Predictors of low response to mild ovarian stimulation initiated on cycle day 5 for IVF.
ABSTRACT Milder stimulation protocols are being developed to minimize adverse effects of ovarian stimulation in in vitro fertilization (IVF) programs. A drawback is the possibility of an increased rate of insufficient ovarian response. This study aimed to develop a prognostic model for the prediction of cycle cancellation due to insufficient response to mild stimulation.
A total of 174 IVF patients aged<38 years and with a body mass index (BMI)<28 Kg/m2 were treated with mild ovarian stimulation using a fixed daily dose (150 IU) of recombinant follicle-stimulating hormone (rFSH) from cycle day 5 and GnRH antagonist from the late follicular phase. In women with mono- or bifollicular growth (17%), the cycle was cancelled and the treatment was adjusted in a second treatment cycle by starting rFSH on cycle day 2.
In a multivariable logistic regression analysis, duration of infertility, menstrual cycle length, secondary infertility and BMI were included in the prediction model. The area under the receiver-operating characteristics curve of the model was 0.69. A probability cut-off for cancellation of 0.3 yielded an expected sensitivity of 33% and specificity of 92%. Analysis of ovarian response in the subsequent treatment cycle showed an improved ovarian response and a significant reduction in the cancellation rate.
With the presented model, it is possible to identify patients at risk for cycle cancellation, during mild ovarian stimulation, due to insufficient response. The contributing factors of the model suggest that ovarian aging and BMI are related to insufficient response to mild stimulation.
- SourceAvailable from: oxfordjournals.org[show abstract] [hide abstract]
ABSTRACT: Ovarian stimulation for IVF is known to affect luteal phase function. The endometrium in IVF cycles is thus subject to an altered endocrinological environment and to a possible direct effect of the ovarian stimulation therapy. Factors influencing the endometrial receptivity in such cycles are poorly understood. Studies comparing the endometrium in IVF cycles with natural cycles as controls have shown premature secretory changes in the post-ovulatory and early luteal phase of IVF cycles, followed by a large proportion of dyssynchronous glandular and stromal differentiation in the mid-luteal phase. These findings suggest a profound modification of luteal endometrial development in stimulated cycles. This hypothesis is further supported by the demonstration of a modified endometrial steroid receptor regulation and a profound antiproliferative effect in IVF cycles. The time of maximal endometrial receptivity is defined as the implantation window and is characterized by the expression of various endometrial products, among which pinopodes, integrins and leukaemia inhibitory factor are best described. Premature expression of pinopodes and integrins are in line with the observation of precocious luteal transformation following ovarian stimulation, although the clinical relevance with respect to the establishment of a clinical pregnancy awaits further validation. Studies exploring the endometrium within the cycle of embryo transfer have shown a deleterious effect of severe peri-ovulatory maturation advancement exceeding 3 days, as no clinical pregnancies were obtained in this condition. Further unravelling of molecules involved in the implantation mechanism is needed for a better comprehension of the link between altered endometrial development and receptivity in IVF cycles.Human Reproduction Update 9(6):515-22. · 8.85 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Obesity was defined by a body mass index more than 30 kg/m2. Many risks were related to this pathology, and sometimes, menstrual disorders or infertility. In order to obtain an adequate response to ovarian stimulation during IVF cycles, higher doses of menotropins are necessary in the group of obese patients. The mechanism of this phenomenon is still unclear. Leptin is one of the main hypothesis, and could act on obesity and reproductive system simultaneously. The likelihood to have an ongoing pregnancy after IVF treatment is less in the group of obese patients because of the increased risk of miscarriage and obstetrical complications. Weight loss prior IVF remains the main advice in order to decrease the risks of the procedure and to treat successfully these patients.Contraception, fertilité, sexualité (1992) 26(7-8):564-7.
