1 Division of Human Reproduction, Center for Reproductive Medicine and Surgery, Department of Obstetrics and Gynecology, University of Pennsylvania Medical Center, 3400 Spruce Street, Philadelphia, PA 19104-4283 and 2 Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania Medical Center and Health System, Philadelphia, PA, USA
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: IVF/moderately elevated FSH/ovarian reserve/screening test
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
One of the first large-scale studies evaluating the predictive value of the basal FSH categorized the threshold values as follows: low <15 mIU/ml; moderate 1524.9 mIU/ml; high 25 mIU/ml (Scott et al., 1989
). This study demonstrated a declining pregnancy rate as basal FSH concentration increased, with a moderate decrease in pregnancy rate for women with an FSH between 15 and 24.9 mIU/ml, and a dramatic decrease in pregnancy rate if the FSH was
25 mIU/ml. Others (Licciardi et al., 1995
; Evers et al., 1998
) have used 17 mIU/ml as a cut-off to define an elevated basal FSH concentration. Converting these numbers to the new 78/549 standard results in a substantially lower FSH threshold. An FSH value of 25 mIU/ml would be equivalent to 16.7 mIU/ml (World Health Organization 78/549 standard), a value of 17 mIU/ml equivalent to 11.4 mIU/ml, and a value of 15 mIU/ml equivalent to 10 mIU/ml.
In this study, the predictive value of FSH at various concentrations was determined in order to identify the optimal cut-off. The cut-off values were defined to coincide with those established in the literature, taking into account the change in reference standard. The predictive ability of a moderately elevated and elevated day 3 FSH was evaluated, both independently and after controlling for confounding factors such as diagnosis, age, and follicular response to gonadotrophins.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A total of 293 subjects (94%) was included whose records were complete. Cycle day 3 FSH concentrations were measured in all patients during a spontaneous cycle. All FSH concentrations from the previous 12 months were reviewed, and the highest value recorded was used as the FSH concentration in this study. Other information collected included: patient age, aetiology of infertility, medication used, number of ampoules used, day of human chorionic gonadotrophin administration, estradiol concentration on the day of human chorionic gonadotrophin (HCG) administration, endometrial thickness, whether intracytoplasmic sperm injection (ICSI) was performed, number of oocytes retrieved, number of embryos transferred, and whether or not pregnancy was achieved. All patients, regardless of age or FSH concentration, underwent stimulation.
Stimulation protocol
Ovulation induction was commenced with either a luteal phase or a flare protocol. In patients undergoing the luteal phase protocol, leuprolide acetate (Lupron; TAP Pharmaceuticals, Deerfield, IL, USA) was started at a dose of 10 U (0.5 mg) daily, 78 days after the LH surge. After the withdrawal bleed, the patient underwent baseline ultrasound, estradiol, and LH measurements to confirm ovarian suppression. The cycle was allowed to proceed if estradiol was <50 pg/ml (conversion factor to SI units, 3.67), LH was suppressed, and ultrasound demonstrated absence of cysts. Once gonadotrophins were started, the dose of Lupron was dropped to 5 units (0.25 mg) daily. Medications utilized by patients included: recombinant (rFSH, Gonal-F; Serono, Norwell, MA, USA), urinary FSH (uFSH, Metrodin; Serono), or highly purified (hp) FSH (Fertinex; Serono).
Patients utilizing the flare protocol started 5 units (0.25 mg) of Lupron daily on cycle day 1 with baseline ultrasound and blood work as above. If all was normal, the patients started gonadotrophins on cycle day 2.
Patients were brought back on cycle day 57 for ultrasound, estradiol, and LH measurements. Based on the results of the above tests, the medication dose was adjusted as necessary. HCG, 10 000 U (Profasi; Serono), was administered when the leading follicles were >2022 mm in mean diameter with an appropriate estradiol concentration of ~200 pg/ml per mature follicle. Ultrasound-guided oocyte retrieval was performed 36 h after HCG administration, and embryo transfer was performed on the third day post retrieval. The luteal phase was supported with progesterone in oil, 50 mg i.m. daily, starting the night of oocyte retrieval, and was continued until a negative pregnancy test was obtained or until week 10 of gestation. A serum HCG test was obtained 14 days following embryo transfer.
