Ovarian response to controlled ovarian hyperstimulation: what does serum FSH say?

Abstract

STUDY QUESTION

Do serum FSH levels on day of hCG trigger differ between women with a poor, normal or hyper response to a fixed daily dose of 150 IU recombinant FSH (rFSH)?

SUMMARY ANSWER

There is no consistent relationship between ovarian response and serum FSH levels on day of hCG trigger in a 150 IU fixed dose treatment protocol.

WHAT IS KNOWN ALREADY

When ovarian response to stimulation for IVF/ICSI is suboptimal, the FSH dose is often adjusted in a subsequent cycle, thereby assuming that serum FSH levels were inadequate for optimal stimulation.

STUDY DESIGN, SIZE, DURATION

Nested cohort study within a randomized controlled trial conducted at the University Medical Centre Utrecht between March 2009 and July 2011. Blood was drawn from 124 women on cycle Day 2 and on day of hCG triggering. Serum FSH level was determined by the Beckman-Coulter Unicel DXi800 chemiluminescence assay. In order to detect a difference of 2 IU/L between poor, normal and hyper responders, a total of 64 participants (16 poor, 32 normal and 16 hyper responders) would provide 80% power, assuming a standard deviation of 2 and an alpha of 0.05.

PARTICIPANTS, SETTING, METHODS

Women aged ≤39 years with a regular cycle and fixed FSH dose of 150 IU. Exclusion criteria: BMI > 32 kg/m² and >2 previous unsuccessful IVF/ICSI cycles. The primary outcome measure was serum FSH level on day of triggering.

MAIN RESULTS AND THE ROLE OF CHANCE

Median [range] body weight was 70.0 kg [55.0–85.6], 68.0 kg [52.0–94] and 60.6 kg [51.0–78.0] for poor (n = 16), normal (n = 94) and hyper (n = 17) responders, respectively. Mean (SD) serum FSH levels on day of triggering were 9.5 IU/L (2.4) in poor, 10.4 IU/L (2.3) in normal and 11.5 IU/L (2.2) in hyper responders. Serum FSH levels on day of hCG in poor responders differed significantly as compared to those in hyper responders (P = 0.03).

LIMITATIONS, REASONS FOR CAUTION

The number of retrieved oocytes is only minimally determined by serum FSH level on the day of hCG trigger. After correction for age, body weight, basal serum FSH and basal anti-Mullerian hormone the correlation between serum FSH level on the day of hCG and ovarian response regarding the number of retrieved oocytes disappeared.

WIDER IMPLICATIONS OF THE FINDINGS

The current study shows that a poor response is not related to inadequate serum FSH levels per se. One could therefore question whether increasing the rFSH dose in women with a suboptimal response is meaningful. In women with a hyper response, however, lowering the dose of rFSH in a subsequent IVF cycle may lead to lower serum FSH levels and thereby mitigate ovarian response and improve safety of the IVF treatment. As this was not a dose–response study, future research should assess whether dose adjustments benefit the poor and hyper responder.

STUDY FUNDING/COMPETING INTEREST(S)

No external funds were obtained for this study. S.C.O, T.C.v.T., O.H., H.L.T., E.G.W.M.L., C.B.L. and M.J.C.E. have nothing to disclose. F.J.M.B. receives monetary compensation: member of the external advisory board for Merck Serono and Ferring, the Netherlands; educational activities for Ferring BV, the Netherlands; consultancy work for Gedeon Richter, Belgium; strategic cooperation with Roche on automated AMH assay development and research cooperation with Ansh Labs.

IVF/ICSI, controlled ovarian stimulation, recombinant FSH, serum FSH level, ovarian response