- [show abstract] [hide abstract]
ABSTRACT: This prospective, randomized trial in normo-ovulatory women was designed to test whether administration of low-dose exogenous FSH initiated during the early, mid to late follicular phase can induce multiple dominant follicle development. Forty normal weight women (age 19–35 years, cycle length 25–32 days) participated. A fixed dose (75 IU/day) of recombinant FSH was started on either cycle day 3 ( n = 13), 5 ( n = 13) or 7 ( n = 14) until the induction of ovulation with human chorionic gonadotrophin. Frequent transvaginal ultrasound scans and blood sampling were performed. Multifollicular growth occurred in all groups (overall in 60%), although day 7 starters showed less multifollicular growth. Age, cycle length and initial FSH and inhibin B concentrations were similar between subjects with single or multiple follicle development. However, for all women the lower the body mass index (BMI), the more follicles emerged ( r = –0.44, P = 0.007). If multifollicular growth occurred, the length of the luteal phase was reduced ( P = 0.002) and midluteal serum concentrations of LH ( P = 0.03) and FSH ( P = 0.004) were decreased and oestradiol ( P = 0.002) and inhibin A ( P = 0.01) were increased. In conclusion, interference with decremental serum FSH concentrations by administration of low dose FSH starting on cycle day 3, 5 or as late as day 7, is capable of disrupting single dominant follicle selection. The role of BMI in determining ovarian response suggests that differences in pharmacokinetics of exogenous FSH are involved. Multifollicular growth per se has a distinct effect on luteal phase characteristics. These observations may be relevant for the design of mild ovarian stimulation protocols.Human Reproduction 06/2001; · 4.67 Impact Factor
Predictors of low response to mild ovarian stimulation
initiated on cycle day 5 for IVF
M.F.G. Verberg1,3, M.J.C. Eijkemans1,2, N.S. Macklon1, E.M.E.W. Heijnen1, B.C.J.M. Fauser1
and F.J. Broekmans1
1Department of Reproductive Medicine and Gynaecology, University Medical Centre Utrecht, Heidelberglaan 100, 3584 CS Utrecht,
The Netherlands;2Department of Public Health, Erasmus MC, University Medical Center, Dr. Molewaterplein 50, PO Box 1738, 3000
DR Rotterdam, The Netherlands
3Correspondence address. Tel: þ31(0)302508373; Fax: þ31(0)302505433; E-mail: firstname.lastname@example.org
BACKGROUND: Milder stimulationprotocolsare being developed tominimize adverseeffects of ovarian stimulation
in in vitro fertilization (IVF) programs. A drawback is the possibility of an increased rate of insufficient ovarian
response. This study aimed to develop a prognostic model for the prediction of cycle cancellation due to insufficient
response to mild stimulation. METHODS: A total of 174 IVF patients aged <38 years and with a body mass index
(BMI) <28Kg/m2were treated with mild ovarian stimulation using a fixed daily dose (150 IU) of recombinant fol-
licle-stimulating hormone (rFSH) from cycle day 5 and GnRH antagonist from the late follicular phase. In women
with mono- orbifolliculargrowth (17%),the cycle was cancelledandthe treatmentwas adjusted ina second treatment
cycleby startingrFSHoncycleday2. RESULTS: Ina multivariablelogisticregressionanalysis,durationofinfertility,
menstrualcyclelength, secondary infertility andBMIwere includedinthe prediction model.Thearea undertherecei-
ver-operating characteristics curve of the model was 0.69. A probability cut-off for cancellation of 0.3 yielded an
expected sensitivity of 33% and specificity of 92%. Analysis of ovarian response in the subsequent treatment cycle
showed an improved ovarian response and a significant reduction in the cancellation rate. CONCLUSIONS: With
the presented model, it is possible to identify patients at risk for cycle cancellation, during mild ovarian stimulation,
due to insufficient response. The contributing factors of the model suggest that ovarian aging and BMI are related to
insufficient response to mild stimulation.
Keywords: mild ovarian stimulation; ovarian response; prediction models; IVF cycle cancellation
As assisted reproductive science progresses, a shift in the focus
of in vitro fertilization (IVF) is occurring from striving for
maximizing instant success ‘at all costs’ to developing safer
and more patient friendly protocols in which the risks of treat-
ment are minimized while optimizing the chance of a singleton
live birth (Edwards et al., 1996; Gerris, 2005). Ovarian stimu-
lation is applied in IVF to generate multiple follicle growth in
order to obtain an increased quantity of oocytes to compensate
for inefficiency of the IVF procedure while maintaining the
potential to select the best embryo (Fleming et al., 1990). Cur-
rently, this goal is usually achieved by a long gonadotrophin-
releasing hormone (GnRH) agonist suppression protocol, in
association with ovarian stimulation with high doses of
exogenous follicle-stimulating hormone (FSH). Disadvantages
of this approach are the high cost, complex stimulation proto-
cols which take several weeks, physical and emotional discom-
uncontrollable degree of ovarian response. The current trend
of limiting the number of embryos to be transferred reduces
the need for large numbers of oocytes. Moreover, there is
increasing evidence of the detrimental effects of ovarian stimu-
lation on corpus luteum function, endometrial receptivity and
embryo quality (Valbuena et al., 2001; Bourgain and
Devroey, 2003). As a consequence, mild ovarian stimulation
protocols are being developed to minimize the adverse treat-
ment effects of ovarian stimulation (Fauser et al., 1999). The
introduction of GnRH antagonists into clinical practice and a
greater understanding of the process of follicle recruitment
and dominant follicle selection have led to new opportunities
for developing mild stimulation protocols.
