Human Reproduction vol.14 no.9 pp.

2207–2215, 1999

Recombinant versus urinary follicle stimulating

hormone for ovarian stimulation in assisted
reproduction

Salim Daya1,2,3 and Joanne Gunby1
1Obstetrics 2Clinical
Departments of and Gynaecology and
Epidemiology and Biostatistics, McMaster University, Hamilton,
Ontario, Canada
3To
whom correspondence should be addressed at: Department of
Obstetrics and Gynaecology, McMaster University, 1200 Main
Street West, Hamilton, Ontario, Canada L8N 3Z5
The recent availability of recombinant follicle stimulating
hormone (rFSH), with its high level of purity and batch-
to-batch consistency has made it an attractive alternative
to urinary FSH (uFSH) for ovarian stimulation. Several
trials have compared the two preparations, but none had
sufficient power to detect a clinically meaningful difference
in pregnancy rates. The purpose of this study was to
determine the clinical pregnancy rates per started cycle
by pooling data from randomized trials which compared
the use of rFSH and uFSH in treatment cycles using in-vitro
fertilization (IVF) or intracytoplasmic sperm injection
(ICSI). A thorough search of the literature identified 12
trials which met the inclusion criteria. In four trials, both
IVF and ICSI were performed, in seven trials only IVF
was performed and in one trial only ICSI was performed.
Data were extracted and pooled using the principles of
meta-analysis. There was no significant heterogeneity of
treatment effect across the trials. The common odds ratio
and the risk difference (and their 95% confidence intervals),
obtained by pooling the data using a fixed effects model,
were 1.20 (1.02–1.42) and 3.7% (0.5–6.9%) respectively,
in favour of rFSH. The pregnancy rate with the alpha
preparation of rFSH was statistically significantly higher
than with uFSH in IVF cycles. The overall conclusion from
this meta-analysis is that the use of rFSH in assisted
reproduction is preferred over uFSH.
Key words: assisted reproductive technology/IVF/meta-
analysis/recombinant FSH
Introduction
The strategy of stimulating the ovaries with exogenous
gonadotrophins to induce multifollicular development in
women undergoing therapy with assisted reproduction tech-
niques such as in-vitro fertilization (IVF) and intracytoplasmic
sperm injection (ICSI) is now well established. The role of
follicle stimulating hormone (FSH) in this process is essential,
whereas luteinizing hormone (LH) plays a relatively minor
role. In fact, too much LH during the period of follicular
development and in the periovulatory phase may have detri-
© European Society of Human Reproduction and Embryology
mental effects on reproductive outcome (Stanger and Yovich,
1985; Howles et al., 1987; Homburg et al., 1988; Regan
et al., 1990).
Human FSH is a glycoprotein hormone consisting of two
non-co-valently linked, non-identical protein chains (α and
β subunits) to each of which are attached two complex
carbohydrate structures. Endogenously secreted FSH consists
of a family of isoforms with identical primary structures, but
with variable composition of the carbohydrate side-chains and
sialic acid residues (Chappel, 1995). This heterogeneity is
responsible for differences in plasma half-life and consequent
biological activity of the isoforms (Chappel, 1995).
Until recently, the main source of exogenous FSH was urine
of postmenopausal women. Menotrophin [human menopausal
gonadotrophin (HMG)], which consists of a mixture of FSH,
LH and urinary proteins, was the first such preparation. Two
decades later, a further refinement in the extraction process
permitted the removal of LH by absorbing it onto an antibody
column. The resulting urofollitrophin, although essentially
devoid of LH activity, still contained urinary protein
contaminants. Over the last decade, further advances in puri-
fication techniques, using monoclonal antibodies, enabled the
extraction of FSH from urine to produce highly purified (HP)
urofollitrophin from which all of the contaminant proteins and
LH had been removed leaving only FSH activity. Recently,
biotechnology has made available a recombinant preparation
of FSH (rFSH), produced by inserting the genes encoding for
α and β subunits of FSH into expression vectors that are
transfected into a Chinese hamster ovary cell line (Howles,
1996). The use of mammalian cells for this purpose is necessary
because glycosylation is required to ensure full biological
activity of the protein (Howles, 1996). There are two rFSH
preparations currently available for clinical use: follitrophin
alpha, marketed as Gonal-F® by Ares-Serono, Geneva,
Switzerland, and follitrophin beta, marketed as Puregon® or
Follistim® by NV Organon, Oss, The Netherlands. Although
both preparations have been developed using the same tech-
nique, the post-translation glycosylation process and purifica-
tion procedures are not identical (Olijve et al., 1996). The
purification procedure used for follitrophin alpha includes the
use of immunochromatographic methods, whereas purification
of follitrophin beta does not involve immunological methods
(Olijve et al., 1996).
The purity and batch-to-batch consistency of rFSH makes
it an attractive alternative to urinary FSH (uFSH). In a
combined analysis of three randomized trials, rFSH was
observed to be associated with a significantly higher pregnancy
rate in IVF treatment cycles compared to urinary gonado-
trophins (Out et al., 1997). However, in this study, the trials
2207
S.Daya and J.Gunby
selected were limited to those in which follitrophin beta was
used. Furthermore, in one of the trials, rFSH use was compared
with HMG (Jansen and Van Os, 1996; Out et al., 1996). The
observation that HMG is not as efficacious as FSH (Daya
et al., 1995a,b; Daya, 1998) suggests that the inclusion of this
trial is likely to have biased the overall conclusion of the
combined analysis. Therefore, the purpose of this study was
to review the evidence from all randomized trials comparing
rFSH (both follitrophin alpha and follitrophin beta) with
uFSH to evaluate the relative efficacy with respect to clinical
pregnancy rates in treatment cycles with IVF or ICSI.
Materials and methods
Identification of trials
Trials were identified using several search strategies. The Medline
