Human Reproduction, Vol.27, No.1 pp.

89 96, 2012
Advanced Access publication on November 7, 2011 doi:10.1093/humrep/der373

ORIGINAL ARTICLE Embryology

Non-invasive metabolomic proling of

Day 2 and 5 embryo culture medium: a
prospective randomized trial
T. Hardarson 1,*, A. Ahlstrom1, L. Rogberg 1, L. Botros 2, T. Hillensjo 1,
G. Westlander1, D. Sakkas2, and M. Wikland 1
1

Fertilitetscentrum, Goteborg, Sweden 2Molecular Biometrics Inc., Norwood, MA, USA

*Correspondence address. E-mail: thorir.hardarson@fcivf.com

background: Near infrared (NIR) spectroscopy is a technology proposed to facilitate non-invasive screening for the most optimal
human embryo for uterine transfer. It has been proposed that the NIR spectral prole of an embryos spent culture medium can be
used to generate a viability score that correlates to implantation potential. As the initial proof of principle studies were all retrospective,
our aim was to investigate whether NIR spectroscopy on spent embryo culture medium in an on-site, prospective setting could improve
the ongoing single embryo transfer (SET) pregnancy rate after Day 2 and 5 transfers.
methods: We conducted a single-centre, double-blinded, randomized controlled trial in which the NIR group was compared with a
control group. The primary outcome was the clinical pregnancy rate after 67 weeks of gestation per randomized patient. In the control
group embryo selection was based only on traditional morphological evaluation while in the treatment group NIR spectroscopy was
added to the morphological evaluation.
results: The study was terminated early as the analysis of the Data Safety Monitoring Board showed a very low conditional power of
superiority for the primary outcome. Of the 752 patients calculated to be included in the study, 164 and 163 patients were randomized into
the NIR and control groups, respectively. No signicant difference in the ongoing pregnancy rate per randomized patient was found between
the NIR and the control group, 34.8 versus 35.6%, (P 0.97). The proportional difference between the study groups mean was 20.8% (95%
condence interval 11.4 to 10.2).

conclusions: This study shows that adding NIR spectroscopy, in its present form, to embryo morphology does not improve the
chance of a viable pregnancy when performing SET. The NIR technology appears to need further development before it can be used as
an objective marker of embryo viability.
clinical trials identifier: ISRCTN23817363.
Key words: RCT / NIR / embryo selection / pregnancy / metabolomic proling

Introduction
During recent years several non-invasive methods have been introduced for choosing the most optimal embryo for uterine transfer
(Brison et al., 2004; Seli et al., 2007; Katz-Jaffe et al., 2009). Among
these new methods is a rapid non-invasive screening technology
using near infrared (NIR) spectroscopy to analyse the chemical
content of spent culture medium. This technology examines NIR
spectra between 920 and 1675 nm and takes advantage of the fact
that molecules absorb specic frequencies that are characteristic of
their structure. Therefore an embryos medium sample will generate
a spectrum due to vibrations of specic functional groups including

N H, C H, OH and S H. The intensity of the light absorbed

can be determined by analysis of the resulting spectra and is directly
proportional to the concentrations of molecules, such as albumin,
lactate, pyruvate, glutamate and glucose, in the sample. It has been
proposed that the NIR spectral prole of an embryos spent culture
medium can be used to predict the implantation potential and can
serve as a method to compare and select between sibling embryos.
The conceptual basis of this approach is not new; many other
studies have demonstrated that embryos modify their environment
(medium) during culture and these modications can be measured
and correlated to an embryos developmental and implantation potential (Renard et al., 1980; Gardner and Leese, 1986). However, by

& 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

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Submitted on July 6, 2011; resubmitted on September 26, 2011; accepted on October 5, 2011

90
comparison these studies have used technologies not available to prospectively test in a clinical setting. Several proof of principle studies
examining results from Day 2, 3 and 5 transfers, including studies
from our clinic, have shown, through spectral analysis of large
numbers of medium samples from embryos with known pregnancy
outcomes (positive versus negative), that a predictive algorithm can
be created to predict the outcomes of a blinded sample population
(Hardarson et al., 2007; Seli et al., 2007; Scott et al., 2008;
Vergouw et al., 2008). These studies have, however, all been retrospective in their design and have relied largely on sending spent
frozen culture medium to a single central laboratory for measurement.
The aim of this study was to investigate if the NIR spectroscopy
system, utilizing an on-site instrument loaded with previously established predictive algorithms, was able to improve the ongoing single
embryo transfer (SET) pregnancy rate.