Introduction

The optimization of ovarian response to controlled ovarian stimulation (COS) for IVF is still an important topic of debate. Recombinant FSH (rFSH) is the most widely used gonadotropin for COS (Macklon et al., 2006). Administration of gonadotropins increases serum FSH levels and extends the period in which the FSH threshold is exceeded (Schipper et al., 1998), which is essential to achieve multiple follicular growth (Schoemaker et al., 1993; Fauser and van Heusden, 1997). Unfortunately, not every woman undergoing COS has the same degree of ovarian response. The optimal response range in terms of live birth rate has been declared to be between 8 and 15 oocytes (van der Gaast et al., 2006; Sunkara et al., 2011). This range may be somewhat different for GnRH antagonist co-treated IVF cycles (Al-Inany et al., 2006; Hamdine et al., 2015). Based on the number of oocytes retrieved, ovarian response is commonly defined as poor, normal or hyper (Ferraretti et al., 2011; Broer et al., 2013). Both a poor and a hyper response are considered inappropriate, for they are associated with higher rates of cycle cancellation and lower pregnancy rates (van der Gaast et al., 2006; Zhen et al., 2008). In hyper responders there may also be an increased risk of developing the ovarian hyperstimulation syndrome (OHSS) (Ji et al., 2013). For patients with an inappropriate ovarian response, the rFSH dose is often adjusted in a subsequent treatment cycle, which is based on the assumption that an inappropriate response is due to inadequate serum FSH levels. A study that, among other outcomes, assessed the relationship between two different recombinant gonadotropin doses for COS in GnRH antagonist cycles and serum FSH levels during stimulation, revealed that a dose of 200 IU compared to 150 IU per day resulted in higher serum FSH levels on the day of hCG trigger (Out et al., 2004). The use of the higher gonadotropin dose also resulted in a higher number of retrieved oocytes (Out et al. 2004), a finding that is confirmed in other dose–response studies (Arce et al., 2014). One may therefore assume that differences in FSH serum levels may, partly, explain variations in ovarian response.

To our knowledge, no studies have been performed that directly evaluated the relationship between serum FSH levels during FSH stimulation and the ovarian response to COS for IVF in GnRH antagonist co-treated cycles. So, it remains to be elucidated if, and if so to what extent, serum FSH levels may explain the variation in ovarian response that is commonly observed in the ART population. If we assume that with the current day standard FSH dosage of 150 IU a maximum stimulation is obtained in the vast majority of women, then other sources for this response variation may be much more likely (Sterrenburg et al., 2011). One of these sources of variation in response could be a variation in serum FSH levels at the end of stimulation with rFSH and thus a variation in FSH to which the ovaries are exposed. An explanation for why serum FSH levels would differ at the end of stimulation in a fixed dose protocol, could be a difference in body weight and consequently in distribution volume of the drug (Mannaerts et al., 1996). A difference in basal serum FSH level could be a second explanation. If one assumes that a fixed dose of rFSH is able to achieve a ‘fixed rise’ in serum FSH level and women vary from each other with respect to their basal serum FSH level, a similar variation in serum FSH at the end of stimulation can be expected. However, these are all assumptions and no evidence is available especially on the relationship with ovarian response to support this. The aim of this study was therefore to assess whether serum FSH levels at the end of stimulation differ significantly between poor, normal and hyper responders to a fixed daily dose of 150 IU rFSH in GnRH antagonist co-treated cycles.

Materials and Methods

Patient population and ethical approval

This was a nested cohort study within a multicentre randomized controlled trial, the CETRO Trial (Hamdine et al., 2013). This Dutch trial on the timing of GnRH antagonist initiation was conducted between March 2009 and July 2011. It was approved by the institutional review board of the University Medical Centre (UMC) Utrecht and registered on clinicaltrials.gov (NCT00866034). Written informed consent was obtained from all women participating in the trial. The present study only concerns women undergoing IVF or ICSI from the IVF outpatient clinic of the department of Reproductive Medicine and Gynaecology of the UMC Utrecht. The inclusion criteria of the CETRO Trial were: women aged ≤39 years with a regular menstrual cycle (average cycle length between 21 and 35 days) and an indication for IVF of ICSI. Exclusion criteria were: BMI > 32 kg/m² and more than two previous unsuccessful IVF/ICSI cycles. For this nested cohort study, we added three exclusion criteria: rFSH starting dose ≠ of 150 IU/L, rFSH dose adjustment during stimulation and absence of a serum sample on cycle Day (CD) 2 or day of hCG trigger.