Definition of outcome
Pregnancy was defined using standard SART definitions. A pregnancy was classified as a chemical pregnancy if the initial HCG concentration exceeded 5 mIU/ml, but no sac was demonstrated on ultrasound. A live birth was defined as a pregnancy resulting in delivery of a viable infant. An ectopic gestation was defined as occurring outside the endometrial cavity, and a spontaneous abortion was defined as a pregnancy lost before week 20 of gestation.
In order to evaluate the predictive value of different ages and FSH values, the data were subdivided into age categories as follows: <35, 3540 and >40 years, and FSH categories of normal (<10 mIU/ml, conversion factor to SI units, 1.00), moderately elevated (1011.4 mIU/ml), elevated (>11.4 mIU/ml) and markedly elevated (>16.7 mIU/ml). These categories were chosen to reflect concentrations of FSH demonstrated to be predictive in other studies, adjusting for the reference preparations used in the respective assays (Scott et al., 1989; Licciardi et al., 1995
; Evers et al., 1998
).
Data analysis
Descriptive analyses were first conducted in the total eligible cohort to examine baseline associations and to assess whether the distributions of continuous variables were consistent with the assumption of normality required for parametric statistical tests. Associations between live births and FSH value, and between live births and other variables, were assessed using odds ratios (OR) calculated using STATA software (College Park, TX, USA). A variable was considered a significant predictor of pregnancy if the 95% confidence interval of the OR did not include 1.0. The sensitivity, specificity, and predictive values were calculated. Multiple 2x2 tables were utilized to obtain the percentage values for sensitivities, specificities, and positive and negative predictive values, with disease defined as not achieving a live birth. Separate tables were constructed to evaluate the sensitivity and predictive value of an FSH concentration at the cut-off value used in this study. The test characteristics of an elevated FSH value were calculated using each cut-off point (>10.0 and >11.4) in the entire population. In order to assess the utility of using an FSH cut-off value of >10, assuming an elevated FSH was predictive, the test characteristics of an FSH value of >10.0, excluding women with an elevated FSH value (>11.4), were also evaluated. All calculations were conducted on the entire sample and in subgroups stratified by age and FSH concentration.
Both unadjusted and stratified analyses were also performed to identify potential confounding variables to the outcome of achieving a live birth. Multivariable logistic regression was then used to estimate the OR for the association of FSH value (above a given cut-off point) and pregnancy, adjusted for all potential confounders. Variables were included in the logistic regression model if: (i) they were biologically plausible, (ii) they demonstrated a P < 0.20 for the association with outcome in the unadjusted analyses, or (iii) adjustment for the given variable changed the OR for the association of FSH with pregnancy by 10% (Mickey and Greenland, 1989; Maldonado and Greenland, 1993
). P < 0.05 was considered statistically significant using two-tailed tests. Data are presented as mean ± SD unless otherwise specified.
Laboratory procedures
FSH concentration was determined using the Bayer Technicon Immuno 1 System, a heterogeneous sandwich magnetic separation assay with a sensitivity of 0.1 mIU/ml. It is standardized against the World Health Organization (WHO) 2nd International Standard (IRP 78/549). The normal FSH value for a reproductive age female in the follicular phase is 1.511.4 mIU/ml using this assay. The intra-assay coefficient (within-run SD; mIU/ml) is 0.13 at an FSH of 5.5, 0.22 at an FSH of 12.3, and 0.49 at an FSH of 30.3. The inter-assay coefficient (total SD; mIU/ml) is 0.18 at an FSH of 5.5, 0.34 at an FSH of 12.3, and 0.85 at an FSH of 30.3.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
The sensitivity, specificity, positive, and negative predictive values for a moderately elevated FSH (1011.4 mIU/ml), an elevated FSH (>11.4 mIU/ml) value in all women, and for an FSH value >10 mIU/ml, excluding women with a value >11.4 mIU/ml, are presented in Table III. Data is presented overall, and stratified by age. The sensitivity of a moderately elevated and elevated FSH value in predicting IVF failure is low for all age groups (6.719.5%). Specificity of moderately elevated and elevated FSH concentrations for all age groups is high, ranging from 90.1 to 100%. The positive predictive value of a moderately elevated FSH concentration ranges from 60 to 100%. The positive predictive value of an elevated FSH was 100%.