data base of the National Library of Medicine covering the period
from 1990 to 1999 was searched on-line using medical subject
(MeSH) headings ‘pregnancy’, ‘gonadotrophin’, and ‘fertilization in
vitro’, and publication type ‘randomized controlled trial’. The search
was performed on titles, abstracts and key words of the listed articles.
The Excerpta Medica CD: Fertility data base, which contains
publications in human reproduction and all publications in the
Excerpta Medica Data base (Embase) related to obstetrics and
gynaecology, was searched, covering the period 1985 to October
1998. The search was performed on titles and abstracts using
‘recombinant FSH’, ‘IVF’ and ‘randomized’ as keywords.
The bibliographies of relevant publications and review articles
were scanned, and abstracts of major scientific meetings from 1992
to 1999 were hand-searched. The Cochrane Menstrual Disorders and
Subfertility Specialized Register was also searched. When necessary,
authors of relevant abstracts were contacted for detailed data on their
studies. Peer consultation was sought for any remaining articles.
Finally, the pharmaceutical companies that manufacture the gonado-
tropin preparations were consulted for additional information.
Reports of clinical trials were selected only if they met the inclusion
criteria and if the outcome information was provided in sufficient
detail to enable the data to be pooled.
Study inclusion
This systematic review was limited to trials reporting random alloca-
tion to rFSH (either follitrophin alpha or follitrophin beta) or uFSH
(either urofollitrophin or urofollitrophin HP) for ovarian stimulation
in infertile women undergoing treatment with IVF or ICSI. Trials
were included whether or not the stimulation protocol included
pituitary down-regulation with gonadotrophin releasing hormone
agonists (GnRHa). The primary outcome of interest was the clinical
pregnancy rate (usually defined as a gestational sac seen by ultrasono-
graphy), which was calculated per treatment cycle commenced.
Validity assessment and data extraction
For many of the trials, additional information concerning the methods
of the trial and the outcome data was obtained from the authors.
The methodological quality of each trial was assessed using a
predetermined scoring system consisting of eight criteria as shown
in Table I. Each trial was assessed independently by two reviewers
and ranked for its methodological rigor and its potential to introduce
bias. The evaluation included how the randomization procedure was
undertaken and whether it was concealed, the use of blinding, the
presence of cointervention, the completeness of follow-up of trial
subjects, whether a sample size calculation had been performed,
2208
whether a crossover design was used, in which case only data from
the first period (i.e. before crossover) were admissible, and whether the
unit of comparison was patient or treatment cycle. Any disagreement
between the two reviewers was resolved by consensus whenever
possible. In the event of persistent disagreement, a third reviewer
was consulted. Data were extracted and checked for accuracy in a
second review.
Statistical analysis
The data on the outcome for each trial selected for inclusion in the
analysis were extracted into two-by-two tables and summarized using
the odds ratio (OR) and the risk difference. The OR was chosen
because its mathematical properties allow for ease in combining data
to provide an overall estimate of the effect size and in testing for
statistical significance. The risk difference, which is a measure of the
absolute treatment effect, was chosen because it is more easily
understood by clinicians and more helpful for decision making when
applying the results to clinical practice. It is calculated as a weighted
(for sample size) mean difference in pregnancy rates between rFSH
and uFSH. Statistical significance was established if, with a two-
tailed test, P ø 0.05.
Effectiveness was evaluated using the Peto modification of the
Mantel-Haenszel method (Mantel and Haenszel, 1959; Peto, 1987),
which is a test of overall association across all trials. A test of the
homogeneity of treatment effect across all trials was performed
(Breslow and Day, 1980). A nonsignificant result (i.e. lack of
heterogeneity) indicates that no trial has an OR that is statistically
significantly worse or better than the overall common OR obtained
by pooling the data. Only when homogeneity of treatment effect was
confirmed were the data pooled using the fixed effects model,
otherwise the random effects model was used.
A funnel plot (in which the effect estimate of each trial was plotted
against the precision of the effect, calculated as the inverse of its
standard error) was used to detect publication bias. The value of the
funnel plot is based on the fact that precision in estimating the
underlying treatment effect will increase as the sample size of the trial
increases (Egger et al., 1997). Thus, results from small trials will
have a wide scatter at the bottom of the scatter plot, the spread
decreasing as the trials become larger. A symmetrically inverted
funnel shape to the scatter plot indicates that publication bias
is unlikely.
Subgroup analyses were performed to identify whether the two
types of follitrophin (i.e. alpha or beta), and the two types of
fertilization procedure (i.e. ICSI or IVF) had any effect on the overall
combined result. The data were recoded according to these variables
and subjected to logistic regression analysis to identify the model
that best predicted clinical pregnancy per started cycle.
Results
Trials included
Out of 17 trials identified, 12 met the selection criteria for this
systematic review and were included in the analysis (O’Dea
et al., 1993; Alvino et al., 1995; Hedon et al., 1995; Out et al.,
1995; RHFSG, 1995; Bergh et al., 1997; Manassiev et al.,
1997; Andersen et al., 1998; Frydman et al., 1998; Franco
et al., 1999; Lenton et al., unpublished results; Schats et al.,
unpublished results). The reasons for excluding the other five
trials were as follows: no pregnancy outcome data were
available in one of the trials (Fisch et al., 1995) despite several
requests for the missing information from the authors; in one
 Recombinant versus urinary FSH