Stimulation protocol (Hamdine et al., 2013)

rFSH (Gonal-f; Merck Serono, the Netherlands) was used for ovarian stimulation in a daily fixed dose of 150 IU and administered subcutaneously (s.c.) from CD 2 onward around 18:00 (ranging between 16:00 and 20:00). The timing of rFSH administration was disclosed in the written treatment procedures of our hospital for which verbal informed consent was obtained. Patients were randomized to start GnRH antagonist co-treatment (Cetrotide 0.25 mg s.c; Merck Serono, the Netherlands) either on CD 2 or CD 6. Patients did not receive hormonal pre-treatment (including oral contraceptives) in the cycle preceding stimulation. From stimulation Day 6 onward, follicular growth was assessed by transvaginal ultrasound and repeated every few days depending on the ovarian response. When at least one follicle of ≥18 mm and two follicles of ≥16 mm in diameter were visualized, final oocyte maturation was induced by administrating 6500 IU of hCG s.c. (Ovitrelle; Merck Serono, the Netherlands). Thirty-six hours later oocyte retrieval was performed.

Hormonal assay

Blood was drawn on CD 2 (before start of rFSH and GnRH antagonist) and day of hCG trigger. Timing of both blood withdrawals was logged and ranged from 8:00 a.m. to 1:00 p.m. Therefore, the time between the last rFSH administration and blood withdrawal on day of hCG trigger was estimated to range between 14 and 19 h. Serum samples were stored at −80°C until day of FSH measurement. The serum FSH levels were analysed using our in-house Beckman-Coulter Unicel DXi800 immunoanalyser (Woerden, the Netherlands) chemiluminescence assay. The interassay coefficient of variation (CV) varies from 5.7 to 6.6% for the concentration range of 7–46 IU/L. CV's were determined over a year of serum FSH measurements.

In addition to FSH, basal (measured 0–11 days prior to start of stimulation) anti-Mullerian hormone (AMH) was determined in a sandwich ELISA (AMH Gen II ELISA A79765, Beckman-Coulter; Inc., USA). The lower limit of detection was 0.16 μg/L. The CV varied from 4.7 to 10% for the concentration range of 0.27–3.9 and 4.7 μg/L.

Response assessment

Because no consensus exists on ovarian response definitions in GnRH antagonist co-treated cycles, we used definitions based on GnRH agonist protocols. In accordance with the Bologna criteria (Ferraretti et al., 2011), poor response was defined as the retrieval of less than four oocytes (irrespective of oocyte maturity) or cancellation due to poor ovarian response (less than three dominant follicles of >12 mm). Normal response was defined as the retrieval of 4–15 oocytes and hyper response was defined as the retrieval of more than 15 oocytes or cancellation due to an anticipated risk of OHSS.

Outcome measure

Our primary outcome measure was the serum FSH level on the day of hCG trigger.

Sample size and statistical analysis

Based on previous literature, we expected normal responders to have a serum FSH level on day of hCG trigger of 11 IU/L (Out et al., 2004). A difference in FSH level on the day of hCG trigger of 2 IU/L between poor and normal responders and between hyper and normal responders was considered to be clinically relevant (9 versus 11 IU/L and 13 versus 11 IU/L, respectively). This difference was chosen because serum FSH is expected to rise at least 2 IU/L when increasing the rFSH dose from 150 to 225 IU (van der Meer et al., 1994; Mannaerts et al., 1996; Out et al., 2004).

A total of 64 participants (16 poor, 32 normal and 16 hyper responders) would provide 80% power to detect a difference in serum FSH level on the day of hCG administration of 2 IU/L, assuming a standard deviation of 2 and an alpha of 0.05.

Statistical analysis was performed using IBM SPSS Statistics for Windows (Version 21.0. Armonk, NY: IBM Corp). If continuous variables were normally distributed, they were presented as mean (standard deviation (SD)). If continuous data were not normally distributed, median [range] were presented. For categorical variables, data were presented as number of women and percentage (%). Between-group statistical comparisons of mean values were done using ANOVA tests and for median values the Kruskal–Wallis test was used. Between-group comparisons of categorical variables were performed using Chi-square tests. Post-hoc analyses were done if these tests showed significant differences and corrected for multiple comparisons using Tukey for parametrical tests and Bonferroni for non-parametrical tests.

To analyse whether a difference in serum FSH levels on day of hCG trigger was associated with a difference in number of retrieved oocytes, a Pearson correlation coefficient was calculated. Possible confounding factors based on available literature (i.e. age, body weight, CD 2 FSH and basal AMH) were included in a multivariable linear regression analysis. Finally, a sensitivity analysis was performed comparing outcome between women who started GnRH antagonist co-treatment on CD 2 or CD 6. P < 0.05 was considered to be statistically significant.