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Clinically, a basal day 3 FSH is commonly used as a screening tool to assess the prognosis of achieving a live birth with IVF. In this case, a clinician uses an FSH value prior to ovarian stimulation, without knowledge of response to gonadotrophin stimulation. Thus, prognosis is based solely upon the FSH concentration. A screening test used to identify those with a poor prognosis would optimally identify fertile patients (high specificity) as well as infertile patients (high sensitivity) (Barnhart and Osheroff, 1998). However, screening tests often sacrifice either sensitivity or specificity in order to maximize the other (Barnhart and Osheroff, 1999
). The data presented here confirm that a day 3 FSH has a high specificity but a poor sensitivity (Barnhart and Osheroff, 1998
). Specifically, it was demonstrated that an elevated FSH had the highest sensitivity in women >35 years of age. However, this sensitivity was only 11.3%. Conversely, it was confirmed that FSH is a very specific screening test, as >90.1% of women who did achieve a live birth did not have an elevated FSH. In our series, no woman with an FSH >11.4 mIU/ml achieved a live birth. Conversely, an FSH of 1011.4 mIU/ml did not predict IVF failure in up to 37% of patients. Based on these data, the use of FSH as a screening test is best employed to provide prognostic information for a patient undergoing IVF for the first time using a high cut-off concentration (>11.4 mIU).
The goal of this paper was to evaluate the usefulness of a moderately elevated FSH and an elevated FSH value beyond the traditional 2x2 tables. This was accomplished using logistic regression, assessing the association of an elevated FSH concentration with the ability to achieve a live birth. The use of multivariate analysis allows a statistical evaluation of the association of an elevated FSH to IVF success, while accounting for other factors that may confound outcome. The data presented here demonstrate that without controlling for confounding variables, many factors are statistically associated with the likelihood of achieving pregnancy including: an FSH >11.4 mIU/ml, age, maximum estradiol concentration, length of stimulation, and the number of oocytes retrieved. Therefore, each could serve as an independent predictor of pregnancy outcome. Many of these variables are inter-related and inter-dependent. Controlling for these factors with logistic regression demonstrates that a moderately elevated FSH (>10 mIU/ml) is not significantly associated with failure to achieve a live birth after IVF. Only an elevated FSH (>11.4 mIU/ml) and age are significantly associated with failure to achieve a live birth after IVF. Without controlling for these variables, the data could have been misleading, and would have been subject to many of the biases limiting other manuscripts evaluating the usefulness of FSH as a predictor of treatment outcome. The current data are consistent with recent studies (Bancsi et al., 2000) but contrast with the results of others who have demonstrated that age predicts pregnancy more strongly than FSH (Sharif et al., 1998
). Thus, while an elevated FSH (>11.4 mIU/ml) is an independent predictor of pregnancy, poor prognosis can also be identified on the basis of age and/or prior stimulation response. In other words, if information on these factors is known to the clinician, the additional information obtained with a day 3 FSH is limited.
An important aspect of the current study is that not only was the FSH value just prior to the IVF cycle assessed, but all baseline FSH values were obtained in the preceding 12 months. The highest FSH value recorded was used for analysis. A large variability in values over this time period was noted. The highest value was more predictive of outcome than the FSH immediately preceding the IVF cycle (data not shown). This is consistent with a prior study demonstrating that the chance of pregnancy after an IVF cycle is dramatically decreased in patients with any previous history of an elevated day 3 FSH concentration (Martin et al., 1996).
In conclusion, comparing the results of studies evaluating the predictive value of FSH is problematic. Differences in assays, techniques, and reference preparations exist, such as the change in the Second IRP from the older HMG standard to the newer 78/549 standard. For example, an FSH of 10 mIU/ml (78/549 standard) equals 15 mIU/ml (HMG standard), and an FSH of 11.4 mIU/ml (78/549 standard) equals 17 mIU/ml (HMG standard). The data presented here demonstrate that a day 3 FSH is only predictive of IVF failure when it is elevated (>11.4 mIU/ml, WHO 2nd IRP 78/549 standard). A moderately elevated FSH concentration (1011.4 mIU/ml) does not have any independent predictive ability. Moreover, much of the predictive value of an elevated FSH is confounded by other variables that affect IVF success such as age, infertility diagnosis, and response to stimulation with gonadotrophins. The poor prognosis conferred by an elevated FSH in younger women (<35 years) may be overcome by maximizing the stimulation protocol.