It has been shown that by interfering with the physiological
decrease in FSH levels during the follicular phase, it is possible
to override the selection of a single dominant follicle (van
Santbrink et al., 1995; Hohmann et al, 2001). Both the
degree and the duration of the FSH elevation will lead to an
extension of the so called ‘FSH window’ that enables the devel-
opment of several rather than just a single dominant follicle
# The Author 2007. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology.
All rights reserved. For Permissions, please email: email@example.com
Human Reproduction Vol.22, No.7 pp. 1919–1924, 2007
Advance Access publication on May 7, 2007
by guest on June 1, 2013
(Zeleznik et al., 1985; Schipper et al., 1998; de Jong et al.,
2000). Indeed a slight but extended elevation of FSH levels
during the mid to late follicular phase has been shown to be suf-
ficient for growth of a modest number of dominant follicles
(Schipper et al., 1998; Hohmann et al., 2001). Yet, there
appears to be an individual variability in the optimal moment
for initiating exogenous FSH supplementation. Hohmann
et al. (2001) observed no difference in multifollicular growth
when daily FSH administration was started on day 3 or 5,
although a tendency toward lower numbers of dominant fol-
licles was seen when cycle day 7 was chosen to initiate FSH
With the use of GnRH antagonists to prevent a premature
luteinizing hormone rise, the IVF treatment cycle can com-
mence at a point in the course of a spontaneous menstrual
cycle where the recruitment of a cohort of antral follicles has
already been established (Fauser and van Heusden, 1997;
Macklon and Fauser, 2000). This approach enables limiting
the use of exogenous FSH in order to extend the FSH
window, allowing multiple dominant follicle development to
take place. As a consequence, the number of treatment days
and the total amount of exogenous FSH required is substan-
tially reduced (de Jong et al., 2000; Hohmann et al., 2003;
Heijnen et al., 2007). A potential drawback of mild stimulation
is a decrease in ovarian response compared with conventional
stimulation, leading to higher cancellation rates (Fauser et al.,
1999; Hohmann et al., 2003). Although low numbers of
oocytes appear to be related to good outcomes in mild stimu-
lation (Hohmann et al., 2003), cancellations should be pre-
vented to optimize the benefit of mild stimulation. The
purpose of this study was the prediction of which mild stimu-
lation cycles are likely to be cancelled due to insufficient
ovarian response. The development of methods to identify
women who may benefit from an earlier start of exogenous
FSH may reduce the number of cancelled cycles and improve
the efficacy of the mild stimulation protocol.
Materials and Methods
Data were derived from the mild stimulation arm of a randomized con-
trolled trial on effectiveness of IVF treatment strategies (Heijnen
et al., 2007). The study was approved by the local ethics review
board of both participating centers. In this study, infertile patients,
with a regular indication for IVF or intracytoplasmic sperm injection
(ICSI) and who attended the Erasmus Medical Centre (Rotterdam, the
Netherlands) or the University Medical Centre Utrecht (Utrecht, The
,38 years of age and had a regular menstrual cycle (25–35 days)
and a body mass index (BMI) between 18–28 kg/m2. Couples who
had been previously treated with IVF were excluded. Study design
and clinical outcomes of the RCT have been reported recently
(Heijnen et al., 2007).