Results

Baseline and stimulation parameters

From the 200 women included in the CETRO trial in the UMC Utrecht, 124 women were included in this study (Fig. 1). From these women, 16 were categorized as poor responders (one cancellation of treatment for poor response), 17 as hyper responders and 91 as normal responders. Baseline characteristics for the three response groups are listed in Table I. Age differed significantly between groups. Post-hoc analysis showed that this was due to a difference between poor and hyper responders (32.5 [28–38] versus 28.0 [22–37] years, respectively; P = 0.024) and normal versus hyper responders (33.0 [22–39] versus 28.0 [22–37] years, respectively; P = 0.024). A significant difference was also found in body weight. Hyper responders had a significantly lower body weight as compared to normal responders (60.6 kg [51.0–78.9] versus 68.0 kg [52.0–94.0], P = 0.036) and as compared to poor responders (70.0 kg [55.0–85.6]), although the latter was not statistically significant, probably due to small sample sizes. Women with a poor response were significantly more often current smokers as compared to normal responders (6 (38%) versus 9 (10%), P = 0.009). Basal serum AMH level differed significantly between all three groups (1.05 μg/L [0.5–3.3] versus 2.1 μg/L [0.2–5.7] versus 5.4 μg/L [1.9–9.3], in poor, normal and hyper responders respectively; P < 0.001 for all comparisons). Stimulation characteristics and serum level results for the various parameters are listed in Table II. No significant differences were found for start of downregulation, duration of stimulation or total rFSH used.

Table I

Baseline characteristics of the women by ovarian response.

	Response			P  value
	Poor (N = 16)	Normal (N = 91)	Hyper (N = 17)
Age (years)	32.5 (28–38)	33.0 (22–39)	28.0 (22–37)	0.014a
BMI (kg/m2), mean (SD)	23.1 (3.5)	23.6 (2.7)	22.4 (2.5)	0.268
Weight (kg)	70.0 (55.0–85.6)	68.0 (52.0–94.0)	60.6 (51.0–78.0)	0.043b
Primary subfertility	14 (88%)	64 (70%)	13 (77%)	0.341
Duration of infertility (years)	2.1 (0.3–3.6)	2.5 (0.2–14.8)	2.4 (0.5–5.6)	0.546
Nulliparous women	14 (89%)	70 (77%)	13 (77%)	0.628
Diagnosis				0.169
Tubal	4 (25%)	6 (7%)	2 (11%)
Male factor	9 (56%)	60 (66%)	12 (71%)
Endometriosis	1 (6%)	1 (1%)	0 (NA)
Idiopathic	2 (13%)	24 (26%)	3 (18%)
Length of cycle (days)	28 (21–30)	28 (23–35)	29 (25–31)	0.065
Smoking	6 (38%)	9 (10%)	5 (29%)	0.006c
Consuming alcohol	8 (50%)	45 (50%)	11 (65%)	0.508
Treatment				0.688
IVF	8 (50%)	40 (44%)	6 (35%)
ICSI	8 (50%)	51 (56%)	11 (65%)
Baseline AMH (μg/L)	1.1 (0.5–3.3)	2.1 (0.2–5.7)	5.4 (1.9–9.3)	<0.001d

Data are N (% of response group) or median (range) unless otherwise specified.

AMH = Anti-Mullerian hormone.

^aStatistically significant between poor versus hyper and normal versus hyper responders.

^bStatistically significant between normal versus hyper responders.

^cStatistically significant between poor versus normal responders.

^dStatistically significant difference across all three groups.

Table II

Stimulation characteristics and FSH serum level results in the women by ovarian response.

	Response			P  value
	Poor (N = 16)	Normal (N = 91)	Hyper (N = 17)
Start downregulation				0.117
CD 2	6 (38%)	49 (54%)	5 (29%)
CD 6	10 (63%)	42 (46%)	12 (71%)
Total stimulation (days), median (range)	9.5 (7–16)	9.0 (6–14)	9.0 (6–11)	0.525
Total dose of rFSH (IU), median (range)	1425 (1050–2400)	1350 (900–2100)	1350 (900–1650)	0.525
No. of oocytes retrieveda, median (range)	3.0 (0–3)	8.0 (4–15)	18.0 (16–23)	<0.001b
FSH on CD 2	7.7 (2.8)	7.4 (1.7)	6.1 (1.2)	0.021c
FSH on day of hCG (IU/L)	9.5 (2.4)	10.4 (2.3)	11.5 (2.2)	0.042d
∆ FSH between days (IU/L)	1.8 (3.8)	3.0 (2.6)	5.4 (2.3)	0.001c

Data are N (% of response group) or mean (SD) unless otherwise specified. CD = Cycle day, r = recombinant.