Table I. Validity criteria used in assessing the methodological rigour of relevant trials

Category Score Method

Randomization procedure 4 Randomization within blocks

Concealment of randomization
Blinding
Co-intervention
Completeness of follow-up
Sample size calculation
Cross-over design
Patients and cycles differentiated
3 Randomization by computer tables
2 Randomization using coin-tossing
1 Randomization method other than above or not stated
3 Central allocation (e.g. by telephone or pharmacy)
2 Allocation locally using concealed method (e.g. envelope,
opaque discs covering the allocation sequence, etc.)
1 No concealment or method not stated
3 Triple
2 Double
1 Single
0 None or not stated
2 No co-intervention
1 Co-intervention unlikely or not stated
0 Co-intervention present
2 All subjects accounted for at completion of study
1 ,5% of subjects lost to follow-up or missing data
0 ù5% of subjects lost to follow-up or missing data
2 Explicitly stated
1 Not done or not stated
2 Cross-over design not used
1 Cross-over design used but data from first period available
0 Cross-over design used
3 Included only first treatment cycles
2 Each subject contributed data from one cycle only
1 Subjects may have contributed data from more than one cycle
0 Number of subjects and cycles not differentiated

Best possible score 5 21; worst possible score 5 3.

trial (Mantzavinos et al., 1998) it was determined, after
contacting the authors, that the patients were not randomly
allocated to treatment, as had been indicated in the abstract;
two trials (Sunde et al., 1994; Mitchell et al., 1996) were
excluded because they were part of a larger multicentre
randomized trial (Out et al., 1995) which had already been
selected for the present analysis; and one trial (Out et al.,
1997) was a report of a combined analysis of three randomized
trials, two of which have been included in this study (Hedon
et al., 1995; Out et al., 1995) whereas the third was a
comparison of rFSH with hMG. In four of the 12 included
trials, additional data were available from the authors so that
the outcomes for IVF and ICSI could be considered separately
(Bergh et al., 1997; Frydman et al., 1998; Lenton et al.,
unpublished results; Schats et al., unpublished results). In one
of these trials (Schats et al., unpublished results), a few patients
underwent both IVF and ICSI treatment in the same cycle,
and these were included as ICSI cycles. Thus, data from
a total of 16 comparisons from 12 trials were available
for analysis.
The validity scores for methodological rigour of the 12 trials
included in the analysis are shown in Table II and indicate
that most trials are of moderate to high grading based on
the predetermined validity criteria used. The methodological
details of these trials are listed in Table III. Apart from
minor differences, the patient profiles were quite similar and
representative of the infertile population requesting treatment,
and the interventions used conform to the currently accepted
standards of care. All trials used GnRHa in a long protocol as
part of the ovarian stimulation regimen.
Meta-analysis
Overall analysis
The overall meta-analysis included a total of 2875 cases, of
which 1556 were allocated to rFSH and 1319 to uFSH. It can
be seen from the OR tree of the trials, shown in Figure 1, that
the direction of the estimate of the treatment effect in all
except two of the trials was in favour of rFSH. The OR for
the 12 trials ranged from 0.83 to 2.27, but no individual
estimate of treatment effect was statistically significant. There
was no significant heterogeneity of treatment effect across all
trials (Breslow–Day statistic 5 7.5, P 5 0.94). The common
OR for clinical pregnancy per started cycle, obtained by
pooling the data using a fixed-effects model, was 1.20 [95%
confidence interval (CI), 1.02–1.42, P 5 0.03] in favour of
rFSH. The risk difference represented a 3.7% (95% CI,
0.5–6.9%) increase in clinical pregnancy rate per cycle started
with rFSH, compared with uFSH.
Subgroup analyses
After separating the data based on the type of fertilization
procedure performed (i.e. IVF or ICSI) and the type of
follitrophin administered (i.e. alpha or beta) several subgroup
analyses were undertaken as shown below.
A funnel plot of the ORs for clinical pregnancy per cycle
started, shown in Figure 2, demonstrates that the data are
distributed along a symmetrical, inverted-funnel shape, indicat-
2209
S.Daya and J.Gunby

Table II. Validity score for each trial selected

Trial Randomization Concealment Blinding Co- Follow-up Sample Cross- Patients/ Total % of max. Rank
intervention over cycle score (21) order

Alvino et al. (1995) 3 1 0 2 2 1 2 2 13 62 3

Andersen et al. (1998) 1 1 0 2 1 2 2 2 10 48 6
Bergh et al. (1997) 3 2 1 2 2 2 2 2 16 76 2
Franco et al. (1999) 3 1 0 2 2 1 2 2 13 62 3
Frydman et al. (1998) 3 2 2 2 2 1 2 2 16 76 2
Hedon et al. (1995) 1 1 1 2 1 1 2 2 11 52 5
Lenton et al. (unpublisheda) 4 2 0 2 2 2 2 3 17 81 1
Manassiev et al. (1997) 1 1 0 2 2 1 2 2 11 52 5
O’Dea et al. (1993) 1 1 0 2 2 1 2 2 11 52 5
Out et al. (1995) 1 1 1 2 1 2 2 2 12 57 4
RHFSG (1995) 3 2 0 2 1 1 2 2 13 62 3
Schats et al. (unpublisheda) 4 2 1 2 2 2 2 2 17 81 1
aData obtained from Dr C.Howles, Ares–Serono, Geneva, Switzerland (see Acknowledgements).

Figure 3. Common odds ratios in subgroup analyses of the

Figure 1. Odds ratio tree of clinical pregnancy per started cycle in
12 trials, comparing recombinant (rFSH) with urinary (uFSH)
follicle stimulating hormone. Breslow and Day test for
homogeneity of treatment effect 5 7.5, P 5 0.94. aData obtained
from Dr C.Howles, Ares–Serono, Geneva, Switzerland (see
Acknowledgements).
Figure 2. Funnel plot of odds ratios for clinical pregnancy in 16
comparisons of rFSH with uFSH.
2210
different types of recombinant follitrophin and fertilization
procedure used. rFSH 5 recombinant follicle stimulating hormone;
uFSH 5 urinary FSH; IVF 5 in-vitro fertilization;
ICSI 5 intracytoplasmic sperm injection.
ing that publication bias was unlikely to be present. Although
the median sample size of the trials was 160, the range
extended from 55 to 981, with the number in all except two
trials (Out et al., 1995; Schats et al., unpublished results)
below 300. Thus, no trial was designed with adequate power
to test the null hypothesis of no difference in pregnancy rates
between the two gonadotrophin preparations.
Follitrophin type
In the subgroup analyses by type of rFSH administered, there
were nine trials (comprising 1639 cycles) representing 13
comparative assessments of follitrophin alpha with uFSH
(Table IV). There was no significant heterogeneity of treatment
effect among the trials (Breslow–Day statistic 7.3, P 5 0.84).
The common OR was 1.21 (95% CI 0.97–1.51, P 5 0.09) in
favour of rFSH (Figure 3). The risk difference was 3.7% (95%
CI, –0.5–7.9%).
There were three trials (with an aggregate sample size of
1236 cycles) in which follitrophin beta was compared with
uFSH. No significant heterogeneity of treatment effect was
observed (Breslow-Day statistic 0.20, P 5 0.91). The common
OR was 1.19 (95% CI 0.93–1.53, P 5 0.16) in favour of rFSH

