathur et al., 1996). Furthermore, the predictive value of the serum E 2 level and the number of retrieved oocytes for the occurrence of OHSS has been questioned (Morris et al., 1995; Chen et al., 1997). Although monitoring the serum E 2 level has been effective in reducing the incidence of OHSS (Varma and Patel, 1988; Golan et al., 1989), the relevance of the use of E 2 levels during COS to predict the occurrence of OHSS has been challenged (Aboulghar, 2003; Orvieto, 2003). In addition, both the E 2 level and number of follicles are determined near the completion of COS. It is not particularly easy to accurately predict OHSS prior to COS for IVF cycles, using only age and body mass index (BMI). If a reliable marker could be identified in an individual s prior menstrual cycles (basal serum levels), mild, patient-friendly protocols for COS that are less aggressive than the present standard long protocol for a GnRH agonist could be utilized to prevent the detrimental effects of OHSS. Serum levels of anti-müllerian hormone (AMH) have been reported recently to be closely related to the ovarian response or ovarian reserve during IVF cycles (Seifer et al., 2002; van Rooij et al., 2002; Penarrubia et al., 2005; Ficicioglu et al., 160 # The Author Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 Hormone prediction of ovarian hyperstimulation syndrome 2006). The serum AMH level would appear to better reflect the level of ovarian aging than other known markers of ovarian reserve, such as basal serum FSH level, inhibin B level and antral follicle count (Fanchin et al., 2003b; van Rooij et al., 2005). Interestingly, OHSS has been reported to be associated with high serum levels of AMH prior to COS (Nakhuda et al., 2006). La Marca et al. (2007) reported that all patients with cancelled IVF cycles due to poor response to COS were in the group with the lowest serum AMH level in their study, whereas those with cancelled cycles due to a high risk of OHSS had serum AMH levels in the highest quartile. Based on this evidence, AMH may be a useful marker to predict OHSS for IVF cycles in addition to being an appropriate marker of ovarian reserve. The purpose of this study was to investigate the reliability of the basal serum AMH level in predicting OHSS for a prospective cohort of infertile couples undergoing COS. The occurrence of OHSS has been reported to be closely related to the likelihood of pregnancy (Papanikolaou et al., 2005); therefore, the ability of serum AMH levels and other risk factors for OHSS to predict pregnancy was also evaluated by a logistic regression model in this study. Materials and Methods Patient selection and stimulation protocols A total of 354 COS cycles performed at the Lee Women s Hospital, Taichung, Taiwan from July to December 2006 inclusively were enrolled into this study. The patients stimulated with the following protocols were excluded from analysis: an ultra-long or ultra-short GnRH agonist protocol (n ¼ 2), GnRH antagonist protocols (n ¼ 11), protocols for the combined use of clomiphene and gonadotropins (n ¼ 53) and protocols for oocyte donation (n ¼ 14). Therefore, patients stimulated with long and short GnRH agonist protocols (n ¼ 274) were all included in this study. Only patients who underwent one stimulated cycle during the study period were recruited for analysis. As a result, a total of 262 IVF procedures were recruited and analyzed for this study. The study protocol was approved by the Institutional Review Board of Chung-Shan Medical University Hospital. The women participating in this study followed a long or short GnRH agonist protocol that began with daily s.c. injections of 0.1 mg leuprolide acetate (LA Lupron, Takeda Pharmaceutics, Stolberg, Germany) on Day 21 of the pre-stimulation cycle (long protocol) or on Day 3 of the stimulation cycle (short protocol). The GnRH agonist Lupron was continued until the day of HCG administration. Gonadotropin (Gonal-F; Serono, Bari, Italy), at 225 IU/day s.c., was administered for five days, after which the level was adjusted according to the associated ovarian response in order to stimulate follicular development. The resultant ovarian response was monitored by transvaginal ultrasound and serum E 2 levels. When two or more follicles reached a maximum diameter of 18 mm, 250 mg of hcg (Ovitrelle, Serono) was administered. Transvaginal oocyte retrieval was performed h after hcg injection. The luteal phase was supported with 1500 IU of hcg (Pregnyl; Organon, Oss, The Netherlands) given i.m. on Days 4, 7 and 10 following oocyte retrieval, together with a daily oral 400 