^aBased on women with ovum pick-up; one cancellation due to poor response.

^bStatistically significant difference across all three groups.

^cStatistically significant between poor versus hyper and normal versus hyper responders.

^dStatistically significant between poor versus hyper responders.

Figure 1

Flowchart. UMC = University Medical Centre, IU/L = International units per litre and CD = cycle day.

Serum FSH levels

CD 2 serum FSH levels were significantly lower in hyper responders as compared to poor and normal responders (P = 0.039 and P = 0.026, respectively). The mean serum FSH level on day of hCG trigger was 10.4 IU/L (2.3) for the total study population. For poor, normal and hyper responders the means were 9.5 IU/L (2.4), 10.4 IU/L (2.3) and 11.5 IU/L (2.2), respectively (P = 0.042) (Table II, Fig. 2). Post-hoc analyses revealed significant differences between poor and hyper responders (P = 0.034). Also, the mean rise in serum FSH levels from CD 2 to day of hCG was significantly smaller in poor versus hyper responders (1.8 IU/L (3.8) versus 5.4 IU/L (2.3), P = 0.001) and in normal versus hyper responders (3.0 IU/L (2.6) versus 5.4 IU/L (2.3), P = 0.004). Of note, cases were observed in whom the FSH level on day of HCG was below the level measured on CD 2 (N = 9 poor responders, N = 7 normal responders, ranging from a decrease in serum FSH of 0.04 to 7.1 IU/L).

Figure 2

Boxplot of serum FSH level at the two moments of sampling. CD = cycle day.

Correlation

We found a weak, but significant correlation between serum FSH levels on day of hCG and number of retrieved oocytes (correlation r = 0.203; P = 0.024, Fig. 3). The r² of 0.041 indicates that the variation in oocyte retrieval is only minimally determined by serum FSH levels on day of hCG trigger. After correction for age, body weight, CD 2 FSH and AMH, the regression coefficient for serum FSH on day of hCG and the number of retrieved oocytes changed to 0.151 (P = 0.063, Table III).

Table III

Multivariable linear regression analysis for association with number of retrieved oocytes.

	Standardized regression coefficient for serum FSH on day of hCG	P value
Serum FSH on day of hCG	0.203	0.024a
Serum FSH on day of hCG + age + body weight + serum FSH on CD 2 + AMH	0.151	0.063

CD = Cycle day, AMH = Anti-Mullerian hormone.

^aStatistically significant. The change of the regression coefficient for the serum FSH level on the day of hCG and P value in the multivariable model indicate that after correction for the other factors its association with the number of retrieved oocytes disappears.

Figure 3

Scatter plot serum FSH level on day of hCG trigger and number of retrieved oocytes. Correlation r = 0.203, P = 0.024, r² = 0.041. This plot shows a weak, but significant correlation between serum FSH level on day of hCG and the number of oocytes retrieved. The r² indicates that the number of oocytes retrieved is only minimally determined by the FSH serum level on day of hCG trigger.

Sensitivity analysis

No significant differences were found when comparing CD 2 to CD 6 initiation of GnRH antagonist co-treatment for baseline, stimulation or outcome characteristics (data not shown).

Discussion

This study demonstrates that slightly higher serum FSH levels on day of hCG trigger are present in hyper responders as compared to poor responders receiving a fixed daily rFSH dose of 150 IU. Poor or hyper responders, however, could not be distinguished from the normal responders, based on the day of hCG serum FSH levels. Overall, when performing a multivariable linear regression analysis for the number of retrieved oocytes, ovarian reserve status appeared clearly more important in determining response to stimulation than serum FSH levels. Not many studies have been performed on the direct relationship between serum FSH levels on day of hCG trigger after a fixed daily dose of 150 IU rFSH and ovarian response categories. Out et al. (2004) did show that, in women predicted to have a normal response (based on age, BMI and basal FSH and luteinizing hormone levels) receiving 200 IU rFSH daily, serum FSH levels were higher on day of hCG trigger than in women treated with 150 IU (13.9 versus 10.8 IU/L respectively, stated to be significant, however, a P-value was not reported). The mean number of retrieved oocytes in that study was 10.3 for women in the 150 IU group and 11.9 for women in the 200 IU group (P = 0.051). However, no significant differences were found by Out et al. when comparing mature oocytes, good quality embryos or pregnancy rates for these two treatment regimes. This study therefore shows that a higher gonadotropin dose leads to subsequently higher serum FSH levels, but with only marginal increase in number of oocytes retrieved, in predicted normal responders.

