Content uploaded by Nick Raine-Fenning
Author content
All content in this area was uploaded by Nick Raine-Fenning
Content may be subject to copyright.
ORIGINAL ARTICLE Infertility
Association between the number of
eggs and live birth in IVF treatment:
an analysis of 400 135 treatment cycles
Sesh Kamal Sunkara1, Vivian Rittenberg1, Nick Raine-Fenning2,
Siladitya Bhattacharya3, Javier Zamora4, and Arri Coomarasamy5,*
1
Assisted Conception Unit, Guy’s and St Thomas’ Foundation Trust, King’s College London, London, UK
2
Nottingham University Research
and Treatment Unit in Reproduction (NURTURE), Division of Human Development, School of Clinical Sciences, University of Nottingham,
Nottingham, UK
3
Division of Applied Health Sciences, School of Medicine and Dentistry, University of Aberdeen, Aberdeen, UK
4
Clinical
Biostatistics Unit, Hospital Ramon y Cajal. IRYCIS. CIBERESP, University Complutense of Madrid, Spain
5
School of Clinical and Experimental
Medicine, College of Medical & Dental Sciences, University of Birmingham, Academic Unit, 3rd Floor, Birmingham Women’s Hospital,
Birmingham B15 2TG, UK
*Correspondence address. Tel: +44-121-623-6835; Fax: +44-121-626-6619; E-mail: a.coomarasamy@bham.ac.uk
Submitted on December 10, 2010; resubmitted on March 2, 2011; accepted on March 10, 2011
background: While live birth is the principal clinical outcome following in vitro fertilization (IVF) treatment, the number of eggs
retrieved following ovarian stimulation is often used as a surrogate outcome in clinical practice and research. The aim of this study was
to explore the association between egg number and live birth following IVF treatment and identify the number of eggs that would optimize
the IVF outcome.
methods: Anonymized data on all IVF cycles performed in the UK from April 1991 to June 2008 were obtained from the Human Ferti-
lization and Embryology Authority (HFEA). We analysed data from 400 135 IVF cycles. A logistic model was fitted to predict live birth using
fractional polynomials to handle the number of eggs as a continuous independent variable. The prediction model, which was validated on a
separate HFEA data set, allowed the estimation of the probability of live birth for a given number of eggs, stratified by age group. We pro-
duced a nomogram to predict the live birth rate (LBR) following IVF based on the number of eggs and the age of the female.
results: The median number of eggs retrieved per cycle was 9 [inter-quartile range (IQR) 6– 13]. The overall LBR was 21.3% per fresh
IVF cycle. There was a strong association between the number of eggs and LBR; LBR rose with an increasing number of eggs up to !15,
plateaued between 15 and 20 eggs and steadily declined beyond 20 eggs. During 2006–2007, the predicted LBR for women with 15 eggs
retrieved in age groups 18– 34, 35 – 37, 38 – 39 and 40 years and over was 40, 36, 27 and 16%, respectively. There was a steady increase in
the LBR per egg retrieved over time since 1991.
conclusion: The relationship between the number of eggs and live birth, across all female age groups, suggests that the number of eggs
in IVF is a robust surrogate outcome for clinical success. The results showed a non-linear relationship between the number of eggs and LBR
following IVF treatment. The number of eggs to maximize the LBR is !15.
Key words: IVF treatment / egg numbers / live birth / nomogram
Introduction
The primary aim of in vitro fertilization (IVF) treatment is to achieve a
term live birth. However, as the number of eggs retrieved is con-
sidered to be an important prognostic variable, IVF treatment proto-
cols aim to optimize this outcome. Studies evaluating IVF treatment
regimens and ovarian reserve tests such as anti-mu
¨llerian hormone
or antral follicle count often use the number of eggs as a surrogate
outcome. However, this practice has been criticized (Vail and Gar-
dener, 2003) as the relationship between the number of eggs and
live birth is poorly understood.