![]() |
Notes |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Barnhart, K.T. and Osheroff, J. (1998) Follicle stimulating hormone as a predictor of fertility. Curr. Opin. Obstet. Gynecol., 10, 227232.[ISI][Medline]
Barnhart, K.T. and Osheroff, J. (1999) We are over interpreting the predictive value of serum FSH. Fertil. Steril., 71, 89.
Evers, J.L.H., Slaats, P., Land, J.A. et al. (1998) Elevated levels of basal estradiol-17ß predict poor response in patients with normal basal levels of follicle-stimulating hormone undergoing in vitro fertilization. Fertil. Steril., 69, 10101014.[ISI][Medline]
Khalifa, E., Toner, J.P., Muasher, S.J. and Acosta, A.A. (1992) Significance of basal follicle-stimulating hormone levels in women with one ovary in a program of in vitro fertilization. Fertil. Steril., 57, 835839.[ISI][Medline]
Lenton, E.A., Sexton, L., Lee, S. and Cooke, I.D. (1988) Progressive changes in LH and FSH and LH:FSH ratio in women throughout reproductive life. Maturitas, 10, 35.[ISI][Medline]
Licciardi, F.L., Liu, H.C. and Rosenwaks, Z. (1995) Day 3 estradiol serum concentrations as prognosticators of ovarian stimulation response and pregnancy outcome in patients undergoing in vitro fertilization. Fertil. Steril., 64, 991994.[ISI][Medline]
Maldonado, G. and Greenland, S. (1993) Simulation study of confounder selection strategies. Am. J. Epidemiol., 138, 923936.[Abstract]
Martin, J.S.B., Nisker, J.A., Tummon, I.S. et al. (1996) Future in vitro fertilization pregnancy potential of women with variably elevated day 3 follicle-stimulating hormone levels. Fertil. Steril., 65, 12381240.[ISI][Medline]
Mickey, R.M. and Greenland, S. (1989) The impact of confounder selection criteria on effect estimation. Am. J. Epidemiol., 129, 125137.[Abstract]
Muasher, S.J., Oehninger, S., Simonetti, S. et al. (1988) The value of basal and/or stimulated serum gonadotropin levels in prediction of stimulation response and in vitro fertilization outcome. Fertil. Steril., 50, 298307.[ISI][Medline]
Navot, D., Rosenwaks, Z. and Margalioth, E.J. (1987) Prognostic assessment of female fecundity. Lancet, ii,645647.
Pearlstone, A.C., Fournet, N., Gambone, J.C. et al. (1992) Ovulation induction in women age 40 and older: the importance of basal follicle stimulating hormone level and chronological age. Fertil. Steril., 58, 674679.[ISI][Medline]
Scott, R.T., Toner, J.P., Muasher, S.J. et al. (1989) Follicle stimulating hormone levels on cycle day 3 are predictive of in vitro fertilization outcome. Fertil. Steril., 51, 651654.[ISI][Medline]
Scott, R.T., Hofmann, G.E., Oehninger, S. and Muasher, S.J. (1990) Intercycle variability of day 3 follicle stimulating hormone levels and its effect on stimulation quality in in vitro fertilization. Fertil. Steril., 54, 297302.[ISI][Medline]
Sharif, K., Elgendy, M., Lashen, H. and Afnan, M. (1998) Age and basal follicle stimulating hormone as predictors of in vitro fertilization outcome. Br. J. Obstet. Gynaecol., 105, 107112.[ISI][Medline]
Tanbo, T., Per Olav, D., Abyholm, T. and Stokke, K.T. (1989) Follicle stimulating hormone as a prognostic indicator in clomiphene citrate/human menopausal gonadotropin stimulated cycles for in-vitro fertilization. Hum. Reprod., 6, 647650.
Toner, J.P., Philput, C.B., Jones, G.S. and Muasher, S.J. (1991) Basal follicle-stimulating hormone level is a better predictor of in vitro fertilization performance than age. Fertil. Steril., 55, 784791.[ISI][Medline]
Submitted on June 18, 2001; accepted on September 28, 2001.