Study design
This was a blinded randomized controlled study comparing the NIR group
with a control group. All consecutive IVF patients at Fertility Center
Scandinavia, Carlanderska Hospital, Gothenburg, between April 2008
and December 2010 were assessed for eligibility. Those patients who fullled the inclusion criteria and signed informed consent forms were
included in the study. Ethical approval was obtained from the ethics committee at Goteborg University. The study was registered as an RCT with
the number of ISRCTN23817363.
The primary end-point was the number of ongoing pregnancies per randomized patient (intention-to-treat, ITT), based on fetal heart activity by
means of ultrasound around 45 days after embryo transfer.
Secondary end-points were the total number of biochemical pregnancies per randomized patient, based on a positive urine pregnancy test
18 days after embryo transfer, and the total number of ongoing pregnancies per randomized patient, based on fetal heart activity around 45 days
after embryo transfer on Day 2 and Day 5, respectively.
The randomization into the two study groups was performed using a
computerized randomization program balancing for all of the following
prognostic variables by way of minimization (1:1): mean female age,
patient age group categorized as ,34, 34 38 and .38; mean number
of good quality embryos (GQE) on the day of transfer; number of GQE
available for transfer categorized by groups 2, 3 4 and .5; mean
number of previous fresh IVF cycles leading to embryo transfer and
reason for infertility (tubal factor, male factor, unexplained and other).
The randomization was performed by the laboratory staff and blinded
from both the patient and physician.

Inclusion and exclusion criteria

All couples who underwent IVF or ICSI at the Fertilitetscentrum and
received verbal and written information about the study.
The couple was excluded from the study if they: planned for two
embryos to be transferred, had less than two GQE on the day of
embryo transfer, had a testicular biopsy performed (testicular sperm
aspiration/testicular sperm extraction), had the embryo transfer on Day
3 or 4 or had previously been randomized to the study. Tight inclusion criteria of at least two GQE were used as we did not feel it ethical to include
all available embryos for the NIR analyses. This study was seen as an initial
step of testing the system and if the results were improved using the NIR
system with GQE, we intended to establish a future study where all
embryos in the cohort were examined.

Sample size calculation

It was calculated that in order to detect an increase of 10% in the ongoing
pregnancy rate (i.e. from 30 to 40%) for all patients receiving a SET, 376
patients in each group would be needed to achieve a power of 80%, with a
two-sided Fishers exact test, at a signicance level of 5%. This increase in
the pregnancy rate was based on previous results at the Fertilitetscentrum,
Gothenburg for the year 2006. To compensate for possible drop-outs (i.e.
patients who may withdraw from the study after they have been allocated), we added 2% to the total number of patients needed for the
study. We therefore estimated that we would need a total of 768 patients
(384 in each arm).

Data Safety Monitoring Board

The study was monitored by an independent Data Safety Monitoring
Board (DSMB). Their role was to look at the data and advise whether
or not to proceed with the study. The DSMB consisted of a statistician
(other than the one performing the nal statistical analysis) and a
medical scientist, not involved in the study. The plan was to perform
this analysis after 50% of the patients had been randomized. The following conditions were set-up in advance, in order to be able to decide
whether or not it would be clinically relevant to proceed with the study:
(i) If the difference between the two study groups is 3% in favour of
the treatment (NIR) group, the study should be stopped.
(ii) If the difference between the two study groups is ,3 10% in favour
of the treatment (NIR) group, a new power analysis should be
performed.
(iii) If the difference between the two study groups is .10% in favour of
the treatment (NIR) group, the study according to this study protocol
should be continued.
According to the study plan, if the study was to be stopped, the data
should be kept blinded until the database was locked. When the clean
le had been locked, the database could be opened for analysis. The following statistical analyses were then to be performed.