Patients in the mild stimulation arm were treated with a fixed daily
starting dose of 150 IU recombinant FSH (rFSH) (Gonal-Fw: Serono
N.V. Organon, Oss, The Netherlands) s.c., initiated on the fifth cycle
day (CD 5 protocol). The dose of exogenous FSH was not adjusted
during the stimulation. GnRH antagonist (ganirelix, Orgalutranw:
N.V. Organon, 0.25 mg/day; or cetrorelix, Cetrotidew: Serono
Benelux, 0.25 mg/day) was administered s.c. from the day that at
least one follicle attained a diameter ?14 mm (Hohmann et al.,
2003). Human chorionic gonadotrophin (hCG) (Profasiw: Serono
Benelux B.V.; or Pregnylw: N.V. Organon) 10 000 IU s.c. was adminis-
tered as a single bolus injection to induce final oocyte maturation, when
the largest follicle had reached at least 18 mm in diameter and at least
one additional follicle ?15 mm was observed. Oocyte retrieval and fer-
tilization ‘in vitro’ was performed according to standard procedures as
embryo transfer of the resulting best quality embryo was performed on
day 3 or 4 after oocyte retrieval. Standard luteal phase support in the
form of intravaginal progesterone (Progestanw: N.V. Organon)
600 mg/day was given from the day of oocyte retrieval until a urine
pregnancy test was performed 18 days later.
Insufficient ovarian response resulting in cancellation of the cycle
was defined as the development of less than three follicles
.12 mm. In these patients, exogenous FSH was initiated on cycle
day 2 (CD 2 protocol) in a subsequent treatment cycle while the
daily dosage remained unchanged. Cycles at risk for ovarian hypersti-
mulation syndrome (OHSS), defined as more than 20 follicles with a
diameter .10 mm or estradiol concentrations .15.400 pmol/l were
also cancelled before hCG injection.
In order to identify a prior predictors of cancellation cycles due to
insufficient response in a mild stimulation protocol, female age, pre-
vious pregnancy, previous childbirth, cause of infertility, menstrual
cycle length, IVF or ICSI treatment and BMI were compared
between patients where ovarian stimulation was cancelled and patients
who had a sufficient ovarian response to proceed to oocyte pick-up in
their first treatment cycle. For this analysis, cycles cancelled due to
increased risk of OHSS were considered as good responders and
included in the analysis. Cycles cancelled due to premature luteiniza-
tion were excluded from the analysis. Multivariable logistic regression
analysis was performed with a backward elimination procedure, a
P-value ,0.3 was used as a criterion for exclusion. The predictive
ability of the model was assessed by determining the area under the
receiver operating characteristics (ROC) curve (AUC).
To assess the amount of overfitting of the created model, internal
validation was performed with bootstrapping, a statistical technique
to create comparable populations. We bootstrapped 200 times. In
each of these 200 new data sets, the same multivariable logistic
regression analysis with backward elimination was performed, and
the resulting model was tested on the original data. In this way the
amount of overfitting can be assessed and expressed as a shrinkage
factor. The shrinkage factor should be taken into account when apply-
ing the model in clinical practice (Van Houwelingen and le Cessie,
1990; Harrell et al., 1996).
To study whether patients who were cancelled due to insufficient
ovarian response in their first treatment cycle with the CD 5 protocol
presented with improved ovarian response when stimulation in the
subsequent cycle was started at CD 2, a ‘within patient comparison’
was performed. In this analysis, the ovarian response of a second treat-
ment is compared with the response in the first treatment cycle of the
same individual. To compare both cycles, duration of stimulation,
total rFSH needed and number of developed follicles and oocytes of
patients who were cancelled for insufficient ovarian response in
their first treatment cycle were included in the analysis.
Comparisons of outcome measures between the groups were per-
formed using the t-test for continuous data and the x2-test for binary
variables unless stated otherwise. Within patient comparison was
done by a paired t-test. Analyses were performed using SPSS
Verberg et al.
by guest on June 1, 2013
version 12.0 (SPSS Inc., Chicago, IL, USA, 1999) and S-plus 2000
(Insightful Corp., Seattle, WA, USA).
Of the 174 first cycles started, 39 (22%) ended in a cancella-
tion: 30 (17%) due to an insufficient response and 9 (5%) for
other reasons (Fig. 1). Univariable analysis of patient charac-
teristics is shown in Table I. A significantly shorter menstrual
cycle length (the number of days of an average menstrual
cycle in the previous year as indicated by the patient) (28.2
versus 27.5 days; P ¼ 0.045) and longer duration of infertility
(4.4 versus 3.6 years; P ¼ 0.022) were observed in patients
with an insufficient response. The number of treatment days
and medication used were also compared between cancellation
for insufficient ovarian response and patients who did continue
for follicle aspiration. As expected, there was a significant
difference in treatment days and medication used in these
cycles (P-values all ,0.001) (Table II).