mg dose of micronized progesterone (Utrogestan; Besins International, Brussels, Belgium) given for the first 14 days after embryo transfer. When the serum level of E 2 rose to.3000 pg/ml or the number of retrieved oocytes exceeded 15, hcg was withdrawn and only oral 800 mg of micronized progesterone (Utrogestan; Besins International) was used for the following 14 days. Serum hcg was checked 14 days after embryo transfer and patients with HCG.50 IU/l were considered pregnant. An ultrasound examination was preformed one week later and then again three weeks later in order to determine the number of intrauterine gestational sacs present and fetal viability, respectively. Classification of OHSS The criteria for classification of OHSS as defined by Navot et al. (1992) were utilized to assess the relative severity of OHSS. Moderate OHSS was characterized by abdominal distension and discomfort, nausea, evidence of ascites and an ovarian size of 8 12 cm on ultrasonography. Severe OHSS consisted of clinical ascites with or without pleural effusion, edema anasarca, oligouria, a serum creatinine of , a creatinine clearance of 50 ml/min, liver dysfunction, a hematocrit.45% and a white blood cell count of Only those patients with a moderate or severe OHSS were classified as having a positive case of OHSS. Monitoring ovarian response and hormone assay Baseline hormone profiles, including serum levels of E 2, FSH, LH and AMH were determined on Day 3 of the pre-stimulation cycle. The ovarian response for all participants was monitored using transvaginal ultrasound from the beginning of Day 7 of the stimulation cycle and then continued on a daily basis until the day of hcg administration. Blood sampling for E 2, LH and progesterone was performed on the day of hcg administration. Serum LH was measured using a specific immunometric assay kit (Immulite; Diagnostic Products Corporation, Los Angeles, CA, USA). The sensitivity and intra/inter-assay coefficients of variation (CVs) for LH assay were 0.1 m IU/ml and 6.5/7.1%, respectively. E 2 and progesterone were also measured by competitive immunoassay using the Immulite kit, with intra/inter-assay CVs of 6.3/6.4% and 6.3/5.8% for E 2 and progesterone, respectively. The minimal detection limit for E 2 was 15 pg/ml (55 pmol/l) and 0.2 ng/ml (0.6 nmol/l) for progesterone. Measurement of serum AMH levels was performed in duplicate using the AMH/MIS enzyme-linked immunosorbent assay kit (Diagnostic Systems Lab, Webster, Texas, USA). The minimal detection limit and intra/inter-assay CV for the AMH assay was ng/ml and,5/,8%, respectively. Statistical analysis The various biological parameters relevant to COS were presented as the mean, SD and range of the data. For rates of pregnancy, OHSS and implantation, the respective 95% confidence interval (CI) was also calculated. Receiver operating characteristic (ROC) curve analysis was used to estimate the predictive power of the measured variables. The relative ability to use age, BMI and serum levels of AMH to predict the occurrence of OHSS was compared by calculating the areas under the respective ROC curves (ROC AUC ) and their 95% CIs. MedCalc software was used to compare the areas under every pair of ROC curves (version 9.2; MedCalc, Broekstraat, Belgium). Logistic regression was performed to determine the independent effect of the associated risk factors on the ability to predict OHSS and pregnancy for individual variables. The patients were further divided into four groups according to the 25th, 50th and 75th percentiles of the AMH levels. One-way analysis of variance (ANOVA) and the post-hoc Bonferroni modification were utilized. All logistic regression analyses and one-way ANOVA were performed using the 161

3 Lee et al. Statistical Package for the Social Sciences software (version 14.0; SPSS Inc., Chicago, IL, USA). Results From July to December 2006, a total of 262 patients undergoing the above-described IVF treatment were included in this study. The basic participant characteristics, such as age, BMI and baseline Day 3 levels of serum AMH, FSH and LH are presented in Table I. The results of COS, including the serum E 2 level and number of follicles (10 mm in diameter) present on the day of hcg administration, number of retrieved oocytes and rates of clinical pregnancy, implantation and OHSS are also shown in Table I. Twenty-one (8.0%) patients developed moderate to severe OHSS. One of 36 patients (2.8%) undergoing the short GnRH agonist protocol and 20 of 226 patients (8.8%) undergoing the long GnRH agonist protocol developed OHSS. In addition, nine patients