Human Reproduction vol.7 no.6 pp.758-764, 1992
Luteal phase support and severe ovarian hyperstimulation syndrome
N.McClureM, J.Leya2, E.Radwanska2, R.Rawlins2 and R.V.Haning, Jr M 'Departments of Obstetrics and Gynecology, Brown University and Women and Infants' Hospital, 101 Dudley Street, Providence, RI 02905 and 2Rush-Presbyterian-St Luke's Medical Center, Chicago, IL, USA 'Present address: Department of Obstetrics and Gynaecology, Monash Medical Centre, Clayton Road, Melbourne, Australia 4
To whom correspondence should be addressed
The incidence and statistical associations of the ovarian hyperstimulation syndrome (OHSS) were studied in 304 egg retrievals with gonadotrophin-releasing hormone agonist suppression, gonadotrophin administration and follkular aspiration. In addition to preserving corpus luteum function, the luteal phase administration of human chorionic gonadotrophin (HCG) was associated with a higher incidence of severe OHSS than was supplementation with progesterone alone (12 versus 0%, P < 0.001). Severe OHSS occurred in 3.7% and 12% of retrievals without and with pregnancy respectively (P < 0.01). Stepwise logistic regression showed that the occurrence of moderate or severe OHSS was statistically predicted by the log of the serum oestradiol on the day the initial HCG was given (P < 0.0001), treatment with luteal phase HCG (P < 0.0003), and fetal number (P < 0.0079). In the late luteal phase of cycles without luteal HCG, the serum oestradiol concentration was one-tenth and the serum progesterone concentration was one-fifth of the luteal phase value with HCG support (P < 0.001). Without luteal phase HCG, oestradiol was two-fold higher (P < 0.001) and progesterone was 1.4-fold higher (P < 0.005) in pregnant than in non-pregnant women. With luteal phase HCG, oestradiol was 1.4-fold higher in pregnant than in nonpregnant women (P < 0.05), and progesterone was 1.7-fold higher (P < 0.001). Oestradiol upper limits of 4400 and 14 700 pmol/1 (1200 and 4000 pg/ml) for cycles with and without luteal phase HCG respectively correspond to ~ 5% risk of moderate or severe OHSS with a singleton pregnancy under these conditions. Key words: corpus luteum/HCG/luteal hyperstimulation syndrome/pregnancy
phase/ovarian
Introduction The risk of the ovarian hyperstimulation syndrome (OHSS) following gonadotrophin therapy is increased in young women (Navot et al., 1988), if pregnancy occurs (Haning et al., 1983, 758
1984), if the plasma or serum oestradiol level is high on the day the first injection of human chorionic gonadotrophin (HCG) is given (Schenker and Weinstein, 1978), or if ultrasound demonstrates the presence of large numbers of follicles (Golan et al., 1989). One practical method of attempting to reduce the incidence of severe OHSS has been to set an upper limit for the serum oestradiol concentration above which a treatment cycle is cancelled. However, oestradiol levels corresponding to a 5% risk of severe OHSS are different for polycystic ovarian disease and hypothalamic amenorrhoea patients, suggesting that a number of factors could affect the predictive effect of the serum oestradiol on the day HCG is given (Haning et al., 1984). Experience suggests that with follicular aspiration, upper limits can be set at a higher level (Golan et al., 1989; Asch et al., 1990; Chenette etal., 1990). A number of studies have implicated gonadotrophin-releasing hormone agonist (GnRHa) as a further risk factor for OHSS (Caspi etal., 1989; Golan etal., 1989; Forman etal., 1990; Herman et al., 1990; Lindner et al., 1990). Furthermore, in a small randomized trial, the use of HCG for luteal phase support, whilst resulting in an improved pregnancy rate, was also associated with a higher rate of OHSS in comparison with similarly treated subjects who received no luteal phase support (Herman et al., 1990). The present larger retrospective study was designed to compare the effects of two different forms of luteal phase support (HCG and progesterone) on the incidence of OHSS in a larger