Previous work on the relationship between the number of eggs
retrieved and pregnancy rates following IVF, based on data from
single centres and involving small sample sizes, has shown conflicting
results (Meniru and Craft 1997;Letterie et al., 2005;Kably Ambe
et al., 2008;Molina Hita Ma. del et al., 2008;Hamoda et al.,
2010). None has reported live birth rates (LBRs), but instead
focused on rates of clinical or ongoing pregnancy. The aim of our
study was to determine the association between the number of
eggs retrieved and the LBR in fresh IVF cycles, based on the
analysis of a large national database involving 400 135 IVF treatment
cycles.
&The Author 2011. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved.
For Permissions, please email: journals.permissions@oup.com
Human Reproduction, Vol.0, No.0 pp. 1– 7, 2011
doi:10.1093/humrep/der106
Hum. Reprod. Advance Access published May 10, 2011
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from
Materials and Methods
Anonymized data were obtained from the Human Fertilization and Embry-
ology Authority (HFEA) for all IVF cycles performed in the UK from April
1991 to June 2008 (www.hfea.gov.uk/5874.html, HFEA authority). The
HFEA, which is the statutory regulator of assisted conception treatment
in the UK, has collected data on all IVF treatment cycles performed in
the UK since its inception in 1991. Overall, 787 030 IVF cycles were
recorded in this period. For the purpose of the study, cycles involving
gamete or zygote intra-fallopian transfer (GIFT, ZIFT), egg donation, egg
sharing, embryo donation or where the source of embryos was not
specified, preimplantation genetic diagnosis, surrogacy, oocyte cryopreser-
vation, frozen embryo replacement, and cycles where no eggs were
retrieved or all embryos were frozen were excluded from the analysis.
Information was obtained on the number of eggs retrieved, age group
(18–34, 35–37, 38 – 39, 40 years and over), treatment period (1991–
2008) and live birth outcome. A live birth is defined as any birth event
in which at least one baby is born alive.
Statistical analysis
We described the characteristics of the cohort using absolute and relative
frequencies for categorical variables, and means and medians with
measures of spread for continuous variables. We computed crude LBRs
for the whole cohort, and stratified by period of treatment and age.
To explore the association between the number of eggs and live birth
outcome, we fitted a maximum likelihood logistic model with live birth
outcome as the dependent variable and using a fractional polynomial to
handle the number of eggs as a continuous independent variable. We
used the closed test procedure for function selection as described by
Royston and Sauerbrei (2008). We also introduced in the model indicator
variables for age and period of treatment. We computed robust standard
errors to account for the non-independence of observations from multiple
treatment cycles in a single participant.
The model calibration and discrimination ability was assessed by the
Hosmer–Lemeshow test and the c-index statistic. The live birth
outcome has substantially improved over the four time periods and
thus, for the development of the prediction model, we used the data
set generated after 2006. As the age of the woman has a significant
impact in determining the probability of a live birth, we computed this
probability stratified by age group.
To validate our model, we split the cohort into two parts according to
the period of treatment. The first, comprising cycles performed between
2006 and 2007, was used to derive the model, while data generated from
2008 onwards were used to validate it. Finally, we constructed a nomo-
gram to calculate the probability of a live birth based on the number of
eggs and age.
Figure 1 Data selection process.
2Sunkara et al.
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from
Results
The data selection process with the numbers of cycles excluded (with
reasons for exclusion) is provided in Fig. 1. Of an initial total of
787 030 cycles, 400 135 were eligible for analysis. Characteristics of
the analysis cohort are given in Table I. Half of all cycles were con-
ducted on women between 18 and 34 years of age, while 12.6%
were in women 40 years or over. The major cause of infertility was
male factor (56.3%), and conventional IVF was used in the majority
(61.9%) of cycles.
The median number of eggs retrieved was 9 [inter-quartile range
(IQR) 6–13; Fig. 2a] and the median number of embryos created
was 5 (IQR 3–8; Fig. 2b). The overall LBR in the entire cohort was
21.3% [95% confidence interval (CI): 21.2–21.4%], with a gradual
rise over the four time periods in this study (14.9% in 1991– 1995,
19.8% in 1996–2000, 23.2% in 2001–2005 and 25.6% in
2006–2008).