In the multivariable analysis, the variables, duration of infer-
tility, menstrual cycle length, primary or secondary infertility
and BMI, were selected into the prediction model for cancella-
tion during the mild stimulation protocol. A longer duration of
infertility, short menstrual cycle length, secondary infertility
and higher BMI were found to be associated with an insuffi-
cient ovarian response. The predictive ability of the model
measured by the area under the ROC curve was 0.69 (95% con-
fidence interval [CI] 0.58–0.79) (Table III). Internal validation
by bootstrapping showed a shrinkage factor of 0.58. This
means that the final model is overfitted to our data and the pre-
dictive probabilities of the model on external data will be less.
The model will give the most reliable predictions when all
regression coefficients are on average 42% smaller in absolute
size (Table III). A probability cut-off for cancellation due to
insufficient response of 0.3 yielded an expected sensitivity of
33%, specificity of 92%, and positive predictive value of
48% on our own data. A probability cut-off of 0.15 yielded
an expected sensitivity of 77%, specificity of 54% and positive
predictive value of 26%. Table IV shows the validity of the
model with a range of cut-of values chosen in the area of the
AUC curve of the model with the most discriminative power.
As expected, the within patient comparison between the first
and second treatment cycles of patients who were cancelled for
insufficient response in their first treatment cycle showed that a
change in treatment protocol from CD 5 to CD 2 resulted in a
clear improvement in ovarian response. Of the 27 patients
whose first treatment cycle was cancelled and who proceeded
with a changed protocol in the second cycle, 3 (11%) had to
be cancelled again.
The aim of the present study was to explore whether it is poss-
ible to identify a subgroup of patients at risk for cancellation
due to insufficient ovarian response in mild ovarian stimulation
for IVF starting exogenous FSH on cycle day 5. Our study con-
firmed the previous finding ofa relatively high cancellation rate
(17%) in the mild stimulation protocol (Hohmann et al., 2003).
To increase the benefit of mild stimulation we analysed charac-
teristics of patients with an insufficient ovarian response.
According to pre-treatment variables we developed a prognos-
tic model to identify patients at risk for cancellation prior to the
start of ovarian stimulation. The predictors in the model were a
longer duration of infertility, a shorter menstrual cycle length,
secondary infertility and a higher BMI.
The finding of an association between a shorter cycle length
and insufficient ovarian response may be the consequence of a
shortened follicular phase in these patients where dominant fol-
licle selection may have occurred prior to the start of the
exogenous FSH (Klein et al., 2002; van Zonneveld et al.,
2003). As a consequence, involvement of the non-dominant
follicles in ongoing growth by exogenous FSH becomes
impossible. Also, cycle shortening may be a subtle first sign
Figure 1. Flow chart showing the number of patients divided according to the degree of ovarian response in the mild CD 5 stimulation protocol,
and follow up in the second treatment cycle
†Premature luteinization was defined as a LH serum level above 15, which was measured on clinical suspicion;
*1 spontaneous pregnancy, 1 pregnancy after escape intrauterine insemination
Predictors of ovarian response to mild stimulation
by guest on June 1, 2013
of advanced ovarian aging and as such be related to a small
cohort size. As only patients with a regular cycle length and
below 38 years of age were included for this study, this
phenomenon may become even more evident in the general
A negative association between BMI and ovarian response
has been observed previously (Wittemer et al., 2000; Nichols
et al., 2003; Fedorcsak et al., 2004). Obese patients usually
require significantly higher doses of gonadotrophin and a
longer duration of stimulation (Dechaud et al., 1998).
However, in a prospective study comparing predictive factors
of ovarian response in IVF, BMI did not associate with the
number of follicles or the number of retrieved oocytes
(Popovic-Todorovic et al., 2003).
Longer duration of infertility has previously been recognized
as an important negative prognostic factor affecting the chance
of natural conception, particularly in unexplained infertility
(Hull et al., 1985). This might be the consequence of subtle,
undiagnosed disorders related to a diminished ovarian
reserve. These disorders might normally be overcome by
fierce stimulation in more conventional stimulation protocols,
but are noticed in natural conception and mild stimulation.