population of women treated with GnRHa and human menopausal gonadotrophin (HMG) with follicular aspiration. The study shows that the incidence is low unless luteal phase HCG is used, and that the use of progesterone alone for luteal phase support permits an excellent pregnancy rate. Materials and methods The study was undertaken retrospectively by utilization of data available from treatment cycles conducted by the authors at their respective institutions. Treatment protocols were similar in both centres. Leuprolide acetate (Lupron, Tap Pharmaceuticals Inc., North Chicago, IL, USA) suppression was started in the mid to late luteal phase and continued for 2 - 4 weeks, at a dose of 0.5 mg s.c. qd, prior to and throughout treatment with HMG. The total daily dose of HMG varied from 75 IU of luteinizing hormone (LH) and 75 IU of follicle stimulating hormone (FSH) to 450 IU of LH and 900 IU of FSH, administered on a bid schedule based on demonstrated gonadotrophin sensitivity or resistance of the individual patients. Oestradiol and progesterone levels were measured at similar times during the treatment cycle. The majority of egg retrievals © Oxford University Press
Luteal HCG and ovarian hyperstimulatkm syndrome
were performed transvaginally under ultrasound guidance, but in both centres a few were performed laparoscopically (a total of 70 retrievals). In both institutions, decisions to cancel cycles due to a perceived risk of OHSS were based solely on the serum oestradiol concentration, with an attempt to keep the serum oestradiol concentration < 14 700 pmol/1 (4000 pg/ml) on the day the first dose of HCG was given. Both institutions aimed to have multiple follicles of at least 16 mm in diameter with at least 744 pmol/1 (200 pg/ml) serum oestradiol per follicle > 15 mm at the time of the first dose of HCG. By combining the data on a total of 304 sequential egg retrievals, we were able to assess data on 138 cycles with luteal phase progesterone support only and 166 cycles with luteal HCG supplementation. Of the latter cycles, 83 were treated with HCG only, and 83 were treated with progesterone plus HCG for luteal phase support. Of the total of 304 patients, 165 patients received 10 000IU of HCG i.m. and 139 patients received 5000 IU of HCG. In all cases, the HCG dose was timed 35 h before egg retrieval. All 138 patients treated with progesterone alone for luteal phase support received an initial 50 mg dose of progesterone in sesame seed oil i.m. on the day of embryo transfer (HCG day 5 with the day of the initial HCG injection counted as HCG day 1). Subsequendy these patients received 25 mg of progesterone i.m. qd unless the serum progesterone/oestradiol molar ratio was < 100 on HCG day 7 - 8 or HCG day 12-14, in which case the dose was increased to 50 mg i.m. qd. Once increased the dose was never decreased until therapy ceased on HCG day 60. A total of 166 patients received HCG for luteal phase support: 140 patients (84%) received 1500 IU of HCG i.m. on HCG day 5, 8, and 11, and 26 other patients (16%) received 2500 IU of HCG i.m. on the same schedule. The higher dose of HCG was reserved for patients with lower oestradiol concentrations. Of the 140 patients on the 1500 IU support dose of HCG, 118 (84%) received 5000 IU of HCG as their initial dose whilst 22 (16%) received 10 000 IU. Of the 26 patients who received the 2500 IU support dose of HCG, 21 (81 %) received the 5000 IU initial dose and five (19%) 10 000 IU. In addition, 83 patients (50%) received 25 or 50 mg of progesterone in oil i.m. with the first dose of luteal HCG. Polycystic ovarian disease (PCOD) was the only diagnosis in the population of infertile women which was thought to increase the incidence of servere OHSS. Therefore we included under this diagnosis, for completeness all anovulatory women with a positive progestational challenge test and who did not have abnormal levels of FSH, prolactin, thyroxin, or thyroid stimulating hormone (TSH). There was a total of 38 treatment cycles in women with PCOD.