The present study indicates that in a fixed dose stimulation protocol, a poor response may not be due to inadequate serum FSH levels, as serum levels on the day of hCG trigger did not differ between poor and normal responders. A question that rises is whether the ‘problem’ in poor responders lies within the antral follicles themselves and can be explained by a decreased sensitivity- or insensitivity of follicles to FSH. FSH stimulates follicular growth by binding to its receptors localized in the granulosa cells of the follicles (Simoni et al., 1997; Zheng et al., 1996; Casarini et al., 2011). It is thought that genetic variations of these receptors influence serum FSH levels and the degree of ovarian response to stimulation (Perez Mayorga et al., 2000; Grigorova et al., 2008). However, various studies performed on this subject show contradictory results and, if there is a difference in response for specific FSH receptor subtypes, this difference may be very small, and potentially of limited clinical significance (Behre et al., 2005; Simoni and Casarini, 2014). Therefore, variation in FSH receptor genotypes is not likely to be the basis for the wide variation in the number of oocytes retrieved in response to standard FSH dose stimulation.

The present study suggests that increasing the dose in poor responders will not improve ovarian response, as the observed FSH levels produce normal ovarian responses in the majority of patients. We have to be careful drawing this conclusion, as this study was not designed as a dose–response study. By excluding women with a different FSH starting dose than 150 IU daily or with a dose adjustment during the treatment cycle, a possible effect of a dose increase cannot be demonstrated nor ruled out. However, a previously published study has demonstrated that increasing the dose of FSH during stimulation when a poor response is observed, does not reduce the cycle cancellation rate nor increase the number of retrieved oocytes (van Hooff et al., 1993). Creating higher levels of FSH therefore may not be the solution. Indeed, previous studies comparing different starting doses of rFSH in predicted poor responders, underpin our hypothesis that the majority of poor responders in fact do not have sufficient numbers of antral follicles present to produce oocyte numbers within what we currently assume to be the normal range of 8–15 oocytes (Klinkert et al., 2005; Lekamge et al., 2008). Therefore, in patients with low ovarian reserve, the problem of low numbers of FSH sensitive follicles cannot be overcome by administering higher doses of rFSH than 150 IU daily. Our findings confirm this suggestion, as poor responders had significantly lower serum AMH levels prior to treatment, reflecting low antral follicle numbers and at the same time produced the same FSH exposure in serum as normal responders. The slightly higher serum FSH levels on day of hCG in hyper responders, however, could suggest that in these women the ovarian response could be mitigated by lowering the dose of rFSH and thereby improve safety. These suggestions about the significance of adjusting FSH dose in (predicted) poor or hyper responders are also underpinned by a recently published systematic review by van Tilborg et al. (2016) and a randomized controlled trial by Nyboe Andersen et al. (2017), which both showed that in predicted poor responders, individualized FSH dosing does not improve ovarian response or pregnancy rates. In predicted hyper responders, a possible positive effect of individualized dosing on safety was suggested without negatively influencing pregnancy rates.

Serum FSH levels on the day of hCG trigger differed less between the response groups than we had hypothesized when designing the study. In some poor and normal responders, a decrease of serum FSH between CD 2 and day of hCG trigger was observed. This means that serum levels dropped despite administration of exogenous FSH. We also noted a great variability in serum levels on the day of hCG. These findings could have a few possible explanations. First, an inter-individual difference in the bioavailability of rFSH after administration could be present. Studies on the pharmacokinetics and -dynamics of rFSH (for both follitropin alpha and beta) showed that the serum FSH level is not only determined by the dose of exogenous FSH, but also by body weight and route of administration (Mannaerts et al., 1996; Steinkampf et al., 2003). The lack of difference in body weight between poor and normal responders may therefore explain comparable serum FSH levels in those groups at the end of exogenous administration. Hence, lower body weight in hyper responders, could explain why there was no decrease in serum FSH level in those women.