The observed association between ovarian response and sec-
ondary infertility has not been described previously. In our
opinion, there is no straight-forward biological explanation
for this relationship. Its appearance in the model could be the
result of the relatively small sample size used to build the
model and might explain the relatively large shrinkage factor
that was found in the internal validity analysis.
The only previous study regarding the current mild stimu-
lation protocol with a similar analysis showed a significant
difference in age and baseline FSH between cancelled patients
and those who met the criteria for oocyte retrieval (Hohmann
et al., 2003). The current study could not confirm the finding
of female age as an important predictor in spite of the larger
sample size. This might be partly the result of the age restric-
tion in the inclusion criteria for the study. Additionally, both
the shorter cycle length and the longer duration of infertility
appearing in the current model might be related to ovarian
aging resulting in age being removed in multivariable analysis.
Unfortunately data from measurement of baseline FSH or other
hormonal markers related to ovarian aging such as Inhibin B or
anti-Mu ¨llerian hormone were not available.
Table I: Univariable analysis of patient characteristics
Cancellation due to
response (n ¼ 30)
response (n ¼ 140)a
Difference (95% CI)
Female age (years)
Primary infertility (%)
Cause of infertility (%)
Duration of infertility (years)
Cycle length (days)
0.45 (21.68, 0.77)
0.85 (21.57, 20.13)*
0.74 (0.15, 1.47)*
0.46 (21.49, 0.58)
Values are means+SD.
*Differences are statistically significant (P-value ,0.05).
aIncluding five cycles that were cancelled as a consequence of increased risk for OHSS.
Table II: Univariable analysis of cycle characteristics
Cancellation due to
responsea(n ¼ 30)
response (n ¼ 135)
Difference (95% CI)
Duration of stimulation (days)
Total rFSH used (IU)
GnRH antagonist administered (days)
2.47 (1.84, 3.01)*
370 (268, 472)*
2.29 (1.72, 2.84)*
Values are means (+SD).
*Differences are statistically significant (P-value ,0.05).
aInsufficient ovarian response was defined as the development of less than three dominant follicles (diameter .12 mm).
Table III: Multivariable analysis for cancellations due to poor response in
the mild CD 5 stimulation protocol; the ability of the model measured by the
area under the ROC curve was 0.69 (95% CI: 0.58–0.79)
P-value Odds Ratio
Duration of infertility
1.24 (1.02, 1.50)
0.75 (0.59, 0.98)
2.08 (0.82, 5.27)
1.10 (0.93, 1.29)
aPredicted probability of cancellation due to insufficient response: 1/(1 þ
Exp[2(0.608 þ 0.126 duration 20.158 [cycle length] þ0.423 [secondary
infertility (yes ¼ 1, no ¼ 0)] þ 0.055*BMI)]).
Verberg et al.
by guest on June 1, 2013
Analysis of ovarian response in a subsequent treatment cycle
with the CD 2 protocol for patients who previously presented
with an insufficient ovarian response showed that these were
likely to meet the criteria for oocyte retrieval. These results
support the hypothesis that in a mild stimulation protocol,
insufficient ovarian response is a consequence of suboptimal
ovarian stimulation for a specific group of patients and can
be overcome by the early commencement of exogenous stimu-
lation. However, it is likely that at least part of the improved
ovarian response is the result of the principle of regression
toward the mean. This is a principle stating that of related
measurements, and selecting those where the first measurement
is either higher or lower than the average, the expected value of
the second is closer to the mean than the observed value of the
first (Davis, 1976). Prospective randomized studies are needed
to establish which part of the improved ovarian response
should be ascribed to the change in the hormonal stimulation
These data indicate that the CD 2 protocol is likely to
improve chances for patients who would be cancelled in the
CD 5 protocol due to insufficient ovarian response. Because
the prediction model is based on a priori parameters, patients
at risk for cancellation can be identified prior to the start of
the treatment. The overall cancellation rate for insufficient
response is therefore likely to be reduced if these patients are
treated with the CD 2 protocol instead of the CD 5 protocol.