progesterone kit at bom institutions (between- and within-assay variation 7.2% and 5.8% respectively). OHSS In order to produce an objective graded response variable for statistical evaluation of OHSS, a hyperstimulation score was assigned to each cycle as previously reported by Haning et al. (1984). Briefly, ovarian enlargement was assigned a score from 0—3 as follows: no enlargement, 0; enlargement up to 5 X 5 cm, 1; enlargement > 5 X 5 cm and up to 10 x 10 cm, 2; and enlargement > 10 X 10 cm, 3. Weight gain was similarly scored from 0 - 3 as follows: no weight gain, 0; weight gain 4.5 kg and :S9 kg, 2; and weight gain > 9 kg, 3. The hyperstimulation score was assigned a value equal to the sum of the ovarian size score and die weight score. Thus a cycle without hyperstimulation was assigned a score of 0, whereas a case of severe OHSS could receive, at most, a score of 6. Hyperstimulation score values of 1 and 2 corresponded to mild OHSS, 3 corresponded to moderate OHSS, and 4, 5 and 6 corresponded to severe OHSS as defined by Jewelewicz et al. (1973). Statistical evaluation Tests of statistical significance for comparison between groups were made using a one-way analysis of variance (ANOVA) for continuously distributed variables and chi-square for discrete variables. Stepwise logistic regression was used to evaluate the predictive effect of variables on the occurrence of OHSS. All data available for each calculation were used; missing data resulted in reduced numbers of values for some results. Data for oestradiol and progesterone were log-transformed prior to statistical analysis to correct for the log-normal distribution. All statistical calculations were performed using the Minitab statistical program (Ryan et al., 1985) except for logistic regression which was performed using SASrelease6.04 (SAS Institute Inc., 1990). All Mests were evaluated using two-tailed tables. Where multiple comparisons were made between the treatment means, the increased probability of finding a significant difference was corrected for by using the sequential method of Newman and Keuls (Snedecor and Cochran, 1967). For the comparison of the two oestradiol assays, log transformation of the results were used because of die wide range of values for oestradiol in serum and the log-normal distribution of values for serum oestradiol. Because of the bivariate normal distribution of these data, the orthogonal regression coefficients for a 'regression of the second kind' were calculated as described by Rodbard (1975).
Assays
Results
Oestradiol was determined by radioimmunoassay using the Pantex (Santa Monica, CA, USA) direct [l25I]oestradiol kit (between- and within-assay variation 7.3% and 4.2% respectively) at the Women and Infants' Hospital and by using the Diagnostic Products Corporation (Los Angeles, CA, USA) CoatA-Count oestradiol kit (between- and within-assay variation 4.5% and 4.2% respectively) at Rush-Presbyterian-St Luke's Medical Center. Progesterone was determined using the Diagnostic Products Corporation (Los Angeles, CA, USA) Coat-A-Count
In die subset of 291 cases for which the number of gestational sacs observed in die initial ultrasound and all data for calculation of die hyperstimulation score were available, diere were 99 pregnancies (positive HCG and uterine sac by ultrasound) and 19 cases of severe OHSS (hyperstimulation score of 4, 5, or 6). The overall incidence of severe OHSS was 3.7% (7/192) in cycles not leading to pregnancy and 12% (12/99) in pregnancy cycles (F < 0.01 by chi-square). Severe OHSS occurred in 9% of single pregnancies (7/77) but in 23% of multiple pregnancies (5/22). 759
N.McClure et al.
Ascites was present in all 19 cases of severe OHSS and five of the cases also had pleural effusions. There were three cases classified as moderate OHSS (hyperstimulation score of 3): two who received only progesterone for luteal phase support and one who received HCG luteal phase support. A small amount of ascites was also present in all three cases of moderate OHSS. All treatment methods from both centres were combined to look for an effect of the three different types of luteal phase support. The rate of severe OHSS was 13 % for HCG alone, 11 % for HCG + progesterone, and 0% for progesterone alone. Accordingly, the HCG only and HCG + progesterone treatments were combined for statistical analysis to obtain the effect of luteal phase HCG treatment. Using HCG for luteal phase support gave a significantly higher incidence of severe OHSS than did the use of progesterone alone, by chi-square analysis [19/153 (12%) versus 0/138 (0%), P < 0.001]. All but one of the 19 women with severe OHSS had received 5000 