featured sonographic and hormonal characteristics of PCOS. Only one of the 21 patients (4.8%) with OHSS had PCOS. The overall clinical pregnancy rate for our study was 40.8% (107/262). The ROC AUC for the predicting factors for OHSS were compared and the results are presented in Table II. The ROC AUC for AMH featured the largest area under the curve among the listed Table I. Patient characteristics, parameters related to COS and IVF outcomes (n ¼ 262). risk factors for OHSS and was significantly larger than the corresponding values for age and BMI (P ¼ and P, 0.001, respectively, Table II). The ROC AUC for BMI characterized the smallest area under the curve among the listed six factors for predicting OHSS and it was significantly lower than the corresponding values for AMH, E 2, number of follicles and number of retrieved oocytes (P, 0.001, P ¼ 0.001, P ¼ and P ¼ 0.002, respectively, Table II). Fig. 1 shows the ROC curve for the three predicting factors for OHSS which can be determined prior to ovarian stimulation: patient age, BMI and basal serum AMH level. Fig. 2 illustrates the cut-off value (3.36 ng/ml) for basal serum AMH level determined by the corresponding ROC curve with the highest sensitivity and specificity for predicting OHSS. The cut-off values for the other listed risk factors for OHSS were selected based upon each ROC curve, and the corresponding sensitivity and specificity for predicting OHSS are depicted in Table II. The basal serum levels of AMH featured a higher specificity (81.3%, 95% CI ) for detecting OHSS than the other risk factors for OHSS near the completion of COS, namely, the E 2 levels and number of follicles on the day of hcg administration, and the number of retrieved oocytes. To check whether a combination of these predicting factors would increase the accuracy for the prediction of OHSS, a Variable Mean (SD) Range Age (years) 33.9 (5.0) Baseline serum FSH level (miu/ml) 7.82 (4.06) Baseline serum LH level (miu/ml) 3.78 (1.33) BMI (Kg/m 2 ) 21.5 (3.3) Basal serum AMH level (ng/ml) 2.44 (1.86) Serum E 2 level on day of HCG administration (pg/ml) 1613 (1138) Number of follicles (10 mm) on day of HCG 11.1 (6.9) 1 33 Number of retrieved oocytes 10.3 (7.0) 0 33 Number of transferred embryos 3.76 (1.9) 0 6 OHSS rate (%) a 21/262 (8.0%) % Implantation rate (%) a 213/987 (21.6%) % Pregnancy rate (%) a 107/262 (40.8%) % a The proportional data were presented as percentage and 95% CI. AMH, anti-müllerian hormone; BMI, body mass index; OHSS, ovarian hyperstimulation syndrome. Table II. The predictability of AMH and other risk factors for OHSS compared by means of ROC curve analysis. Variable Area under ROC curve (95% CI) Cut-off value Sensitivity (95% CI) Specificity (95% CI) Positive predictive value Negative predictive value Age (years) ( ) a ( ) 56.0 ( ) BMI (Kg/m 2 ) ( ) b,c,d,e ( ) 90.3 ( ) Basal serum AMH level ( ) a,b ( ) 81.3 ( ) (ng/ml) Serum E 2 level on day of ( ) c ( ) 64.6 ( ) hcg administration (pg/ml) Number of follicles ( ) d ( ) 63.3 ( ) (10 mm) on day of hcg administration Number of oocytes collected ( ) e ( ) 66.0 ( ) a P ¼ 0.002, b P, 0.001, c P ¼ 0.001, d P ¼ 0.002, e P ¼ by pairwise comparison of the area under the ROC curve for all variables. 162

4 Hormone prediction of ovarian hyperstimulation syndrome Figure 1: Comparison of predictive values for ovarian hyperstimulation syndrome using the ROC AUC The ROC AUC of the basal serum AMH level was larger than that of age (P ¼ 0.002) and BMI (P, 0.001). Figure 2: Selection of cut-off value for basal serum AMH to predict OHSS by ROC curve analysis The selected value was 3.36 ng/ml, with a sensitivity of 90.5% (95% CI ) and a specificity of 81.3% (95% CI ). multivariate logistic regression model was used for further analysis (Table III). Only serum levels of basal AMH and E 2 on the day of hcg administration revealed significant correlations with OHSS (odds ratio: , P ¼ and odds ratio: , P ¼ , respectively). Furthermore, the serum levels of E 2 revealed a borderline significance as a predictor for OHSS, whereas the serum AMH level appeared to be strongly correlated with OHSS. Although the ROC AUC for the overall multivariate logistic regression model (0.906, 95% CI ) featured a greater area than those for AMH and E 2, the difference between these factors was not statistically significant. The overall model revealed a sensitivity of (5/9) 55.6% and a specificity of (231/247) 93.5%, which was a lower sensitivity but a better specificity than the serum AMH level alone. The six risk factors for OHSS were analyzed with a logistic regression model