(Figure 3 and Table IV). The risk difference was 3.7% (95%
CI, –1.5–8.8%).
Fertilization procedure
Data for IVF were available from seven trials in which IVF
was the only procedure performed, and from four trials in
which the IVF cycles could be separated from the ICSI
cycles. The aggregate sample size was 2308 cycles. There was
no significant heterogeneity of treatment effect among the
trials in which IVF was performed (Breslow–Day statistic 5.0,
P 5 0.89). The common OR was 1.26 (95% CI 1.05–1.52,
P 5 0.02) in favour of rFSH (Figure 3). The risk difference
was 4.4% (95% CI, 0.9–8.0%).
Data for ICSI were available from four trials in which the
ICSI cycles could be separated from the IVF cycles, and from
one trial in which ICSI was the only procedure performed. In
the five trials in which ICSI was performed (with an aggregate
sample size of 567 cycles), there was no significant hetero-
geneity of treatment effect (Breslow-Day statistic 1.4,
P 5 0.84). The common OR was 1.02 (95% CI, 0.72–1.45,
P 5 0.92) in favour of rFSH (Figure 3). The risk difference
was 0.3% (95% CI, –7.4–7.9%).
Follitrophin type and fertilization procedure
Among the trials in which IVF was performed, follitrophin
alpha was used in eight (representing 1072 cycles) and folli-
trophin beta was used in three (representing 1236 cycles)
(Table IV). There were five trials (representing 567 cycles) in
which ICSI was performed and in all of them follitrophin
alpha was used (Table IV). Thus, no trials compared follitrophin
beta with uFSH in patients undergoing treatment with ICSI.
The common OR and the risk differences, together with their
respective 95% CI, broken down by follitrophin type and
fertilization procedure, are shown in Table V.
Logistic regression analysis
The independent variables tested to determine which one(s)
would enter the final model that predicted clinical pregnancy
per cycle started were: fertilization procedure (IVF or
ICSI), type of FSH (follitrophin alpha, follitrophin beta,
urofollitrophin HP or urofollitrophin), source of FSH (recom-
binant or urinary) and GnRHa protocol (long follicular, long
luteal or long unspecified). There were 2875 cases available for
analysis. The variables that entered the model were fertilization
procedure (F to enter 6.9, P 5 0.009) and source of FSH
(F to enter 6.2, P 5 0.013). The clinical pregnancy rate was
significantly higher when ICSI was performed, compared with
IVF (OR 1.3, 95% CI 1.1–1.6). Similarly, rFSH was associated
with a significantly higher clinical pregnancy rate, compared
with uFSH (OR 1.2, 95% CI 1.1–1.5).
Discussion
The overall conclusion from this meta-analysis of all the
available randomized trials which compared rFSH and uFSH
for ovarian stimulation in infertility treatment cycles is that
Recombinant versus urinary FSH
the clinical pregnancy rate per cycle started is statistically
significantly higher with rFSH. The total sample size on which
this conclusion is based was 2875, which is large enough to
demonstrate, with power of ù0.8, a difference in clinical
pregnancy rate of the size observed in this meta-analysis.
The possibility of publication bias influencing the results is
a potential concern with any meta-analysis, and every effort
has been made to identify all the trials that have been published
and as many of the unpublished trials as is humanly possible.
The funnel plot is a useful graphical assessment of the
likelihood of publication bias being present (Egger et al.,
1997). The scatter plot of the odds ratios from the trials in
this review demonstrated a symmetrical, inverse-funnel shape
(Figure 2), thereby providing reassurance that selective publica-
tion is unlikely to be a source of bias in this meta-analysis.
Although the patient profiles in the trials were relatively
similar, and the IVF and ICSI procedures used were standard,
there were differences in the type of gonadotrophin preparations
administered. Two different types of follitrophin (i.e. alpha
and beta) were compared with two different types of uFSH
(i.e. urofollitrophin and urofollitrophin HP), thus producing
four pair-wise comparisons. Subgroup analyses of each of
these comparisons is not reliable because only one trial
compared follitrophin beta with urofollitrophin HP (Andersen
et al., 1998). Nevertheless, the common OR and risk differences
for clinical pregnancy started per cycle for each of the
comparisons are as follows: follitrophin alpha versus
urofollitrophin (OR 5 1.19, risk difference 2.9%), follitrophin
alpha versus urofollitrophin HP, IVF cycles only (OR 5 1.42,
risk difference 6.1%), follitrophin beta versus urofollitrophin
(OR 5 1.21, risk difference 3.9%) and follitrophin beta versus
urofollitrophin HP (OR 5 1.10, risk difference 2.1%). Although
some of the differences in the magnitude of these estimates
of the treatment effect can be explained by the small number
of trials in each comparison, an important factor to be
considered is the FSH isoform heterogeneity among the various
preparations.
Pituitary gonadotrophins secreted into the circulation consist
of a mixture of gonadotrophin molecules with a similar
peptide structure, but with wide differences in the carbohydrate
moieties (Wide, 1997). When synthesized in the cell, the
carbohydrate chains contain sites for the addition of negatively
charged groups of either terminal sialic acid residues or
sulphate. The number of negatively charged sialic acid and
sulphate groups produces variability in the overall charge (as
revealed by electrophoresis) of the isoforms. The physiological
significance of the charge heterogeneity is not well understood,
but is believed to influence the biological properties of the
isoform, in terms of its metabolic clearance rate and endo-
crinological effects at the target organ (Wide, 1997). The less
negatively charged (i.e. more basic) isoforms have a shorter
half-life in the circulation, but have increased biological activity
(Ulloa-Aguirre et al., 1988). In contrast, the more negatively
charged (i.e. more acidic) forms have a longer half-life but
lower bioactivity. These observations suggest that the biological
activity of the pharmaceutical preparations of FSH also may
be dependent on their charge distribution (Ulloa-Aguirre
et al., 1988).
2211
2212
Table III. Details of trials comparing rFSH with uFSH for ovarian stimulation in treatment with IVF or ICSI