IU of HCG as the first dose and all 19 received the 1500 IU HCG dose for luteal phase support. This was due to a clinical decision to utilize lower doses of HCG for women with high serum oestradiol levels. Because the serum oestradiol concentration was measured with the Pentex kit in Providence and with the Diagnostic Products kit in Chicago, we calibrated one assay against the other as described by Rodbard (1975). Using 56 samples from IVF patients with a range and distribution of values typical of the IVF cycles reported here, the slopes of the regression lines of the log-transformed data were not significantly different from unity, allowing calculation of the log10 (Pantex oestradiol/Diagnostic Products oestradiol), the mean and SE of that value, and the geometric mean and 95% confidence limits for Pantex oestradiol/ Diagnostic Products oestradiol as described (Rodbard, 1975). The Pantex method was found to give oestradiol results [geometric mean (95% confidence limits)] 1.24 (1.17-1.31; n = 56) times higher than those provided by the Diagnostic Products kit. To evaluate further whether this difference was constant over time or was related only to the single assays utilized for the above comparison, we utilized data obtained incidental to the American College of Pathologists quality control programme conducted on a nationwide basis in the USA. The oestradiol values obtained in the Providence laboratory were compared with the Pantex kit to those obtained in multiple laboratories across the country for 1990 and the first quarter of 1991. The identical geometric mean (95% confidence limits) of 1.24 (1.10-1.40, n = 10) was obtained for the ratio of a value obtained in our laboratory in comparison with the Diagnostic Products result for the same sample on a nationwide basis, confirming that this between-assay difference had been constant for the past year. Thus, comparison of oestradiol values between patients where different methods were used for measuring oestradiol was only possible after correcting for this systematic difference between the two methods. This was done by dividing all results obtained with the Pantex assay by 1.24 to express the results in terms of the Diagnostic Products oestradiol assay. Because of the possibility that multiple factors might be involved in development of OHSS, we analysed eight independent variables as statistical predictors of OHSS using logistic regression. Analysis of the data by logistic regression would have been impossible with all occurrences of severe OHSS in the group 760
with luteal phase HCG support; also, since the patients with moderate OHSS under this classification were physiologically similar to those with severe OHSS (formation of ascites), we combined all cases with hyperstimulation scores of 3 - 6 (moderate and severe OHSS), assigning a value of one for logistic regression analysis. All cases with hyperstimulation scores of < 2 were considered not to have OHSS for the purposes of logistic regression analysis and were assigned the value of zero. The eight variables evaluated as statistical predictors (or independent variables in the model) were the log of the corrected oestradiol concentration, pregnancy, fetal number, age, FSH dose, LH dose, duration of gonadotrophin treatment and treatment with luteal phase HCG. We used stepwise logistic regression with the logit link function, and the P value of 0.05 for both entry and retention of any predictor in the model. Three of the eight potential predictors were found to be predictive of the occurrence of moderate or severe OHSS as defined above. In the order in which they were added to the model and of decreasing statistical significance, these were the log of serum oestradiol on the day that HCG was subsequently given (P < 0.0001), treatment with luteal phase HCG (P < 0.0003), and fetal number (P < 0.0079) (Table I). In the presence of these three variables, none of the other potential statistical predictors were statistically significant at a level of P < 0.05. The addition of either the diagnosis of polycystic ovarian disease (PCOD) or laparoscopic versus vaginal egg retrieval to the model did not alter this conclusion. Using
Table I. Results from the stepwise logistic regression analysis used to identify potential predictors of the occurrence of moderate or severe ovarian hyperstimulation syndrome (OHSS) Variable
Parameter estimate
Standard error
Wald chi-square
P-value from chi-square
Intercept Log l0 [oestradiol] Luteal phase HCG Fetal no.
26.2697 -6.3915
5.4914 1.5739
22.8843 16.4921
0.0001 0.0001
-2.8835 -0.8274
0.8000 0.3114
12.9924 7.0597
0.0003 0.007S
HCG = human chorionic gonadotrophin.