in order to determine their independence in predicting pregnancy (Table IV). Only subject age revealed a significant relationship with pregnancy outcome (odds ratio: , P ¼ ) when all the risk factors for OHSS were inserted into the logistic regression model. The basal serum AMH level was not a factor in effectively predicting clinical pregnancy (odds ratio: , P ¼ ). The cut-off value for the AMH levels was found to be similar to the levels of the highest quartile (75%) for AMH. In addition, the AMH values were not normally distributed; therefore, the patients were further subdivided into four groups according to the 25th, 50th and 75th percentiles of the AMH levels. The COS related parameters and outcome are demonstrated in Table V. The amount of exogenous gonadotropin administered was inversely related to AMH level. Although patients in the highest quartile group of the AMH level were associated with high risk for OHSS, the gonadotropin dose was lower than in the lowest quartile (P, 0.01). It was revealed that the patients in the third quartile (50 75%) of the AMH level had the highest pregnancy rate (36/66, 54.5%) among the four groups and a significantly higher rate than patients in the lowest quartile (P, 0.01). Discussion The incidence of OHSS (8.0%, 95% CI ) in this study was similar to that in studies reported by Chen et al. (1997, 2000, respectively, 5.5 and 7.7%). However, it was higher than the incidence of 3.3% reported by Navot et al. (1992) and 1.71% reported by Papanikolaou et al. (2005). Luteal support of IVF cycles with hcg, as was done in our study and the studies by Chen et al., may account for the higher incidence of OHSS (Chen et al., 1997; Papanikolaou et al., 2006). Furthermore, discontinuing hcg for luteal support of IVF cycles in patients at high risk for OHSS has been reported to reduce the prevalence of OHSS (Chen et al., 1997; Papanikolaou et al., 2006). Unfortunately, even though luteal hcg support was withdrawn in patients with a high risk of OHSS (serum E 2 level 3000 pg/ ml and retrieved oocytes.15), the incidence of OHSS for this study remained fairly high. This fact further emphasizes the difficulty of predicting and preventing OHSS using either serum E 2 levels or number of retrieved oocytes alone (Morris et al., 1995; Chen et al., 1997). The traditional determinants for OHSS prior to gonadotropin stimulation appear to include only subject age, lean habitus and signs of PCOS (both hormonal and ultrasonographic characteristics; Navot et al. 1992). In this study, nine patients had signs of PCOS, of which only one developed OHSS. This suggests that the predictive value of signs of PCOS for OHSS was somewhat limited in this study (the sensitivity for PCOS to predict OHSS was only 1/21, 4.8%). In addition, the baseline ratio of LH to FSH level (hormonal characteristic of PCOS) did have a smaller area under the corresponding curve than that for the basal serum AMH level (P ¼ 0.005, data not shown). 163

5 Lee et al. Table III. The correlation between the risk factors for OHSS and its occurrence as analyzed by a multivariate logistic regression model (n ¼ 262; cases with OHSS 21). Variable Coefficient Odds ratio (95% confidence interval) P-value Age (years) BMI (Kg/m 2 ) Basal serum AMH level (ng/ml) Serum E 2 level on day of hcg administration (pg/ml) Number of follicles on day of hcg administraton Number of oocytes collected Table IV. The correlation between the risk factors for OHSS and clinical pregnancy as analyzed by a multivariate logistic regression model (n ¼ 262; cases of clinical pregnancy 107). Variable Coefficient Odds ratio (95% CI) P-value Age (years) BMI (Kg/m 2 ) Basal serum AMH level (ng/ml) Serum E 2 level on day of hcg administration (pg/ml) Number of follicles on day of hcg administraton Number of oocytes collected Table V. The COS related parameters and clinical outcomes for the studied population divided according to the 25th, 50th and 75th percentile of serum AMH levels. Variable 0 25% (n ¼ 66) 25 50% (n ¼ 65) 50 75% (n ¼ 66) % (n ¼ 65) Basal serum AMH level (ng/ml) Age (years) a BMI (Kg/m 2 ) Basal serum FSH level (IU/l) a Basal serum LH level (IU/l) a Amount of recombinant FSH (ampoules) a Serum E 2 level on day of hcg administration (pg/ml) b a Number of follicles on day of hcg administraton b a Number of oocytes collected b a Rate of OHSS (%) c (3.0) 19 (29.2) Rate of poor response (follicle number,3; %) c 12(18.2) 3(4.6%) 0 0 Rate of clinical pregnancy rate (%) c 13(19.7) 27(40.9) 36(54.5) 31(47.7) a Statistically significant (P, 0.01) after one-way ANOVA analysis with post-hoc