Association between the number of eggs
and live birth
There was a strong association between the number of eggs and the
LBR (Fig. 3a) which rose with increasing number of eggs up to !15,
plateaued between 15 and 20 eggs and steadily declined beyond 20
eggs. The same pattern was observed in all four of the time
periods. For a given number of eggs, LBRs increased over time
(Fig. 3b) but decreased with increasing age (Fig. 3c).
Predicting live birth
To ensure that the predicted LBR was relevant to current practice, the
predictive model was derived from observations generated from data
on treatments from 2006 to 2007. The data from 2008 were used for
model validation. The final model, which includes non-linear terms for
the number of eggs and age as an indicator variables, closely fits with
observed data (Fig. 4). The functional form of the model with coeffi-
cients and their robust standard errors is provided in Appendix
(Supplementary data). The model was well calibrated (Hosmer–
Lemeshow
x
2
¼3.92, df ¼8, P¼0.86) and the c-index was 0.65.
........................................................................................
Table I Characteristics of the cohort (n5400 135).
Characteristic n(%)
Age (given categories)
18–34 years 200 982 (50.2)
35–37 years 97 345 (24.3)
38–39 years 51 385 (12.8)
40 years and over 50 423 (12.6)
Number of previous IVF cycles
0 230 924 (58.8)
1 87 471 (22.3)
2 40 994 (10.4)
3 or more 33 157 (8.5)
Previous LB (yes) 18 633 (4.7)
Cause of infertility
a
Male factor 221 047 (56.3)
Tubal disease 117 722 (30.3)
Ovulatory disorder 46 071 (11.9)
Endometriosis 29 804 (7.5)
Unexplained 131 652 (33.7)
Treatment type
IVF 247 640 (61.9)
ICSI 151 788 (37.9)
Unknown 707 (0.2)
Eggs retrieved (Fig. 2a)
Median (IQR) 9 (6–13)
Embryos created (Fig. 2b)
Median (IQR) 5 (3–8)
Treatment cycles in each period
1991– 1995 72 682 (18.2)
1996– 2000 117 050 (29.3)
2001– 2005 129 402 (32.3)
2006 onwards 81 001 (20.2)
a
The causes of infertility are not mutually exclusive.
Figure 2 Number of eggs retrieved and embryos created.
Number of eggs and live birth in IVF treatment 3
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from
The predicted probability of live birth for a given number of eggs
and age group is provided in Table II. This information is summarized
in the nomogram (Fig. 5), which provides a graphic depiction for easy
interpretation of the results.
Validation was performed on 17 366 IVF cycles and 4863 live births.
Predictive ability of the model does not differ between the derivation
and validation cohorts. Although the Hosmer–Lemeshow
x
2
¼16.3
(df ¼8, P¼0.04) is statistically significant due to the large sample
size, the differences between predicted and observed live birth prob-
abilities are clinically unimportant (Fig. 6). The c-index was 0.66 for the
temporal validation cohort.
Discussion
Our results show a strong relationship between the number of eggs
and the LBR in a fresh IVF cycle. The best chance of live birth was
associated with the number of eggs of around 15 and showed a
decline with .20 eggs. LBRs were seen to decline with advancing
maternal age although a global increase over time was noted across
all age groups.
We used the largest available clinical IVF database to assess the
association between the number of eggs and live birth in a fresh IVF
cycle. Although the clinical heterogeneity within the data set may be
considered a drawback, such differences increase the generalizability
of our findings. The model has been derived using more recent data
(2006–2007) which closely represent current practice and validated
using the most recent subset of IVF cycles within the cohort (2008)
constituting a temporal external validation as current recommen-
dations advocate.
Although the size of the database was large, we encountered pro-
blems with missing data and loss to follow-up; such data were
excluded from the analysis. Data involving cycles where all embryos
were frozen for reasons such as risk of ovarian hyperstimulation
Figure 3 Association between egg number and live birth rate.
Figure 4 Observed versus predicted live birth rate in data from
2006 to 2007.
4Sunkara et al.
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from
......................................................... ........................................................... .......................................................... ..........................................................
..........................................................................................................................................................................................................................................................
Table II Predicted probabilities for live birth.