Table II. Effect of number of fetuses and human chorionic gonadotrophin (HCG) luteal phase support on values for oestradiol, calculated from the logistic regression equation, predictive of a 5% incidence of moderate or severe ovarian hyperstimulation syndrome (OHSS) Number of fetuses
Serum oestradiol concentration Without luteal phase HCG support (pmol/1)
With luteal phase HCG support (pmol/l)
0 1 2 3
16376 12155 9023 6696
5797 4302 3194 2371
The value of logit^ (P) was set equal to 2.9444390 to solve for oestradiol (in pg/ml), using values for luteal HCG of 0 and 1 and values for fetal number of 0. 1. 2. and 3. The logistic regression equation was: ) = 26 2691 -6.3915(logi 0 oes
Luteal phase support and severe ovarian hyperstimulation syndrome
N.McClureM, J.Leya2, E.Radwanska2, R.Rawlins2 and R.V.Haning, Jr M 'Departments of Obstetrics and Gynecology, Brown University and Women and Infants' Hospital, 101 Dudley Street, Providence, RI 02905 and 2Rush-Presbyterian-St Luke's Medical Center, Chicago, IL, USA 'Present address: Department of Obstetrics and Gynaecology, Monash Medical Centre, Clayton Road, Melbourne, Australia 4
To whom correspondence should be addressed
The incidence and statistical associations of the ovarian hyperstimulation syndrome (OHSS) were studied in 304 egg retrievals with gonadotrophin-releasing hormone agonist suppression, gonadotrophin administration and follkular aspiration. In addition to preserving corpus luteum function, the luteal phase administration of human chorionic gonadotrophin (HCG) was associated with a higher incidence of severe OHSS than was supplementation with progesterone alone (12 versus 0%, P < 0.001). Severe OHSS occurred in 3.7% and 12% of retrievals without and with pregnancy respectively (P < 0.01). Stepwise logistic regression showed that the occurrence of moderate or severe OHSS was statistically predicted by the log of the serum oestradiol on the day the initial HCG was given (P < 0.0001), treatment with luteal phase HCG (P < 0.0003), and fetal number (P < 0.0079). In the late luteal phase of cycles without luteal HCG, the serum oestradiol concentration was one-tenth and the serum progesterone concentration was one-fifth of the luteal phase value with HCG support (P < 0.001). Without luteal phase HCG, oestradiol was two-fold higher (P < 0.001) and progesterone was 1.4-fold higher (P < 0.005) in pregnant than in non-pregnant women. With luteal phase HCG, oestradiol was 1.4-fold higher in pregnant than in nonpregnant women (P < 0.05), and progesterone was 1.7-fold higher (P < 0.001). Oestradiol upper limits of 4400 and 14 700 pmol/1 (1200 and 4000 pg/ml) for cycles with and without luteal phase HCG respectively correspond to ~ 5% risk of moderate or severe OHSS with a singleton pregnancy under these conditions. Key words: corpus luteum/HCG/luteal hyperstimulation syndrome/pregnancy
phase/ovarian
Introduction The risk of the ovarian hyperstimulation syndrome (OHSS) following gonadotrophin therapy is increased in young women (Navot et al., 1988), if pregnancy occurs (Haning et al., 1983, 758
1984), if the plasma or serum oestradiol level is high on the day the first injection of human chorionic gonadotrophin (HCG) is given (Schenker and Weinstein, 1978), or if ultrasound demonstrates the presence of large numbers of follicles (Golan et al., 1989). One practical method of attempting to reduce the incidence of severe OHSS has been to set an upper limit for the serum oestradiol concentration above which a treatment cycle is cancelled. However, oestradiol levels corresponding to a 5% risk of severe OHSS are different for polycystic ovarian disease and hypothalamic amenorrhoea patients, suggesting that a number of factors could affect the predictive effect of the serum oestradiol on the day HCG is given (Haning et al., 1984). Experience suggests that with follicular aspiration, upper limits can be set at a higher level (Golan et al., 1989; Asch et al., 1990; Chenette etal., 1990). A number of studies have implicated gonadotrophin-releasing hormone agonist (GnRHa) as a further risk factor for OHSS (Caspi etal., 1989; Golan etal., 1989; Forman etal., 1990; Herman et al., 1990; Lindner et al., 1990). Furthermore, in a small randomized trial, the use of HCG for luteal phase support, whilst resulting in an improved pregnancy rate, was also