Although the area under the ROC curve was modest, the
model has, in our opinion, the ability to predict sufficient
numbers of cancellations to reduce the remaining number of
cancellations to an acceptable level. At a cut-off level of
0.30, the model will predict 10 cancellations correctly (sensi-
tivity 33% (95% CI 16.5–50.2)). According to our CD
2-started second cycles, 1 out of these 10 would still be can-
celled (11%). This means that with a model based treatment
adaptation the percentage cancels would be lowered from 17
to 12% (21/174). This reduction would equal the proportional
difference in cancellation rates between the days 2 and 5 arm of
the study by Hohmann et al. (2003), and as such may be con-
sidered clinically useful. Due to the high specificity at the
cut-off level (92% (95% CI 87.7–96.6)), the number of
patients that will be unnecessarily treated with the CD 2 proto-
col is limited, so that the majority of cases can therefore still
benefit from the advantages of the mild stimulation day 5 pro-
tocol. Still, one should be aware that the model is developed on
a modest number of patients leading to the relative broad con-
fidence interval for sensitivity (16.5–50.2%).
Overall, the clinical use of the prediction model may render
mild ovarian stimulation more patient tailored. The concept of
‘cycle cancellation’ should be viewed in the context of just a
few days of medication in the current mild stimulation
approach compared with the conventional approach. In such
stimulation protocols (GnRH agonist long protocol, sometimes
proceeded by oral contraceptives and followed by extended
ovarian stimulation), the time elapsed before an insufficient
response can be identified is much longer, resulting in an
increased waste of medication and the delay of at least two
menstrual cycles. The cancellation rate during mild stimulation
should also be balanced against the extended gain (for multiple
cycles, if needed) of a later start of stimulation in the great
majority of women resulting in fewer injections and reduced
patient discomfort and cost.
In conclusion, midfollicular initiation of rFSH in combi-
nation with a GnRH antagonist leads to a mild ovarian stimu-
lation protocol, but yields a distinct risk of cancellation due
to insufficient response. We developed a model that could
predict 33% of the cancellations (sensitivity) with a false posi-
tive rate of 8% on our own data and so equalize the cancellation
rate for insufficient response to that observed in a standard
GnRH antagonist ovarian stimulation protocol for IVF. Insuffi-
cient response in mild stimulation may be related to ovarian
aging and increased BMI. After external validation, the
model might be used to identify patients at risk for insufficient
ovarian response prior to the start of the treatment cycle. Treat-
ment with ovarian stimulation initiated in the early follicular
phase in these patients may reduce the number of cancelled
cycles and therefore improve the efficacy of the mild stimu-
lation protocol. Further prospective randomized trials should
evaluate the clinical use of adjusting the starting day of
ovarian stimulation based on the current prediction model.
This study was funded by ZonMw (The Netherlands), program Doel-
matigheidsonderzoek (grant number 945-12-010).
Bourgain C, Devroey P. The endometrium in stimulated cycles for IVF. Hum
Reprod Update 2003;9:515–522.
Davis CE. The effect of regression to the mean in epidemiologic and clinical
studies. Am J Epidemiol 1976;104:493–498.
Dechaud H, Ferron G, Anahory T et al. Obesity and assisted reproduction
techniques. Contracept Fertil Sex 1998;26:564–567.
Table IV: Clinical value of the model for cancel prediction with test characteristics at several probability cut-offs
Cut-off value for the probability of cancel0.10 0.15 0.200.25 0.30
% of patients that will change protocol
Number of cancels unpredicted (n (%))
20 (67%) 4 (13%)
Values are percentages unless stated otherwise.
PPV, positive predictive value; NPV, negative predictive value.
Predictors of ovarian response to mild stimulation
by guest on June 1, 2013
Edwards RG, Lobo R, Bouchard P. Time to revolutionize ovarian stimulation.
Hum Reprod 1996;11:917–919.
Fauser BC, Devroey P, Yen SS et al. Minimal ovarian stimulation for IVF:
appraisal of potentialbenefits
Fauser BC, van Heusden AM. Manipulation of human ovarian function:
Fedorcsak P, Dale PO, Storeng R et al. Impact of overweight and underweight
on assisted reproduction treatment. Hum Reprod 2004;19:2523–2528.
Fleming A, Jamieson ME, Coutts JRT. The use of GnRH-analogs in assisted
reproduction. In: Matson PL, Lieberman BA (eds). Clinical IVF Forum.
Manchester: Manchester University Press, 1990, pp. 1–19.