Bonferroni modification and the value is greater than the other three groups; b Statistically significant (P, 0.01) after one-way ANOVA analysis with post-hoc Bonferroni modification and the value is lower than the other three groups; c P, by chi-squared test. Furthermore, for women with PCOS, the serum AMH level appeared to be strongly elevated (Cook et al., 2002) and also appeared to be associated significantly with ultrasonographic features of PCOS, such as the presence of multiple small follicles in the ovaries. This evidence suggests that the basal serum AMH level might also correlate well with ultrasonographic signs of PCOS. The basal serum AMH level appeared to be a more efficient predictor of OHSS than subject age and BMI in this study. These results indicate that, of those parameters investigated, basal serum AMH level is the factor most likely to reveal the possibility of OHSS prior to COS, and worked as well as, if not better than, the number of follicles and serum levels of E 2 (Table II and Fig. 1). According to this finding, the serum AMH level prior to gonadotropin administration could provide useful information to direct the application of mild, patientfriendly COS protocols to avoid moderate and severe OHSS. The cut-off point, or threshold, for serum levels of AMH in predicting OHSS was not easy to establish, as found by others in attempts to use serum E 2 level as a predictor for OHSS (Aboulghar, 2003; Orvieto, 2003). The range of serum AMH levels ( ng/ml) in our study was similar to that in one previous report ( ng/ml, n ¼ 80) (Smeenk et al., 2007), but slightly different from another study (0 6.2 ng/ ml, n ¼ 119) (van Rooij et al., 2002). In one study, the serum level of AMH was higher in OHSS patients (n ¼ 16; 3.62 ng/ml with a SE of 0.87 ng/ml) than in controls (Nakhuda et al., 2006). In our study, the mean serum AMH level for OHSS patients (n ¼ 21) was 5.02 ng/ml (SE ¼ 0.45 ng/ml). The highest quartile of the AMH level was 164

6 Hormone prediction of ovarian hyperstimulation syndrome 7ng/ml for 48 patients in one study (La Marca et al., 2007), which was different from 3.35 ng/ml for the 262 patients in our study. This seems to indicate that different kits for detecting AMH and/or different study designs might result in substantial variation in the serum level of AMH. The cut-off value of 3.36 ng/ml for the AMH serum level (almost the same as 3.35 ng/ml for the highest quartile) as selected by the ROC curve in our study would likely yield the greatest predictor value for OHSS (Fig. 2 and Table II). The serum AMH level either for the highest quartile group or selected by the ROC curve might provide a threshold level to determine the risk of OHSS for IVF patients. A multivariate logistic regression model showed that the basal serum level of AMH and E 2 level on the day of hcg administration were the significant predictors of OHSS. In order to clarify the effect of age and BMI for predicting OHSS, a further logistic regression by forced entry method was performed. The results were similar to that shown in Table III: the odds ratios for age, BMI, AMH and E 2 levels were (P ¼ 0.097), (P ¼ 0.146), (P, 0.001) and (P ¼ 0.020), respectively. These results appear to be contrary to two studies (Orvieto, 2003; Papanikolaou et al., 2006), which indicated that the number of follicles present was superior to the serum E 2 level as a predictor for OHSS. The difference between our study and the study by Papanikolaou et al. (2006) might be related to different protocols (GnRH agonist versus GnRH antagonist), because the number of follicles present and the serum E 2 level on the day of hcg injection appeared to have been modified by stimulation protocols for GnRH antagonist compared with protocols for GnRH agonist (Lee et al., 2005). Another possible explanation is that the serum E 2 level and the number of follicles are closely related and certainly not independent of each other in predicting for OHSS. The advantage of logistic regression was that the overall model provided higher specificity compared with the single AMH level (93.5 versus 81.3%, respectively). Therefore, to avoid false positives and subsequent cycle cancellations due to poor response to mild stimulation protocols (Hendriks et al., 2004), the overall logistic regression model may be more applicable than the AMH level only in predicting OHSS. Namely, the basal serum AMH levels in combination with serum E 2 levels during COS would be more helpful in preventing OHSS. In this study, all the patients with false positive results according to the cut-off value of AMH