associated with a higher rate of OHSS in comparison with similarly treated subjects who received no luteal phase support (Herman et al., 1990). The present larger retrospective study was designed to compare the effects of two different forms of luteal phase support (HCG and progesterone) on the incidence of OHSS in a larger population of women treated with GnRHa and human menopausal gonadotrophin (HMG) with follicular aspiration. The study shows that the incidence is low unless luteal phase HCG is used, and that the use of progesterone alone for luteal phase support permits an excellent pregnancy rate. Materials and methods The study was undertaken retrospectively by utilization of data available from treatment cycles conducted by the authors at their respective institutions. Treatment protocols were similar in both centres. Leuprolide acetate (Lupron, Tap Pharmaceuticals Inc., North Chicago, IL, USA) suppression was started in the mid to late luteal phase and continued for 2 - 4 weeks, at a dose of 0.5 mg s.c. qd, prior to and throughout treatment with HMG. The total daily dose of HMG varied from 75 IU of luteinizing hormone (LH) and 75 IU of follicle stimulating hormone (FSH) to 450 IU of LH and 900 IU of FSH, administered on a bid schedule based on demonstrated gonadotrophin sensitivity or resistance of the individual patients. Oestradiol and progesterone levels were measured at similar times during the treatment cycle. The majority of egg retrievals © Oxford University Press
Luteal HCG and ovarian hyperstimulatkm syndrome
were performed transvaginally under ultrasound guidance, but in both centres a few were performed laparoscopically (a total of 70 retrievals). In both institutions, decisions to cancel cycles due to a perceived risk of OHSS were based solely on the serum oestradiol concentration, with an attempt to keep the serum oestradiol concentration < 14 700 pmol/1 (4000 pg/ml) on the day the first dose of HCG was given. Both institutions aimed to have multiple follicles of at least 16 mm in diameter with at least 744 pmol/1 (200 pg/ml) serum oestradiol per follicle > 15 mm at the time of the first dose of HCG. By combining the data on a total of 304 sequential egg retrievals, we were able to assess data on 138 cycles with luteal phase progesterone support only and 166 cycles with luteal HCG supplementation. Of the latter cycles, 83 were treated with HCG only, and 83 were treated with progesterone plus HCG for luteal phase support. Of the total of 304 patients, 165 patients received 10 000IU of HCG i.m. and 139 patients received 5000 IU of HCG. In all cases, the HCG dose was timed 35 h before egg retrieval. All 138 patients treated with progesterone alone for luteal phase support received an initial 50 mg dose of progesterone in sesame seed oil i.m. on the day of embryo transfer (HCG day 5 with the day of the initial HCG injection counted as HCG day 1). Subsequendy these patients received 25 mg of progesterone i.m. qd unless the serum progesterone/oestradiol molar ratio was < 100 on HCG day 7 - 8 or HCG day 12-14, in which case the dose was increased to 50 mg i.m. qd. Once increased the dose was never decreased until therapy ceased on HCG day 60. A total of 166 patients received HCG for luteal phase support: 140 patients (84%) received 1500 IU of HCG i.m. on HCG day 5, 8, and 11, and 26 other patients (16%) received 2500 IU of HCG i.m. on the same schedule. The higher dose of HCG was reserved for patients with lower oestradiol concentrations. Of the 140 patients on the 1500 IU support dose of HCG, 118 (84%) received 5000 IU of HCG as their initial dose whilst 22 (16%) received 10 000 IU. Of the 26 patients who received the 2500 IU support dose of HCG, 21 (81 %) received the 5000 IU initial dose and five (19%) 10 000 IU. In addition, 83 patients (50%) received 25 or 50 mg of progesterone in oil i.m. with the first dose of luteal HCG. Polycystic ovarian disease (PCOD) was the only diagnosis in the population of infertile women which was thought to increase the incidence of servere OHSS. Therefore we included under this diagnosis, for completeness all anovulatory women with a positive progestational challenge test and who did not have abnormal levels of FSH, prolactin, thyroxin, or thyroid stimulating hormone (TSH). There was a total of 38 treatment cycles in women with PCOD.