First author/ Setting Cycles Drugs compared Randomization Inclusion criteria Exclusion Patient GnRHa FSH HCG criteria Embryo transfer Luteal phase Main Definifition of Ongoing
year criteria characteristics support outcome pregnancy pregnancy

Alvino et al. IVF centre in 55 IVF rFSH alpha s.c. 1:1, open age 18–38 years, PCOS, severe mean age 31 long luteal 225 IU daily for 5,000 IU 2 or 3, 2 days 3 injections not given ‘clinical’ live birth
(1995) Australia uFSH im BMI ,27, endometriosis, years, mean protocol, 5 days, then after retrieval HCG 1000 IU
S.Daya and J.Gunby

ovulatory, ,4 severe male duration of leuprolide adjusted

previous ART factor infertility 3.6 acetate s.c.,
cycles years, 57% tubal 0.5 mg daily
factor, 49%
primary
infertility

Andersen 3 IVF centres 165 IVF rFSH beta 1:1, open age 18–39 years, male factor mean age 33 long rFSH 150 IU 10 000 IU when maximum 3 number of sac seen on
et al. (1998) in Europe uFSH-HP ovulatory years, mean (unspecified) daily, uFSH 225 ù3 follicles embryos oocytes ultrasound
duration of protocol, IU daily, fixed ù17 mm retrieved, total
infertility 5.5 buserelin dose FSH dose
years

Bergh et al. 2 IVF centres 107 IVF, rFSH alpha sc 1:1, assessor- age 18–38 years, PCOS, mean age 32 long luteal 112.5 or 150 IU 10 000 IU usually 2, but HCG sc 1500 IU number of sac seen on viable ù18
(1997) in Sweden 126 ICSI uFSH-HP sc blind, concealed ovulatory, ø3 azoospermia years, mean protocol, daily for 6 days, when 1 follicle maximum 3, every 3–4 days oocytes retrieved ultrasound weeks
allocation previous ART duration of buserelin, i.n. then adjusted .18 mm and usually 2 (or 3) or progesterone gestation
cycles infertility 4.6 33300 µg daily ù2 others ù16 days after 50 mg daily
years, 24% tubal mm retrieval
factor, 53% male
factor, 61%
primary
infertility

Franco et al. IVF centre in 120 ICSI rFSH alpha 1:1, open age ø37 years, mean age 31 long luteal 150 or 225 IU 10 000 IU when maximum 4, 2 4 injections number of sac seen on viable ù12
(1999) Brazil uFSH-HP male factor years, mean protocol, daily (based on ù1 follicle ù17 days after HCG 1000 IU or follicles ù16 ultrasound weeks
infertility duration of leuprolide age) for 7 days, mm retrieval progesterone 50 mm on HCG gestation
infertility 4.7 acetate 0.5 mg then adjusted mg i.m. daily day
years, 93% daily
primary
infertility

Frydman multicentre in 180 IVF, rFSH alpha s.c. 1:1 double-blind, age 18–38 years, mean age 31 long follicular 150 IU daily for 10 000 IU when usually 2 or 3, usually ? number of increasing serum viable ù12
et al. (1998) France 98 ICSI uFSH-HP s.c. concealed ovulatory, ,3 years, mean protocol, 6 days, then 1 follicle but maximum 5, progesterone oocytes received βHCG weeks
allocation previous ART duration of triptorelin depot adjusted .18 mm and 2 or 3 days after gestation
cycles infertility 4.6 (90%) or daily ù2 others ù16 retrieval
years, 36% tubal s.c. or buserelin mm and
factor, 51% male i.n. oestradiol
factor, 65% appropriate
primary
infertility

Hedon et al. 6 IVF centres 90 IVF rFSH beta im 3:2, assessor- age 18–39 years, PCOS, mean age 32 long follicular 150–225 IU 10 000 IU when maximum 3, 2 ù3 injections number of urinary HCG viable ù12
(1995) in France uFSH im blind ovulatory, ø3 spermatozoa years, mean protocol, daily for 4 days, ù3 follicle days after HCG 1500 IU or oocytes retrieved .1000 IU/l or weeks post
previous ART ,103106/ml or duration of triptorelin s.c., then adjusted ù17 mm retrieval ù50 mg sac seen on transfer
cycles ,40% motility infertility 4.8 100 µg daily progesterone ultrasound
years, 63% tubal i.m. daily or 400
factor, 33% mg progesterone
primary vaginally daily
infertility

Lenton et al. 5 IVF centres 118 IVF, rFSH alpha s.c. 1:1, open, age 18–38 years, PCOS, mean age 32 long luteal 150 IU daily for 10 000 IU when maximum 3 progesterone number of sac seen on live birth
(unpublished)a in UK 37 ICSI uFSH-HP s.c. concealed ovulatory, first moderate/severe years, mean protocol, 6 days, then 1 follicle ù18 oocytes retrieved ultrasound
allocation ART cycle endometriosis, duration of buserelin i.n., adjusted mm and ù2
azoospermia infertility 4 900 µg daily or others ù16 mm
years, 42% buserelin s.c.
tubal factor, 0.5 mg/ml
35% male daily
factor, 61%
primary
 Table III. continued

First author/ Setting Cycles Drugs compared Randomization Inclusion criteria Exclusion Patient GnRHa FSH HCG criteria Embryo transfer Luteal phase Main Definifition of Ongoing
year criteria characteristics support outcome pregnancy pregnancy

Manassiev IVF centre in 65 IVF rFSH alpha quasi-randomized, age 20–38 years, mean age 32 long luteal 150–225 IU 10 000 IU when progesterone number of ‘clinical’
et al. (1997) UK uFSH-HP open, allocation ovulatory, ,3 years protocol, daily ù4 follicles 400 mg oocytes retrieved
by patient’s area previous ART buserelin s.c., ù16 mm vaginally 32
of residence cycles 0.5 mg daily daily