18–34 years 35– 37 years 38–39 years 40 years and over
Eggs nObserved
live birth
(%)
Predicted
live birth
(%)
95%CI
predicted
(%)
nObserved
live birth
(%)
Predicted
live birth
(%)
95%CI
predicted
(%)
nObserved
live birth
(%)
Predicted
live birth
(%)
95%CI
predicted
(%)
nObserved
live birth
(%)
Predicted
live birth
(%)
95%CI
predicted
(%)
1 253 8 7 7, 8 275 7 6 6, 7 280 5 4 4, 5 541 1 2 2, 3
2 540 17 16 15, 17 579 14 14 13, 14 509 9 9 9, 10 774 5 5 5, 5
3 819 21 22 21, 22 840 18 19 18, 19 718 12 13 13, 14 1002 6 7 7, 8
4 1221 29 26 25, 26 1091 22 22 22, 23 817 17 16 15, 17 1025 9 9 8, 9
5 1486 29 29 28, 29 1245 24 25 24, 26 899 18 18 17, 19 1058 11 10 10, 11
6 1684 30 31 30, 31 1298 27 27 26, 28 854 18 20 19, 21 980 9 11 11, 12
7 1809 35 33 32, 33 1321 29 29 28, 30 846 21 21 20, 22 901 11 12 11, 13
8 1904 34 34 34, 35 1278 29 30 30, 31 729 23 22 22, 23 771 11 13 12, 14
9 1898 35 36 35, 36 1207 31 31 31, 32 672 23 23 23, 24 627 15 14 13, 14
10 1805 36 37 36, 37 1168 31 33 32, 33 630 25 24 23, 25 538 14 14 13, 15
11 1795 36 38 37, 38 1035 34 33 33, 34 549 23 25 24, 26 466 17 15 14, 15
12 1639 38 38 38, 39 872 34 34 33, 35 474 26 26 25, 27 401 15 15 14, 16
13 1484 38 39 38, 40 703 34 35 34, 36 411 26 26 25, 27 298 16 15 15, 16
14 1291 40 40 39, 40 675 37 35 34, 36 329 26 27 26, 28 252 16 16 15, 17
15 1155 40 40 39, 41 526 41 36 35, 37 256 26 27 26, 28 229 17 16 15, 17
20 487 41 41 41, 42 219 36 37 36, 38 93 29 28 27, 29 74 18 17 16, 18
25 172 42 41 40, 43 63 43 37 36, 38 37 30 28 27, 30 19 26 17 16, 18
30 67 31 40 38, 42 20 50 36 33, 38 4 0 27 25, 29 12 25 16 14, 18
35 14 29 37 33, 41 7 29 33 29, 37 5 0 25 22, 28 0 – 15 13, 17
40 15 27 33 28, 40 7 43 30 24, 35 2 50 22 18, 27 0 – 13 11, 16
Number of eggs and live birth in IVF treatment 5
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from
syndrome (OHSS) also could not be analysed. Also, our study only
analysed the outcome of fresh IVF cycles, and did not take into
account the impact of frozen– thawed cycles on the cumulative LBR
(due to data currently not being available as the current HFEA data
set does not allow linkage of fresh and frozen cycles in the same
woman). It is possible that the declining effect of higher number of
eggs on the outcome of a fresh IVF cycle becomes attenuated by
the increasing likelihood of a pregnancy in a subsequent frozen–
thawed transfer cycle. The existing format of the anonymized data
set precluded detailed exploration of age-related outcomes other
than comparison of the existing age categories. This has certain
drawbacks; for example, over half of all women were in the same
age group (18–34 years). At the other end of the spectrum, all
women over 40 years were treated as a homogeneous group although
outcomes in older women change significantly with small increases in
age, with LBRs of 11.9% in women aged 40–42 years falling to 3.4% in
women aged 43 – 44 years (http://www.hfea.gov.uk/ivf-figures, HFEA
authority). No information regarding type of stimulation or gonado-
trophins used in IVF treatment was collected by the HFEA, and
these data were therefore unavailable for analysis.