progesterone kit at bom institutions (between- and within-assay variation 7.2% and 5.8% respectively). OHSS In order to produce an objective graded response variable for statistical evaluation of OHSS, a hyperstimulation score was assigned to each cycle as previously reported by Haning et al. (1984). Briefly, ovarian enlargement was assigned a score from 0—3 as follows: no enlargement, 0; enlargement up to 5 X 5 cm, 1; enlargement > 5 X 5 cm and up to 10 x 10 cm, 2; and enlargement > 10 X 10 cm, 3. Weight gain was similarly scored from 0 - 3 as follows: no weight gain, 0; weight gain 4.5 kg and :S9 kg, 2; and weight gain > 9 kg, 3. The hyperstimulation score was assigned a value equal to the sum of the ovarian size score and die weight score. Thus a cycle without hyperstimulation was assigned a score of 0, whereas a case of severe OHSS could receive, at most, a score of 6. Hyperstimulation score values of 1 and 2 corresponded to mild OHSS, 3 corresponded to moderate OHSS, and 4, 5 and 6 corresponded to severe OHSS as defined by Jewelewicz et al. (1973). Statistical evaluation Tests of statistical significance for comparison between groups were made using a one-way analysis of variance (ANOVA) for continuously distributed variables and chi-square for discrete variables. Stepwise logistic regression was used to evaluate the predictive effect of variables on the occurrence of OHSS. All data available for each calculation were used; missing data resulted in reduced numbers of values for some results. Data for oestradiol and progesterone were log-transformed prior to statistical analysis to correct for the log-normal distribution. All statistical calculations were performed using the Minitab statistical program (Ryan et al., 1985) except for logistic regression which was performed using SASrelease6.04 (SAS Institute Inc., 1990). All Mests were evaluated using two-tailed tables. Where multiple comparisons were made between the treatment means, the increased probability of finding a significant difference was corrected for by using the sequential method of Newman and Keuls (Snedecor and Cochran, 1967). For the comparison of the two oestradiol assays, log transformation of the results were used because of die wide range of values for oestradiol in serum and the log-normal distribution of values for serum oestradiol. Because of the bivariate normal distribution of these data, the orthogonal regression coefficients for a 'regression of the second kind' were calculated as described by Rodbard (1975).
Assays
Results
Oestradiol was determined by radioimmunoassay using the Pantex (Santa Monica, CA, USA) direct [l25I]oestradiol kit (between- and within-assay variation 7.3% and 4.2% respectively) at the Women and Infants' Hospital and by using the Diagnostic Products Corporation (Los Angeles, CA, USA) CoatA-Count oestradiol kit (between- and within-assay variation 4.5% and 4.2% respectively) at Rush-Presbyterian-St Luke's Medical Center. Progesterone was determined using the Diagnostic Products Corporation (Los Angeles, CA, USA) Coat-A-Count
In die subset of 291 cases for which the number of gestational sacs observed in die initial ultrasound and all data for calculation of die hyperstimulation score were available, diere were 99 pregnancies (positive HCG and uterine sac by ultrasound) and 19 cases of severe OHSS (hyperstimulation score of 4, 5, or 6). The overall incidence of severe OHSS was 3.7% (7/192) in cycles not leading to pregnancy and 12% (12/99) in pregnancy cycles (F < 0.01 by chi-square). Severe OHSS occurred in 9% of single pregnancies (7/77) but in 23% of multiple pregnancies (5/22). 759
N.McClure et al.
Ascites was present in all 19 cases of severe OHSS and five of the cases also had pleural effusions. There were three cases classified as moderate OHSS (hyperstimulation score of 3): two who received only progesterone for luteal phase support and one who received HCG luteal phase support. A small amount of ascites was also present in all three cases of moderate OHSS. All treatment methods from both centres were combined to look for an effect of the three different types of luteal phase support. The rate of severe OHSS was 13 % for HCG alone, 11 % for HCG + progesterone, and 0% for progesterone alone. Accordingly, the HCG only and HCG + progesterone treatments were combined for statistical analysis to obtain the effect of luteal phase HCG treatment. Using HCG for luteal phase support gave a significantly higher incidence of severe OHSS than did the use of progesterone alone, by chi-square analysis [19/153 (12%) versus 0/138 (0%), P < 0.001]. All but one of the 19 women with severe OHSS had received 5000 IU of HCG as the first dose and all 19 received the 1500 IU HCG dose for luteal phase support. This was due to a clinical decision to utilize lower doses of HCG for women with high serum oestradiol levels. Because the serum oestradiol concentration was measured with the Pentex kit in Providence and with the Diagnostic Products kit in Chicago, we