Statistical methods
The following demographics and baseline variables were analysed by treatment group and subsequently corrected for in the statistical analysis as
possible confounders: maternal age, BMI, number of previous IVF treatment cycles, number of fertilized oocytes, method of fertilization,
number of aspirated follicles and oocytes, number of GQE on the day
of transfer, number of embryos transferred, type of pituitary regulation,
total dosage of gonadotrophins administered, day of transfer, medical
cause of infertility, number of cells on transfer Day 2, degree of fragmentation on transfer Day 2, degree of expansion on transfer Day 5, morphological grade of the inner cell mass (ICM) and trophectoderm cells (TE)
and viability score (only NIR group).
All data were analysed from all of the randomized couples, i.e. ITT analysis and from couples who were without any protocol violations that
might have had a signicant effect on the analysis, i.e. as per protocol
(PP) analysis.
The main analysis performed compared between the two groups of the
ITT population. Unadjusted comparison between the two groups was analysed using the Fishers exact test for dichotomous variables, Mann
Whitney U-test for continuous variables, Mantel Haenszel x 2-test for
ordered categorical variables and x 2-test for non-ordered categorical variables. The two groups could be either the two study groups or pregnancy/not pregnancy. Complementary unadjusted analyses were also
performed between the two groups on the PP population. A confounder
was dened as a baseline variable signicantly correlated with the outcome

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Materials and Methods

Hardarson et al.

91

NIR fails to predict pregnancy: an RCT

variable and signicantly different between the study arms. Continuous

variables were described with the mean, SD, median and range and categorical variables by number and percentages.
All signicance tests were two-sided and conducted at the signicance
level of 0.05.
To study if increasing the viability score increases the pregnancy rate, a
logistic regression analysis was performed.
A complementary analysis of the viability score versus pregnancy
outcome adjusted for confounders was performed with multiple logistic
regression.

IVF and embryo culture

NIR spectroscopy
In the NIR group, all GQE on Day 2 or Day 5 were analysed using a prototype NIR spectroscopy system intended for IVF use (Molecular Biometrics
Inc., Norwood, MA, USA). From the two or more GQE randomized to
the NIR group, a small aliquot (13 ml) was aspirated from the culture
medium droplet of each GQE and injected into a sample cell (Molecular
Biometrics Inc.) for analysis by the instrument. A sample cell was only
used for one embryo and then discarded. The instrument used the NIR
spectroscopic technology as previously described (Seli et al., 2007;
Botros et al., 2008; Scott et al., 2008). An algorithm had previously
been developed using 62 samples for the Day 2 and Day 5 algorithms;
both algorithms were trained from 50% fetal cardiac positive and negative
SET medium samples. The Day 2 algorithm contained four predictive
wavelet regions, two below the water band (1300 1400 nm) and two
above the water band (1500 1600 nm), while the Day 5 algorithm contained three predictive wavelet regions all below the water band
(1300 1400 nm). Both algorithms were developed on a different instrument from that used in the trial. When the NIR analysis had been performed on the GQE cohort, the embryo with the highest numerical
score was chosen for SET. The other GQEs were cryopreserved according