Previous studies looking at the relationship between the number of
eggs and pregnancy rates have reported inconsistent results in showing
that pregnancy rates increased with an increasing number of eggs
(Meniru and Craft, 1997), best pregnancy rates being obtained with
number of eggs of 10–15 (Kably Ambe et al., 2008), or 7–16
(Molina Hita Ma. del M et al., 2008). Furthermore, these studies
involved small numbers and were reported from single centres,
which limited their generalizability. Our study is the first to provide
vital information on predicting the LBR on the basis of eggs retrieved
in women of different age groups. The simplicity of the nomogram
facilitates interpretation of this information by clinicians as well as
couples seeking IVF treatment.
Knowledge of factors predicting IVF success is critical to patients and
service providers in informing decisions to embark on IVF treatment and
the choice of ovarian stimulation regimens. Such information is also
helpful in counselling couples about deciding against further IVF treat-
ment or plans to opt for donor eggs. To date, most clinical decisions
on ovarian stimulation in IVF have been based on ovarian reserve
tests which are good at predicting numbers of eggs retrieved but poor
in terms of predicting live birth (Broekmans et al., 2006;Broer et al.,
2009). By allowing clinicians to link the (predicted) number of eggs to
live birth, the nomogram generated by this study is likely to facilitate
use of these tests to optimize outcomes in IVF while preventing compli-
cations relating to production of an excessive number of eggs.
Figure 5 Nomogram to calculate predicted live birth probability given egg number and age.
Figure 6 Calibration plot of the validation model. Circles indicate
the observed proportion of live births per tenth of predicted prob-
ability. The dashed line represents perfect calibration.
6Sunkara et al.
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from
Our data suggest that around 15 eggs may be the optimal number
to aim for in a fresh IVF cycle in order to maximize treatment
success while minimizing the risk of OHSS which is associated with
high number of eggs of .18 (Lyons et al., 1994;Verwoerd et al.,
2008;Lee et al., 2010). The decline in the LBR observed with
higher number of eggs could be due to the deleterious effect of
the raised serum estradiol levels affecting embryo implantation (Val-
buena et al., 2001;Mitwally et al., 2006;Joo et al., 2010). Even in
cases where the aim is to freeze surplus embryos for future use,
existing data suggest that the numbers of embryos frozen after a
fresh IVF cycle are not enhanced by retrieving .18 eggs (Hamoda
et al., 2010). On the other hand, there has been a recent trend
towards mild ovarian stimulation in IVF with the emphasis on reco-
vering fewer eggs than previously deemed optimal (Fauser et al.,
2010). Our findings support the use of moderate stimulation proto-
cols over either mild or aggressive stimulation protocols in IVF
treatment.
The nomogram that we have established is the first of its kind that
allows prediction of live birth for a given number of eggs and female
age group. This is potentially valuable for patients and clinicians in plan-
ning IVF treatment protocols and counselling regarding the prognosis
for a live birth occurrence, especially in women with either predicted
or a previous poor ovarian response.
The relationship observed between the number of retrieved eggs
and live birth in a fresh IVF cycle, across all female age groups,
suggests that number of eggs is a reasonable surrogate outcome
to use in IVF practice and research. Future research should focus
on establishing the relationship between retrieved eggs and the
cumulative LBR per IVF cycle by including the outcome following
replacement of all frozen embryos generated from a single fresh
IVF treatment.
Authors’ roles
S.K.S. undertook the task of verifying and validating the HFEA data
and contributed to writing the manuscript. V.R. undertook the task
of verifying and validating the HFEA data. N.R.-F. contributed to
writing the manuscript. S.B. contributed to writing the manuscript
and appraised it critically for important intellectual content. J.Z.
undertook the analysis of the data and contributed to writing the
manuscript. A.C. conceived the idea and contributed to writing
the manuscript.
Supplementary data
Supplementary data are available at http://humrep.oxfordjournals.
org/.
Acknowledgements
We thank all the centres in the UK for their work in completing and
forwarding all the treatment and outcome details to the HFEA, and
the staff at the HFEA for validating this data.
Conflict of interest: none declared.
References
Broekmans FJ, Kwee J, Hendriks DJ, Mol BW, Lambalk CB. A systematic
review of tests predicting ovarian reserve and IVF outcome. Hum
Reprod Update 2006;12:685–718.