Gerris JM. Single embryo transfer and IVF/ICSI outcome: a balanced
appraisal. Hum Reprod Update 2005;11:105–121.
Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in
developing models, evaluating assumptions and adequacy, and measuring
and reducing errors. Stat Med 1996;15:361–387.
Heijnen EM, Eijkemans MJ, de Klerk C et al. A mild treatment strategy for in-
vitro fertilisation: a randomised non-inferiority trial randomized trial. Lancet
Hohmann FP, Laven JS, de Jong FH et al. Low-dose exogenous FSH initiated
during the early, mid or late follicular phase can induce multiple dominant
follicle development. Hum Reprod 2001;16:846–854.
Hohmann FP, Macklon NS, Fauser BC. A randomized comparison of two
(GnRH) antagonist cotreatment for in vitro fertilization commencing
recombinant follicle-stimulating hormone on cycle day 2 or 5 with the
standard long GnRH agonist protocol. J Clin Endocrinol Metab
Huisman GJ, Fauser BC, Eijkemans MJ et al. Implantation rates after in vitro
fertilization and transfer of a maximum of two embryos that have undergone
three to five days of culture. Fertil Steril 2000;73:117–122.
Hull MG, Glazener CM, Kelly NJ et al. Population study of causes,
treatment, and outcome of infertility. Brit Med J (Clin Res Ed)
de Jong D, Macklon NS, Fauser BC. A pilot study involving minimal ovarian
stimulation for in vitro fertilization: extending the “follicle-stimulating
hormone window” combined with the gonadotropin-releasing hormone
antagonist cetrorelix. Fertil Steril 2000;73:1051–1054.
Kastrop PM, Weima SM, Van Kooij RJ et al. Comparison between
intracytoplasmic sperm injection and in-vitro fertilization (IVF) with high
insemination concentration after total fertilization failure in a previous
IVF attempt. Hum Reprod 1999;14:65–69.
Klein NA, Harper AJ, Houmard BS et al. Is the short follicular phase in older
women secondary to advanced
development? J Clin Endocrinol Metab 2002;87:5746–5750.
Macklon NS, Fauser BC. Regulation of follicle development and novel
approaches to ovarian stimulation for IVF. Hum Reprod Update
Nichols JE, Crane MM, Higdon HL et al. Extremes of body mass index reduce
in vitro fertilization pregnancy rates. Fertil Steril 2003;79:645–647.
Popovic-Todorovic B, Loft A, Bredkjaeer HE et al. A prospective randomized
clinical trial comparing an individual dose of recombinant FSH based
on predictive factors versus a ‘standard’ dose of 150 IU/day in
‘standard’ patients undergoing IVF/ICSI treatment. Hum Reprod 2003;18:
van Santbrink EJ, Donderwinkel PF, van Dessel TJ et al. Gonadotrophin
induction of ovulation using a step-down dose regimen: single-centre
clinical experience in 82 patients. Hum Reprod 1995;10:1048–1053.
Schipper I, Hop WC, Fauser BC. The follicle-stimulating hormone (FSH)
threshold/window concept examined by different interventions with
exogenous FSH during the follicular phase of the normal menstrual cycle:
duration, rather than magnitude, of FSH increase affects follicle
development. J Clin Endocrinol Metab 1998;83:1292–1298.
Valbuena D, Martin J, de Pablo JL et al. Increasing levels of estradiol are
deleterious to embryonic implantation because they directly affect the
embryo. Fertil Steril 2001;76:962–968.
Van Houwelingen JC, le Cessie S. Predictive value of statistical models. Stat
Wittemer C, Ohl J, Bailly M et al. Does body mass index of infertile women
have an impact on IVF procedure and outcome? J Assist Reprod Genet
ZeleznikAJ, Hutchison JS,Schuler
gonadotropin-suppressing actions of estradiol in macaques overrides the
selection of a single preovulatory follicle. Endocrinology 1985;117:991–
van Zonneveld P, Scheffer GJ, Broekmans FJ et al. Do cycle disturbances
explain the age-related decline of female fertility? Cycle characteristics of
women aged over 40 years compared with a reference population of
young women. Hum Reprod 2003;18:495–501.
Submitted on November 30, 2006; resubmitted on February 26, 2007; accepted
on March 5, 2007
Verberg et al.
by guest on June 1, 2013