Results
The DSMB analysis was performed when 287 patients had been randomized. The analysis was performed earlier than planned due to a
slower enrolment rate than originally expected. After comparing the
two groups, the advice from the DSMB was to stop the study in
accordance with alternative one in the pre-set study protocol. Recruitment into the study was therefore stopped. During the study period, a
total of 1430 patients underwent IVF treatment at the clinic. A total of
1072 (75%) patients met the inclusion criteria and 972 (68%) patients
signed an informed consent. On the day of randomization, 327 (23%)
had two or more GQEs available for transfer and were randomized
(Fig. 1). The two major reasons for drop-outs between those that
were eligible and those that were randomized were either lack of
two or more GQEs at the day of transfer or administrative reasons.
Of the randomized patients, 164 were allocated to the NIR group
and 163 patients to the control group. No signicant differences
between these groups were found for patient and treatment characteristics (Table I). However, there were signicant differences in the
morphological quality of embryos transferred on Day 5 with higher
grades found in the control group, i.e. grade of expansion
(P 0.0003), ICM (P 0.0096) and TE (P 0.025). The degree of
fragmentation and cell number were also improved in the Day 2
control group; however, these failed to reach signicance. From all
embryos (n 484) randomized to the NIR group, spent medium
samples were successfully analysed by NIR spectroscopy for 454
(93.8%) embryos. For 12 patients, failed NIR analysis led to embryo
selection by morphology alone.
As shown in Table II, no signicant difference in the ongoing pregnancy rate per randomized patient was found between the NIR and
the control group for both the ITT and PP populations, 34.8 versus
35.6%, (P 0.97) and 36.2 versus 36.3% (P 1.00), respectively.
The proportional difference between the study groups mean was
20.8% [95% condence interval (CI) 11.4 to 10.2]. Furthermore,
no signicant difference was found between the two study groups
for the biochemical pregnancy rate (P 0.35) or the ongoing pregnancy rate for Day 2 (P 0.63) or Day 5 transfer (P 0.55).
When adjusting for morphology on Day 5, the odds ratio for the
primary efcacy variable, ongoing pregnancies or live birth was 1.06
(95% CI: 0.54 2.08, P 0.86) and the odds ratio for pregnancies
was 0.62 (95% CI: 0.31 1.26, P 0.19) between the two groups.
For the NIR group, viability scores of transferred embryos ranged
from 20.015 to 1.677. No statistical signicance was found for the
odds of biochemical pregnancy and ongoing pregnancy separated by
the day of embryo transfer in relation to the viability score, as
shown in Table III.

Discussion
The DSMB analysis resulted in a recommendation to stop the study in
accordance with the criteria set out in the study protocol. The analysis
showed that the study had a very small chance of showing any
improvement in the pregnancy rate by using the prototype NIR

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Stimulation protocols were performed as previously described (Wikland

et al., 2010). Briey, in the majority of cases (.70%), an antagonist
(Orgalutranw or Cetrotidew) was used as well as down-regulation with
either the GnRH agonist buseralin (Suprecurw; Hoescht, Germany) or
the GnRH antagonist nafarelin (Synarelaw; Pzer, USA) nasal spray.
Ovarian stimulation was achieved with recombinant FSH (Gonal-Fw;
Merck Serono, Germany or Puregonw, MSD, USA). hCG (Ovitrellew;
Merck Serono, Germany), 10 000 IU s.c., was administered when patients
had two or more follicles 18 mm or greater in mean diameter. Follicular
aspiration was performed using vaginal ultrasonography 36 38 h after
hCG administration.
Retrieved oocytes were rinsed and placed in Cook fertilization medium
(Cook Medical, Australia) at 378C, pH 7.35 + 0.5, 5% O2 and 90% N2.
Fertilization was achieved by IVF or ICSI, following standard techniques.
At 16 18 h after insemination (Day 1), fertilization was conrmed by
the presence of two pronuclei and zygotes were placed into individual droplets of 25 ml of Cook cleavage medium for culture to Day 2. On Day 2,
the embryos were graded according to standard laboratory methods
(Hardarson et al., 2001). A GQE was determined to be an embryo with
4 6 cells, ,20% fragmentation and without multinucleation. In general,
blastocyst culture was performed when the number of fertilized oocytes
exceeded four. If the embryos were to be cultured on to Day 5, they
were transferred on Day 2 into individual droplets of 25 ml of Vitrolife
CCM medium (Vitrolife Sweden AB, Sweden). The blastocysts were
graded according to Gardner and Schoolcraft (1999). A good quality
blastocyst was considered to be a 3BB or better. During the study
period no changes were made in either embryo culture and/or grading
systems.

to standard protocols used at the laboratory. In the control group, the

embryo selected for SET was solely based on morphological criteria
used routinely at our clinic.

92

Hardarson et al.