calibrated one assay against the other as described by Rodbard (1975). Using 56 samples from IVF patients with a range and distribution of values typical of the IVF cycles reported here, the slopes of the regression lines of the log-transformed data were not significantly different from unity, allowing calculation of the log10 (Pantex oestradiol/Diagnostic Products oestradiol), the mean and SE of that value, and the geometric mean and 95% confidence limits for Pantex oestradiol/ Diagnostic Products oestradiol as described (Rodbard, 1975). The Pantex method was found to give oestradiol results [geometric mean (95% confidence limits)] 1.24 (1.17-1.31; n = 56) times higher than those provided by the Diagnostic Products kit. To evaluate further whether this difference was constant over time or was related only to the single assays utilized for the above comparison, we utilized data obtained incidental to the American College of Pathologists quality control programme conducted on a nationwide basis in the USA. The oestradiol values obtained in the Providence laboratory were compared with the Pantex kit to those obtained in multiple laboratories across the country for 1990 and the first quarter of 1991. The identical geometric mean (95% confidence limits) of 1.24 (1.10-1.40, n = 10) was obtained for the ratio of a value obtained in our laboratory in comparison with the Diagnostic Products result for the same sample on a nationwide basis, confirming that this between-assay difference had been constant for the past year. Thus, comparison of oestradiol values between patients where different methods were used for measuring oestradiol was only possible after correcting for this systematic difference between the two methods. This was done by dividing all results obtained with the Pantex assay by 1.24 to express the results in terms of the Diagnostic Products oestradiol assay. Because of the possibility that multiple factors might be involved in development of OHSS, we analysed eight independent variables as statistical predictors of OHSS using logistic regression. Analysis of the data by logistic regression would have been impossible with all occurrences of severe OHSS in the group 760
with luteal phase HCG support; also, since the patients with moderate OHSS under this classification were physiologically similar to those with severe OHSS (formation of ascites), we combined all cases with hyperstimulation scores of 3 - 6 (moderate and severe OHSS), assigning a value of one for logistic regression analysis. All cases with hyperstimulation scores of < 2 were considered not to have OHSS for the purposes of logistic regression analysis and were assigned the value of zero. The eight variables evaluated as statistical predictors (or independent variables in the model) were the log of the corrected oestradiol concentration, pregnancy, fetal number, age, FSH dose, LH dose, duration of gonadotrophin treatment and treatment with luteal phase HCG. We used stepwise logistic regression with the logit link function, and the P value of 0.05 for both entry and retention of any predictor in the model. Three of the eight potential predictors were found to be predictive of the occurrence of moderate or severe OHSS as defined above. In the order in which they were added to the model and of decreasing statistical significance, these were the log of serum oestradiol on the day that HCG was subsequently given (P < 0.0001), treatment with luteal phase HCG (P < 0.0003), and fetal number (P < 0.0079) (Table I). In the presence of these three variables, none of the other potential statistical predictors were statistically significant at a level of P < 0.05. The addition of either the diagnosis of polycystic ovarian disease (PCOD) or laparoscopic versus vaginal egg retrieval to the model did not alter this conclusion. Using
Table I. Results from the stepwise logistic regression analysis used to identify potential predictors of the occurrence of moderate or severe ovarian hyperstimulation syndrome (OHSS) Variable
Parameter estimate
Standard error
Wald chi-square
P-value from chi-square
Intercept Log l0 [oestradiol] Luteal phase HCG Fetal no.
26.2697 -6.3915
5.4914 1.5739
22.8843 16.4921
0.0001 0.0001
-2.8835 -0.8274
0.8000 0.3114
12.9924 7.0597
0.0003 0.007S
HCG = human chorionic gonadotrophin.
Table II. Effect of number of fetuses and human chorionic gonadotrophin (HCG) luteal phase support on values for oestradiol, calculated from the logistic regression equation, predictive of a 5% incidence of moderate or severe ovarian hyperstimulation syndrome (OHSS) Number of fetuses
Serum oestradiol concentration Without luteal phase HCG support (pmol/1)
With luteal phase HCG support (pmol/l)
0 1 2 3
16376 12155 9023 6696
5797 4302 3194 2371
The value of logit^ (P) was set equal to 2.9444390 to solve for oestradiol (in pg/ml), using values for luteal HCG of 0 and 1 and values for fetal number of 0. 1. 2. and 3. The logistic regression equation was: ) = 26 2691 -6.3915(logi 0 oes