would be in the highest quartile of the AMH level and cancellation of COS due to poor response would not be likely. Cancellation was more probable in the patients in the lowest quartile of the AMH level in this study (Table V) and that reported by La Maca et al. (2007). However, the possibility of inadequate ovarian response following milder stimulation for these false positive patients remains a clinical issue and deserves further elucidation. The combination of basal AMH and E 2 levels might better predict OHSS than AMH levels alone when trying to decrease the false positive rate. One alternative is to elevate the cut-off value of AMH in order to avoid misclassification of patients as false positive. Unfortunately, the sensitivity of basal AMH levels at a 5% false positive level would drop dramatically to only 40%. As a result, more precise prediction of OHSS prior to COS warrants further detailed investigation. The AMH level has been reported to affect two important regulatory steps during normal folliculogenesis: the initial recruitment of primordial follicles and the cyclic recruitment of FSH-dependent follicles (Durlinger et al., 2002; Gruijters et al., 2003). The serum level of AMH may precisely reflect the number of small to intermediate follicles following COS since serum AMH levels has been shown to correlate strongly with the number of small antral follicles present (diameter,12 mm) (Fanchin et al., 2003a; Gruijters et al., 2003). In addition, AMH protein is not expressed in pre-ovulatory or atretic follicles in either mouse or rat ovaries (Durlinger et al., 2002). The basal serum levels of AMH may effectively represent the number of follicles recruited from the pool of primordial follicles (at the period for transition of follicular development from initial recruitment to cyclic recruitment). Following gonadotropin stimulation as part of the IVF process, the number of active growing follicles may be more responsible for the occurrence of OHSS during IVF cycles than the number of large antral follicles. This scenario would be helpful in explaining why the serum E 2 level on the day of hcg administration has demonstrated only a mild significance when compared with basal serum levels of AMH, since granulosa cells from large antral and pre-ovulatory follicles are considered to secrete more E 2 than from small follicles. To the best of our knowledge, the specific role of AMH in follicular sensitivity to FSH (cyclic recruitment) has not always been consistently reported. In an AMH null mouse model with supplemental FSH stimulation, follicle growth was more pronounced than in wild-type mice in terms of the numbers of follicles present (Durlinger et al., 2001). This suggests that ovarian follicles were more sensitive to FSH in the absence of AMH. In contrast, in in vitro cultures of preantral follicles from rat ovaries, supplementation of AMH enhanced FSH effects on follicular growth (McGee et al., 2001). The difference between these two studies was explained by the use of different species and different culture environments in the experiments (Gruijters et al., 2003). It has been shown that the serum level of AMH correlated positively with the number of retrieved oocytes during an IVF cycle (Seifer et al., 2002; van Rooij et al., 2002). In addition, a high serum level of AMH was associated with OHSS and a low total recombinant FSH dose in our study and in some other reports (Nakhuda et al., 2006; La Marca et al., 2007). In humans, AMH seemed to enhance the effect of FSH stimulation on follicular growth. Nonetheless, the relationship between follicular sensitivity to FSH and the serum AMH level in humans warrants further investigation. The serum AMH levels observed during normal menstrual cycles appears to feature either a small, but statistically significant, fluctuation (Cook et al., 2000) or no substantial fluctuation (Hehenkamp et al., 2006; La Marca et al. 2006). Furthermore, it has been reported that the serum AMH levels on any day of the menstrual cycle can predict the ovarian response to COS (La Marca et al., 2007). This might help to at least partly explain why the basal serum AMH level could 165