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Figure 1 Flow chart of patients for the study.

spectroscopy technique as an adjunct to morphology when compared

with morphology alone. Interestingly, when using the NIR technology
on Day 2 embryos, a minor non-signicant improvement of greater
than the 3% DSMB setting was observed, while Day 5 morphology
alone was better compared with NIR. However, when analysing the
combined data, no differences between the two groups were found
with regard to the ongoing pregnancy rate. Consequently, the

results from this study failed to support the concept that the NIR technique, in its current state, can improve implantation rates for SET. Our
results are somewhat surprising given the background of initial proof of
principle reports that suggested that the NIR technique should work
to select the best embryo for transfer (Hardarson et al., 2007; Seli
et al., 2007; Scott et al., 2008; Vergouw et al., 2008). Although the
use of the NIR technology showed some benet in retrospective

93

NIR fails to predict pregnancy: an RCT

Table I Demographics and baseline characteristics (ITT population).

NIR (n 5 164)

NIR control (n 5 163)

.............................................................................................................................................................................................
Female age

35.5 (4.2) (35.0, 26.044.0)

BMI

23.8 (4.0) (22.5, 17.837.0)

35.6 (4.2) (36.0, 24.0 44.0)

23.4 (3.5) (22.7, 18.4 39.3)

Number of previous IVF treatment cycles

0.61 (1.0) (0.0, 0.0 5.0)

0.61 (1.0) (0.0, 0.0 5.0)

Number of fertilized oocytes

5.91 (3.5) (5.0, 2.0 19.0)

5.93 (3.3) (5.0, 2.0 18.0)

ICSI

43.3% (71)

41.7% (68)

STD IVF

56.7% (93)

58.3% (95)

Method used to fertilize oocytes

Number of aspirated oocytes

9.22 (5.2) (8.0, 2.0 27.0)

9.07 (4.8) (8.0, 2.0 31.0)

Number of follicles aspirated

12.7 (6.4) (12.0, 3.0 35.0)

12.7 (6.1) (12.0, 4.0 40.0)

Number of GQE on day of transfer

2.96 (1.3) (3.0, 2.0 8.0)

2.92 (1.2) (3.0, 2.0 10.0)

0.0% (0)

1.8% (3)

100% (164)

98.2% (160)

Cetrotide

1.8% (3)

1.2% (2)

Orgalutran

72.0% (118)

76.1% (124)

Suprecur

18.9% (31)

15.3% (25)

Suprefact

0.0% (0)

1.2% (2)

Synarela

7.3% (12)

6.1% (10)

2059 (822) (2000, 6644850)

1960 (764) (1800, 700 4125)

53.0% (87)

50.9% (83)

47.0% (77)

49.1% (80)

Number of embryos transferred

Total dosage of gonadotrophins administered

Day of transfer

Medical cause of infertility

Male factor

30.5% (50)

32.5% (53)

Other

22.6% (37)

24.5% (40)

Tubal factor

11.0% (18)

9.2% (15)

Unexplained

36.0% (59)

33.7% (55)

Number of cells on transfer Day 2

4

90.8% (79)

97.5% (79)

6.9% (6)

1.2% (1)

2.3% (2)

1.2% (1)

66.7% (58)

80.2% (65)

33.3% (29)

19.8% (16)

33.8% (26)

8.9% (7)

64.9% (50)

89.9% (71)

1.3% (1)

1.3% (1)

61.0% (47)

81.0% (64)

39.0% (30)

19.0% (15)

45.5% (35)

64.6% (51)

54.5% (42)

35.4% (28)

Degree of fragmentation on transfer Day 2

Degree of expansion on transfer Day 5

Morphological grade of the ICM

Morphological grade of the TC cells

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Type of pituitary regulation

94

Hardarson et al.

Table II Outcome variables by group for ITT and PP populations.