Another explanation for the large variation in observed FSH levels could lie within the timing of serum sampling. The estimated minimum time interval between administration and blood withdrawal was 14 h and the maximum was 19 h, but no control system for timing of injection was present. The maximum concentration of rFSH after repeated administration is reported to be reached after 7.3–10 h (Mannaerts et al., 1996; Olsson et al., 2014). The elimination half-time after repeated administration is reported to be 24–48 h. A kinetic study performing frequent serum sampling in women during repeated FSH administration, showed a rise in serum FSH level of only ~1 IU/L above the steady state level after the last administration (Voortman et al., 2000). Serum levels then reached steady state again approximately 24 h after the last injection. This creates a window of 24 h in which the serum FSH level does not rise more than 1 IU/L above the steady state and it does not drop below the steady state level. If of influence, the difference in timing of blood withdrawal (between 14 and 19 h after the last administration) could only cause a slight, overestimation of serum FSH levels and does not explain the observed variation in serum FSH levels.

A last explanation for the variation in serum FSH levels may be the validity of the assay used for analysing exogenous FSH. If the assay does not pick up both exogenous and endogenous FSH systematically, this will cause unreliable serum FSH levels. Current literature does not provide information on this topic.

A strength of this nested cohort study is that data and blood samples were collected prospectively in the CETRO trial. Therefore, there was no missing data for the included women. Because all blood samples were obtained and stored at one centre (UMC Utrecht), processing of the samples was standardized. Serum FSH levels were measured in one batch of samples to reduce interbatch variability. The present study also has some limitations. Firstly, we used response definitions based on the experience in GnRH agonist co-treated cycles, something that has been debated before (Hamdine et al., 2015). It is thought that, because GnRH antagonists are associated with a slightly lower oocyte yield as compared to GnRH agonists (Al-Inany et al., 2006; Kolibianakis et al., 2006), the cut-off for defining a poor response should be changed to a lower number than four oocytes. We were unable to test this, because our sample size of poor responders became too small to generate a minimal power of 80% if a number of retrieved oocytes of three or less was used. In addition, the Bologna criteria are the best available response classification at this moment. Initially, no distinction was made between women who began GnRH antagonist co-treatment on CD 2 or CD 6. Starting antagonist co-treatment later in the stimulation period could result in higher serum FSH levels during stimulation, for endogenous production of FSH is not yet suppressed in early stimulation. This, in turn, could lead to a higher number of retrieved oocytes. Unfortunately, only serum before and at the end of stimulation was available for our population. However, the CETRO Trial showed no significant difference in number of oocytes between the CD 2 and CD 6 group (Hamdine et al., 2013). Also, Kolibianakis et al. (2003) showed that there was no significant difference in FSH levels during stimulation when comparing a CD 1 to CD 6 start of antagonist treatment. In the present sensitivity analysis no significant difference in mean serum FSH levels on day of hCG trigger was found when comparing the CD 2 to CD 6 group.

In conclusion, the results of the current study show that there is no consistent relationship between ovarian response and serum FSH levels on the day of hCG trigger in a 150 IU fixed dose treatment protocol. This may imply that increasing the dose of rFSH in women who respond poorly, will not lead to a higher oocyte yield. For hyper responders, lowering the dose of rFSH in a subsequent cycle in hyper responders may improve safety of IVF/ICSI treatment. However, these issues should be studied in a larger trial with a true dose comparison design.

Authors’ roles

O.H. performed the original CETRO trial. F.J.M.B, H.L.T. and T.C.v.T. designed the study. E.G.W.M.L. carried out the serum FSH measurements. S.C.O. carried out all necessary data analyses in collaboration with M.J.C.E. and wrote the article. All authors participated in the interpretation of the data, provided significant revisions and read and approved the final version of the article.

Funding

No external funds were obtained for this study.

Conflict of interest

S.C.O, T.C.v.T, O.H., H.L.T., E.G.W.M.L., C.B.L. and M.J.C.E. have nothing to disclose. F.J.M.B. receives monetary compensation: member of the external advisory board for Merck Serono and Ferring, the Netherlands; educational activities for Ferring BV, the Netherlands; consultancy work for Gedeon Richter, Belgium; strategic cooperation with Roche on automated AMH assay development and research cooperation with Ansh Labs.