7 Lee et al. function as a predictor for OHSS in subsequent stimulation cycles. The post-stimulation Day 5 serum AMH level has been shown to be a better predictor of ovarian response to COS than the basal serum AMH level (Penarrubia et al., 2005). In addition, the serum AMH level declined progressively during COS when using a GnRH agonist protocol (Fanchin et al., 2003a; La Marca et al. 2004). For our study, the basal serum AMH was measured on the same day as baseline Day 3 FSH measurements, when suitable COS protocols for individual infertile patients were determined. Therefore, the basal serum AMH level would be helpful in directing the application appropriate stimulation protocols. Whether the AMH level in the early follicular phase during COS or on the day of hcg administration would likely have a greater predictive value for OHSS than the basal serum AMH level deserves further investigation. This study has shown that the serum level of AMH was not able to accurately predict pregnancy, similar to the results of some studies (Penarrubia et al., 2005; Smeenk et al., 2007). The serum AMH level has been regarded as a quantitative marker of ovarian reserve and response in COS. To achieve successful pregnancy following IVF, a number of additional parameters, such as sperm-related factors, embryo development and endometrium receptivity, need to be considered. In contrast, subject age remained a good predictor of pregnancy as shown in this study and other reports (Chuang et al., 2003; Lee et al., 2006). These results suggest that AMH is an excellent marker for the quantitative aspects of ovarian reserve and response, but is not necessarily a good predictor of successful pregnancy. In conclusion, the baseline serum AMH level is a more reliable marker than age and BMI in predicting OHSS prior to COS. According to logistic regression, both the basal serum AMH level and serum E 2 level on the day of hcg administration were significant predictors of moderate and severe OHSS. In addition to functioning as an indicator of ovarian reserve/aging, the serum AMH level also could be utilized to predict OHSS during IVF cycles. In order to prevent OHSS, a mild, patient-friendly stimulation protocol may be applied in patients with a high basal serum AMH level. Acknowledgements We appreciate the help in biostatistics from Chi-Lin Chen, Assistant professor, Graduate Institute in Clinical Medicine, College of Medicine, National Taiwan University. References Aboulghar M. Prediction of ovarian hyperstimulation syndrome (OHSS). Estradiol level has an important role in the prediction of OHSS. Hum Reprod 2003;18: Chen CD, Wu MY, Chao KH, Chen SU, Ho HN, Yang YS. Serum estradiol level and oocyte number in predicting severe ovarian hyperstimulation syndrome. J Formos Med Assoc 1997;96: Chen CD, Chen HF, Lu HF, Chen SU, Ho HN, Yang YS. Value of serum and follicular fluid cytokine profile in the prediction of moderate to severe ovarian hyperstimulation syndrome. Hum Reprod 2000;15: Chuang CC, Chen CD, Chao KH, Chen SU, Ho HN, Yang YS. Age is a better predictor of pregnancy potential than basal follicle-stimulating hormone levels in women undergoing in vitro fertilization. Fertil Steril 2003;79: Cook CL, Siow Y, Brenner AG, Fallat ME. Relationship between serum mullerian-inhibiting substance and other reproductive hormones in untreated women with polycystic ovary syndrome and normal women. Fertil Steril 2002;77: Cook CL, Siow Y, Taylor S, Fallat ME. Serum mullerian-inhibiting substance levels during normal menstrual cycles. Fertil Steril 2000;73: Delvigne A, Dubois M, Battheu B, Bassil S, Meuleman C, De Sutter P, Rodesch C, Janssens P, Remacle P, Gordts S et al. The ovarian hyperstimulation syndrome in in-vitro fertilization: a Belgian multicentric study. II. Multiple discriminant analysis for risk prediction. Hum Reprod 1993;8: Durlinger AL, Gruijters MJ, Kramer P, Karels B, Kumar TR, Matzuk MM, Rose UM, de Jong FH, Uilenbroek JT, Grootegoed JA et al. Anti-Mullerian hormone attenuates the effects of FSH on follicle development in the mouse ovary. Endocrinology 2001;142: Durlinger AL, Visser JA, Themmen AP. Regulation of ovarian function: the role of anti-mullerian hormone. Reproduction 2002;124: Fanchin R, Schonauer LM, Righini C, Frydman N, Frydman R, Taieb J. Serum anti-mullerian hormone dynamics during controlled ovarian hyperstimulation. Hum Reprod 2003a;18: Fanchin R, Schonauer LM, Righini C, Guibourdenche J, Frydman R, Taieb J. Serum anti-mullerian hormone is more strongly related to ovarian follicular status than serum inhibin B, estradiol, FSH and LH on day 3. Hum Reprod 2003b;18: Ficicioglu C, Kutlu T, Baglam E, Bakacak Z. Early follicular antimullerian hormone as an indicator of ovarian reserve. Fertil Steril 2006;85: Golan A, Ron-el R, Herman A, Soffer Y, Weinraub Z, Caspi E. Ovarian hyperstimulation syndrome: an update review. Obstet Gynecol Surv 1989;44: Gruijters MJ, Visser JA, Durlinger AL, Themmen AP. Anti-Mullerian hormone and its role in ovarian function. Mol Cell Endocrinol 