O’Dea et al. 7 IVF centres in 114 IVF rFSH alpha s.c. 1:1, open age 18–38 years, male factor long luteal 225 IU daily for 10 000 IU when progesterone im, number of sac seen on
(1993) USA uFSH i.m. ovulatory protocol, 5 days, then ù2 follicles 25 or 50 mg follicles ù14 ultrasound
leuprolide adjusted (max ù16 mm and daily mm n HCG day
acetate sc, 0.5 6000 IU total) oestradiol
mg daily appropriate

Out et al. 18 IVF centres 981 IVF rFSH beta im 3:2, assessor- age 18–39 years, PCOS, mean age 32 long follicular 150 or 225 IU 10 000 IU when maximum 3 ù3 injections number of viable ù12
(1995) in Europe uFSH i.m. blind ovulatory, spermatozoa years, mean protocol, daily for 4 days, ù3 follicles HCG 1500 IU or oocytes weeks post
ø3 previous ,103106/ml or duration of buserelin, i.n. then adjusted ù17 mm ù50 mg retrieved, transfer
ART cycles ,40% motility infertility 6.2 43150 µg daily progesterone ongoing
years, 64% i.m. daily or 400 pregnancy
tubal factor, mg progesterone
44% primary vaginally daily
infertility

RHFSG 9 IVF centres 123 IVF rFSH alpha s.c. 1:1, open age 18–38 years, PCOS, mean age 32 long luteal 225 IU daily for 10 000 IU when usually not given sac seen on live birth
(1995) in Europe uFSH i.m. ovulatory, spermatozoa years, mean protocol, 5 days, then 1 follicle ù18 maximum 3, but ultrasound
ø3 previous ,103106/ml or duration of buserelin acetate adjusted mm and ù2 up to 5
ART cycles ,20% motility infertility 5.4 s.c., 200 µg others ù16 mm
years, 69% daily and estradiol
tubal factor appropriate

Schats et al. several IVF 310 IVF, RFSH alpha s.c. 1:1 assessor- age 18-38 years, moderate/severe mean age 31 long luteal 150 IU daily, maximum 3, vaginal not given viable ù12 viable ù 12
(unpublished)a centres in 186 CSI uFSH-HP s.c. blind, concealed ovulatory ,3 endometriosis years, 23% tubal protocol, fixed dose 2–4 days after progesterone weeks gestation weeks
Belgium & The allocation previous ART factor, 57% male buserelin or (reductions retrieval 200 mg 3 3 (clinical gestation
Netherlands cycles factor, 70% Lucrin 1.0 mg permitted) daily pregnancy n/a)
primary s.c. daily
infertility

aData obtained from Dr C..Howles, Ares–Serono, Geneva, Switzerland (see Acknowledgements).

rFSH 5 recombinant follicle stimulating hormone; uFSH 5 urinary FSH; IVF 5 in-vitro fertilization; ICSI 5 intracytoplasmic sperm injection; BMI 5 body mass index (kg/m2); ART 5 assisted reproduction techniques; PCOS 5 polycystic ovarian syndrome; HCG 5 human
chorionic gonadotrophin; i.n. 5 intra-nasally.
Recombinant versus urinary FSH

2213
S.Daya and J.Gunby

Table IV. Clinical pregnancy per started cycle comparing rFSH with uFSH: data and odds ratio for each
trial, grouped by fertilization procedure and recombinant follitrophin type

Trial rFSH (pregnant/cycle) uFSH (pregnant/cycle) Odds ratio (95% CI)

IVF: follitrophin alpha

Alvino et al. (1995) 8/27 6/28 1.53 (0.46–5.09)
Bergh et al. (1997) 28/55 21/52 1.52 (0.71–3.25)
Frydman et al. (1998) 16/89 19/91 0.83 (0.40–1.74)
Lenton et al. (unpublished)a 17/57 16/61 1.19 (0.54–2.66)
Manassiev et al. (1997) 19/44 5/21 2.27 (0.78–6.61)
O’Dea et al. (1993) 12/56 13/58 0.94 (0.39–2.29)
RHFSG (1995) 12/60 10/63 1.32 (0.53–3.31)
Schats et al. (unpublished)a 39/154 25/156 1.76 (1.02–3.05)
Subtotal 151/542 115/530 1.36 (1.02–1.80)
IVF: follitrophin beta
Andersen et al. (1998) 25/83 23/82 1.10 (0.57–2.16)
Hedon et al. (1995) 20/57 9/33 1.42 (0.57–3.55)
Out et al. (1995) 179/585 107/396 1.19 (0.90–1.57)
Subtotal 224/725 139/511 1.19 (0.93–1.53)
ICSI: follitrophin alpha
Bergh et al. (1997) 25/63 21/63 1.31 (0.64–2.70)
Franco et al. (1998) 22/60 19/60 1.25 (0.59–2.64)
Frydman et al. (1998) 16/50 18/48 0.79 (0.34–1.80)
Lenton et al. (unpublished)a 10/23 7/14 0.77 (0.21–2.89)
Schats et al. (unpublished)a 23/93 25/93 0.89 (0.46–1.72)
Subtotal 96/289 90/278 1.02 (0.72–1.45)
Overall total 471/1556 344/1319 1.20 (1.02–1.42)
aData obtained from Dr C.Howles, Ares–Serono, Geneva, Switzerland (see Acknowledgements).
rFSH 5 recombinant follicle stimulating hormone; uFSH 5 urinary FSH; OR 5 odds ratio;
CI 5 confidence interval; IVF 5 in-vitro fertilization; ICSI 5 intracytoplasmic sperm injection.