Broer SL, Mol BW, Hendriks D, Broekmans FJ. The role of antimullerian
hormone in prediction of outcome after IVF: comparison with the
antral follicle count. Fertil Steril 2009;91:705–714.
Fauser BC, Nargund G, Andersen AN, Norman R, Tarlatzis B, Boivin J,
Ledger W. Mild ovarian stimulation for IVF: 10 years later. Hum
Reprod 2010;25:2678– 2684.
Hamoda H, Sunkara S, Khalaf Y, Braude P, El-Toukhy T. Outcome of fresh
IVF/ICSI cycles in relation to the number of oocytes collected: a review
of 4,701 treatment cycles. Hum Reprod 2010;25:147.
HFEA authority. Access Anonymised HFEA Data. (19 November 2010, date
last acccesed) http://www.hfea.gov.uk/5874.html.
HFEA authority. http://www.hfea.gov.uk/ivf-figures-2006.html. (19
November 2010, date last acccesed).
Joo BS, Park SH, An BM, Kim KS, Moon SE, Moon HS. Serum oestradiol
levels during controlled ovarian hyperstimulation influence the
pregnancy outcome of in vitro fertilization in a concentration-
dependent manner. Fertil Steril 2010;93:442 – 446.
Kably Ambe A, Estevez Gonzalez S, Carballo Mondragon E, Dura
´n
Monterrosas L. Comparative analysis of pregnancy rate/captured
oocytes in an in vitro fertilization program. Ginecol Obstet Mex 2008;
76:256–260. Spanish
Lee KH, Kim SH, Jee BC, Kim YJ, Suh CS, Kim KC, Lee WD. Comparison
of clinical characteristics between early and late patterns in hospitalized
patients with ovarian hyperstimulation syndrome. Fertil Steril 2010;
93:2274–2280.
Letterie G, Marshall L, Angle M. The relationship of clinical response,
oocyte number, and success in oocyte donor cycles. J Assist Reprod
Genet 2005;22:115–117.
Lyons CA, Wheeler CA, Frishman GN, Hackett RJ, Seifer DB, Haning RV
Jr. Early and late presentation of the ovarian hyperstimulation syndrome:
two distinct entities with different risk factors. Hum Reprod 1994;
9:792–799.
Meniru GI, Craft IL. Utilization of retrieved oocytes as an index of the
efficiency of superovulation strategies for in-vitro fertilization
treatment. Hum Reprod 1997;12:2129– 2132.
Mitwally MF, Bhakoo HS, Crickard K, Sullivan MW, Batt RE, Yeh J.
Estradiol production during controlled ovarian hyperstimulation
correlates with treatment outcome in women undergoing in vitro
fertilization-embryo transfer. Fertil Steril 2006;86:588–596.
Molina Hita Ma. del M, Lobo Martinez S, Gonzalez Varea, Montejo
Gadea JM, Garijo Lopez E, Cuadrado Mangas C. Correlation between
the number of oocytes and the pregnancy rate in IVF-ICSI cycles.
Revista Iberoamericana de Fertilidad y Reproduccion Humana 2008;
25:153–159. Spanish
Royston P, Sauerbrei W. Multivariate Model-Building: a Pragmatic Approach
to Regression Analysis Based on Fractional Polynomials for Modelling
Continuous Variables. Chichester, West Sussex, England: John Wiley &
Sons, 2008.
Vail A, Gardener E. Common statistical errors in the design and analysis of
subfertility trials. Hum Reprod 2003;18:1000–1004.
Valbuena D, Martin J, de Pablo JL, Remohı
´J, Pellicer A, Simo
´nC.
Increasing levels of estradiol are deleterious to embryonic
implantation because they directly affect the embryo. Fertil Steril
2001;76:962–968.
Verwoerd GR, Mathews T, Brinsden PR. Optimal follicle and oocyte
numbers for cryopreservation of all embryos in IVF cycles at risk of
OHSS. Reprod Biomed Online 2008;17:312– 317.
Number of eggs and live birth in IVF treatment 7
at McGill University Libraries on May 10, 2011humrep.oxfordjournals.orgDownloaded from