2003;211: Hehenkamp WJ, Looman CW, Themmen AP, de Jong FH, Te Velde ER, Broekmans FJ. Anti-Mullerian hormone levels in the spontaneous menstrual cycle do not show substantial fluctuation. J Clin Endocrinol Metab 2006;91: Hendriks DJ, Klinkert ER, Bancsi LFJM, Looman CWN, Habbema JDF, te Velde ER, Broekmans FJ. Use of stimulated serum estradiol measurements for the prediction of hyeprrespinse to ovarian stimulation in in vitro fertilization (IVF). J Assist Reprod Genet 2004;21: La Marca A, Malmusi S, Giulini S, Tamaro LF, Orvieto R, Levratti P, Volpe A. Anti-Mullerian hormone plasma levels in spontaneous menstrual cycle and during treatment with FSH to induce ovulation. Hum Reprod 2004;19: La Marca A, Stabile G, Artenisio AC, Volpe A. Serum anti-mullerian hormone throughout the human menstrual cycle. Hum Reprod 2006;21: La Marca A, Giulini S, Tirelli A, Bertucci E, Marsella T, Xella S, Volpe A. Anti-Mullerian hormone measurement on any day of the menstrual cycle strongly predicts ovarian response in assisted reproductive technology. Hum Reprod 2007;22: Lee TH, Wu MY, Chen HF, Chen MJ, Ho HN, Yang YS. Ovarian response and follicular development for single-dose and multiple-dose protocols for gonadotropin-releasing hormone antagonist administration. Fertil Steril 2005;83: Lee TH, Chen CD, Tsai YY, Chang LJ, Ho HN, Yang YS. Embryo quality is more important for younger women whereas age is more important for older women with regard to in vitro fertilization outcome and multiple pregnancy. Fertil Steril 2006;86: Mathur RS, Joels LA, Akande AV, Jenkins JM. The prevention of ovarian hyperstimulation syndrome. Br J Obstet Gynaecol 1996;103: McGee EA, Smith R, Spears N, Nachtigal MW, Ingraham H, Hsueh AJ. Mullerian inhibitory substance induces growth of rat preantral ovarian follicles. Biol Reprod 2001;64: Morris RS, Paulson RJ, Sauer MV, Lobo RA. Predictive value of serum oestradiol concentrations and oocyte number in severe ovarian hyperstimulation syndrome. Hum Reprod 1995;10: Nakhuda GS, Chu MC, Wang JG, Sauer MV, Lobo RA. Elevated serum mullerian-inhibiting substance may be a marker for ovarian hyperstimulation syndrome in normal women undergoing in vitro fertilization. Fertil Steril 2006;85: Navot D, Bergh PA, Laufer N. Ovarian hyperstimulation syndrome in novel reproductive technologies: prevention and treatment. Fertil Steril 1992;58: Orvieto R. Prediction of ovarian hyperstimulation syndrome. Challenging the estradiol mythos. Hum Reprod 2003;18:

8 Hormone prediction of ovarian hyperstimulation syndrome Papanikolaou EG, Tournaye H, Verpoest W, Camus M, Vernaeve V, Van Steirteghem A, Devroey P. Early and late ovarian hyperstimulation syndrome: early pregnancy outcome and profile. Hum Reprod 2005;20: Papanikolaou EG, Pozzobon C, Kolibianakis EM, Camus M, Tournaye H, Fatemi HM, Van Steirteghem A, Devroey P. Incidence and prediction of ovarian hyperstimulation syndrome in women undergoing gonadotropinreleasing hormone antagonist in vitro fertilization cycles. Fertil Steril 2006;85: Penarrubia J, Fabregues F, Manau D, Creus M, Casals G, Casamitjana R, Carmona F, Vanrell JA, Balasch J. Basal and stimulation day 5 anti-mullerian hormone serum concentrations as predictors of ovarian response and pregnancy in assisted reproductive technology cycles stimulated with gonadotropin-releasing hormone agonist gonadotropin treatment. Hum Reprod 2005;20: Schenker JG. Prevention and treatment of ovarian hyperstimulation. Hum Reprod 1993;8: Seifer DB, MacLaughlin DT, Christian BP, Feng B, Shelden RM. Early follicular serum mullerian-inhibiting substance levels are associated with ovarian response during assisted reproductive technology cycles. Fertil Steril 2002;77: Smeenk JM, Sweep FC, Zielhuis GA, Kremer JA, Thomas CM, Braat DD. Antimullerian hormone predicts ovarian responsiveness, but not embryo quality or pregnancy, after in vitro fertilization or intracyoplasmic sperm injection. Fertil Steril 2007;87: van Rooij IA, Broekmans FJ, te Velde ER, Fauser BC, Bancsi LF, de Jong FH, Themmen AP. Serum anti-mullerian hormone levels: a novel measure of ovarian reserve. Hum Reprod 2002;17: van Rooij IA, Broekmans FJ, Scheffer GJ, Looman CW, Habbema JD, de Jong FH, Fauser BJ, Themmen AP, te Velde ER. Serum antimullerian hormone levels best reflect the reproductive decline with age in normal women with proven fertility: a longitudinal study. Fertil Steril 2005;83: Varma TR, Patel RH. Ovarian hyperstimulation syndrome. A case history and review. Acta Obstet Gynecol Scand 1988;67: Submitted on March 27, 2007; resubmitted on June 30, 2007; accepted on July 11,