follitrophins. The morphological quality of embryos (Phillips

Table V. Clinical pregnancy per started cycle comparing rFSH with uFSH: et al., 1999) and the number of ‘good embryos’ obtained (Von
subgroup analyses of the different types of fertilization procedure performed
and follitrophin administered During et al., 1999) were significantly higher with follitrophin
alpha compared to follitrophin beta.
Type of procedure and Common OR Risk difference (%) The ideal method of evaluating the relative efficacy of the
follitrophin used (95% CI) (95% CI)
four different gonadotropin preparations is to conduct a four-
IVF arm randomized trial with sufficient power and with several
Follitrophin alpha 1.36 (1.02–1.80) 5.2 (0.2–10.2)
Follitrophin beta 1.19 (0.93–1.53) 3.7 (–1.5–8.8)
biological endpoints, including cycle performance character-
ICSI istics, oocyte and embryo quality, and incidence of ovarian
Follitrophin alpha 1.02 (0.72–1.45) 0.3 (–7.4–7.9) hyperstimulation syndrome, spontaneous abortion, clinical
Follitrophin beta no published trials no published trials
available available
pregnancy and live birth. Until then, or until more comparative
data become available to supplement the data in this meta-
Results shown are all in favour of the recombinant follitrophin preparation analysis, these results indicate that there is a statistically
identified.
significant difference in clinical pregnancy rates when rFSH
is compared with uFSH. This finding, and the knowledge that
Based on chromatofocusing studies, the charge distribution the recombinant preparations have batch-to-batch consistency,
patterns for follitrophin beta, urofollitrophin and urofollitrophin are free from urinary protein contaminants and have the
HP demonstrated that the amount of material with isoelectric
potential of being produced in limitless quantities, indicate
point (pI) ,4 was ,24%, 40% and 74%, respectively (i.e.
that rFSH is more appealing for clinical use than uFSH. A
progressively more acidic) (Robertson, 1997). The two follitro-
cost-effectiveness analysis is currently being undertaken to
phins had a similar pattern, except that follitrophin alpha
determine if there are additional advantages, in terms of cost
contained glycoforms with a narrower pI band (i.e. 4–5) than
savings, of using one preparation versus the other.
follitrophin beta (3.5–5.5) (Robertson, 1997). Thus, if terminal
charge pattern dictates in-vivo biological activity, the follitro-
phins should have greater activity than the urofollitrophins.
Unfortunately, information on biological endpoints including Acknowledgements
oocyte quality, and amount of oestradiol production was not We are grateful to the investigators of the included trials, who
available consistently in all the trials in this meta-analysis, provided additional information to address our questions regarding
their studies. We are also grateful to Dr Colin Howles, Ares–Serono,
making it difficult to test this hypothesis. Instead, clinical
Geneva, Switzerland for making available data from trials recently
pregnancy was used as a surrogate measure of biological conducted by E.Lenton in the UK and R.Schats in The Netherlands.
activity. Interestingly, differences in outcome have been We understand that these data are currently being prepared for
reported recently in randomized studies comparing the two publication.
2214

References
Alvino, H., Norman, R.J. and Matthews, C.D. (1995) Recombinant human
follicle stimulating hormone (Gonal-F, Serono) compared to urinary follicle
stimulating hormone (Metrodin) in IVF cycles: a randomised control study.
Fertility Society of Australia/Australian Gynecological Endoscopy Society
1995 Annual Meeting (abstract FSA 46).
Andersen, A.N., Loft, A., Leerentveld, R. et al. (1998) A prospective trial
comparing Puregon 150 IU and Metrodin-HP 225 IU as a fixed-dose
regimen in IVF treatment. Hum. Reprod., 13 (Abstract book 1), 185 (abstract
P-112).
Bergh, C., Howles, C.M., Borg, K. et al. (1997) Recombinant human follicle
stimulating hormone (r-hFSH; Gonal-F) versus highly purified urinary FSH
(Metrodin HP): results of a randomized comparative study in women
undergoing assisted reproductive techniques. Hum. Reprod., 10, 2133–2139.
Breslow, N.E. and Day, N.E. (1980) Statistical Methods in Cancer Research,
Vol I. Analysis of Data from Retrospective Studies of Disease. AIRC
Scientific Publications, Lyon.
Chappel, S.C. (1995) Heterogeneity of follicle stimulating hormone: control
and physiological function. Hum. Reprod. Update, 1, 479–487.
Daya, S. (1998) HMG versus FSH — is there a difference? In Filicori, M. and
Flamigni, C. (eds), Ovulation Induction Update 98. Parthenon, Carnforth, pp.
183–192
Daya, S., Gunby, J., Hughes, E.G. et al. (1995a) Follicle-stimulating hormone
versus human menopausal gonadotropin for in vitro fertilization cycles: a
meta-analysis. Fertil. Steril., 64, 347–354.
Daya, S., Gunby, J., Hughes, E.G. et al. (1995b) Randomized controlled trial
of follicle stimulating hormone versus human menopausal gonadotropin in
in-vitro fertilization. Hum. Reprod., 10, 1392–1396.
Egger, M., Smith, G.D., Schneider, M. et al. (1997) Bias in meta-analysis
detected by a simple graphical test. Br. Med. J., 315, 629–634.
Fisch, B., Avrech, O.M., Pinkas, H. et al. (1995) Superovulation before IVF
by recombinant versus urinary human FSH (combined with a long GnRH
analog protocol): a comparative study. J. Assist. Reprod. Genet., 12, 26–31.
Franco, J.G., Jr, Baruffi, R.L.R., Coelho, J. et al. (1999) A prospective and
randomized study of ovarian stimulation for ICSI with recombinant FSH
versus highly purified urinary FSH. Abstract Book of the 11th World
Congress on In vitro Fertilization and Human Reproductive Genetics, 89
(abstract O-055).
Frydman, R., Avril, C., Camier, B. et al. (1998) A double-blind, randomized
study comparing the efficacy of recombinant human follicle stimulating
hormone (rhFSH/Gonal-F) and highly purified urinary FSH (uhFSH/
Metrodin HP) in inducing superovulation in women undergoing assisted
reproductive techniques. Hum. Reprod., 13 (Abstract book 1), 94 (abstract
O-185).
Hedon, B., Out, H.J., Hugues, J.N. et al. (1995) Efficacy and safety of
recombinant FSH (Puregon) in infertile women pituitary-suppressed with
tiptorelin undergoing in-vitro fertilization: a prospective, randomized,
assessor-blind, multicenter trial. Hum. Reprod., 10, 3102–3106.
Homburg, R., Armar, N.A., Eshel, A. et al. (1988) Influence of serum
luteinizing hormone concentrations on ovulation, conception, and early
pregnancy loss in polycystic ovary syndrome. Br. Med. J., 297, 1024–1026.
Howles, C., Macnamee, M.C. and Edwards, R.G. (1987) Follicular
development and early luteal function of conception and non-conceptional
cycles after human in vitro fertilization: endocrine correlates. Hum. Reprod.,
2, 17–21.
Howles, C.M. (1996) Genetic engineering of human FSH (Gonal-F). Hum.
Reprod. Update, 2, 172–191.
Jansen, C.A.M. and Van Os, H.C. (1996) Puregon without analogues: an
oxymoron. Gynecol. Endocrinol., 10 (Suppl. 1), 34.
Manassiev, N.A., Davies, W.A.R., Leonard, T. et al. (1997) Initial results
from the comparison of recombinant FSH and urinary FSH in an IVF
programme. Hum. Reprod., 12 (Abstract book 1), 265 (abstract #R-068).
Mantel, M. and Haenszel, W. (1959) Statistical aspects of the analysis of data
from retrospective studies of disease. J. Natl. Cancer Inst., 22, 719–748.
Mantzavinos, T., Kanakas, N., Hassiakos, D. et al. (1998) Recombinant
follicle-stimulating hormone (Puregon) versus urinary FSH in in vitro
fertilization. Fertil. Steril., 70 (Suppl. 1), S438–S439 (abstract P1006).
Mitchell, R., Buckler, H.M., Matson, P. et al. (1996) Oestradiol and
immunoreactive inhibin-like secretory patterns following controlled ovarian
hyperstimulation with urinary (Metrodin) or recombinant follicle stimulating
hormone (Puregon). Hum. Reprod., 11, 962–967.
O’Dea, L., Loumaye, E. and Liu, H. (1993) A randomized, comparative,
multicenter clinical trial of recombinant and urinary human FSH in in vitro
fertilization and embryo transfer (IVFET). The American Fertility Society
Recombinant versus urinary FSH
and The Canadian Fertility and Andrology Society 1993 Annual Meeting
Program Supplement, S50-S51 (abstract O-106).
Olijve, W., DeBoer, W., Mulders, J.W.M. et al. (1996) Molecular biology and
biochemistry of human recombinant follicle stimulating hormone (Puregon).
Mol. Hum. Reprod., 2, 371–382.
Out, H.J., Mannaerts, B.M.J.L., Driessen, S.G.A.J. et al. (1995) A prospective,
randomized, assessor-blind, multicentre study comparing recombinant and
urinary follicle-stimulating hormone (Puregon vs Metrodin) in in-vitro
fertilization. Hum. Reprod., 10, 2534–2540.
Out, H.J., Mannaerts, B.M.J.L., Driessen, S.G.A.J. et al. (1996) Recombinant
follicle stimulating hormone (rFSH; Puregon) in assisted reproduction:
more oocytes, more pregnancies. Results from five comparative studies.
Hum. Reprod. Update, 2, 162–171.
Out, H.J., Driessen, S.G.A.J., Mannaerts, B.M.J.L. et al. (1997) Recombinant
follicle-stimulating hormone (follitropin beta, Puregon) yields higher
pregnancy rates in in vitro fertilization than urinary gonadotropins. Fertil.
Steril., 68, 138–142.
Peto, R. (1987) Why do we need systematic overviews of randomized trials?
Statist. Med., 6, 233–240.
Phillips, E., Page, M. and Fleming, S.D. (1999) A prospective comparison of
two different recombinant FSH preparations. Abstract Book of the 11th
World Congress on In vitro Fertilization and Human Reproductive Genetics,
88 (abstract O-053).
Recombinant Human FSH Study Group (abbreviated to RHFSG) (1995)
Clinical assessment of recombinant human follicle-stimulating hormone in
stimulating ovarian follicular development before in vitro fertilization.
Fertil. Steril., 63, 77–86.
Regan, L., Owen, E.J. and Jacobs, H.S. (1990) Hypersecretion of luteinizing
hormone, infertility and miscarriage. Lancet, 336, 1141–1144.
Robertson, W.R. (1997) Gonadotrophin isoform patterns in different
pharmaceutical preparations. In Kahn, J.A. (ed.), Gonadotropin Isoforms:
Facts and Future. Serono Fertility Series, Vol. 2. Ciconia Foundation,
Copenhagen, pp. 53–60.
Stanger, J. and Yovich, J.J. (1985) Reduced in vitro fertilization of human
oocytes from patients with raised basal luteinizing hormone levels during
the follicular phase. Br. J. Obstet. Gynaecol., 92, 385–393.
Sunde, A., Kahn, J.A., von Düring, V. et al. (1994) In IVF, administration of
recombinant FSH (Puregon) gives a significantly higher mean number of
oocytes, fertilized oocytes and good embryos compared with administration
of purified FSH (Metrodin). Hum. Reprod., 9 (Suppl. 4), 61 (abstract 144).
Ulloa-Aguirre, A., Espinoza, R., Damian-Matsumura, P. et al. (1988)
Immunological and biological potencies of the different molecular species
of gonadotropins. Hum. Reprod., 3, 491–501.
Von Düring, V., Kahn, J.A., Sunde, A. et al. (1999) Results of a prospective,
randomized study comparing two recombinant FSH preparations (Gonal-F,
Puregon) in IVF and ICSI treatments. Abstract Book of the 11th World
Congress on In vitro Fertilization and Human Reproductive Genetics, 265
(abstract P-200).
Wide, L. (1997) Isoforms of human gonadotrophins under different
physiological conditions. In Kahn, J.A. (ed.), Gonadotrophin Isoforms:
Facts and Future. Serono Fertility Series, Vol 2. Ciconia Foundation,
Copenhagen, pp. 43–52.
Received on September 21, 1998; accepted on June 24, 1999
2215