ITT population

NIR
(n 5 164)

NIR control
(n 5 163)

P-value

Difference in proportions between

groups mean (95% CI)

.............................................................................................................................................................................................
Primary efcacy variable
Number of ongoing pregnancies or live births

34.8% (57)

35.6% (58)

0.97

20.8% (211.4 to 10.2)

Secondary efcacy variables

Number of pregnancies

54.3% (89)

48.5% (79)

0.35

5.8% (25.3 to 16.7)

Number of ongoing pregnancies or live births,

embryo transfer Day 2

31.0% (27)

26.5% (22)

0.63

4.5% (210.7 to 19.5)

Number of ongoing pregnancies or live births,

embryo transfer Day 5

39.0% (30)

45.0% (36)

0.55

26.0% (221.8; 9.5)

NIR
(n 5 152)

NIR Control
(n 5 160)

P-value

Difference in proportions between

groups mean (95% CI)

36.2% (55)

36.3% (58)

1.00

PP population

.............................................................................................................................................................................................
Primary efcacy variable
20.1% (211.2; 11.1)

Number of pregnancies

55.9% (85)

49.4% (79)

0.30

6.5% (24.6; 17.6)

Number of ongoing pregnancies or live births,

embryo transfer Day 2

33.3% (27)

27.2% (22)

0.49

6.2% (29.8; 21.9)

Number of ongoing pregnancies or live births,

embryo transfer Day 5

39.4% (28)

45.6% (36)

0.55

26.1% (221.9; 9.9)

For categorical variables % (n) is presented. For comparison between groups Fishers exact test was used for dichotomous variables. The condence interval for dichotomous variables
is the unconditional exact condence limits. If no exact limits can be computed the asymptotic Wald condence limits with continuity correction is calculated instead.

Table III Viability score versus pregnancy outcome,

NIR group (ITT population).
OR (95% CI)

P-value

Area
under the
curve

........................................................................................
Ongoing pregnancies or
live births, embryo
transfer Day 2

0.97 (0.19 5.07)

0.97

0.52

Ongoing pregnancies or
live births, embryo
transfer Day 5

4.22 (0.43 41.6)

0.22

0.57

studies, it appears that a wide scale application is at this stage not feasible. Instrument variability, threshold of the analytical signal in the
medium and instrument noise to signal may have all inuenced both
algorithm development and its utilization. Variability between different
clinics may also play a signicant role in the global application of this
technology.
It is not uncommon to observe discrepancies between proof of
principle studies and larger RCT studies. This concept was recently
experienced when also examining the use of preimplantation genetic
screening, using uorescence in situ hybridization, in women undergoing IVF (Staessen et al., 2004; Mastenbroek et al., 2007; Hardarson
et al., 2008). Preliminary ndings suggesting a benet of a technique
are often based on smaller retrospective studies. The problem with
the NIR spectroscopy for predicting both disease, and in this case
pregnancy, is mainly related to the creation of the algorithm to be

used. The algorithm itself can create models that inadvertently

conceal problems in a particular platform. In doing so, a method
that has been established and tested on both positive and negative
outcomes (Hardarson et al., 2007; Seli et al., 2007, 2010; Scott
et al., 2008) and then cross-validated (Ahlstrom et al., 2011) on a
larger scale can still be problematic as the variation can still lie
within the technical platform itself. This was evidenced by the subsequent withdrawal of the commercial version of the NIR instrument for
analysis of spent embryo culture medium due to the wide variability in
performance between clinics. Indeed, it has been proposed that rigorous criteria for accepting any test as a validated marker of embryonic
reproductive competence should be established, and practitioners
should be cautious about applying these tests clinically before the
availability of comprehensive and convincing evidence (Hollywood
et al., 2006; Scott and Treff, 2010). It can be argued that this trial in
itself has served as a rigorous assessment of this particular platform
and that modications are needed if this platform is to be accepted.
Prior to more widespread clinical application, a similar trial would
need to be performed.
In this study both Day 2 and Day 5 SET were included. Although not
signicant, the results indicate a possible benet of embryo selection
through addition of NIR on Day 2 transfer. If true, one explanation
might be that on Day 2 there are more embryos to choose from compared with Day 5, where a considerable de-selection has already
occurred. One limitation of our study design is that the morphology
in the control group was better than in the NIR group. This was
expected as we did not balance the study group morphology
through the randomization program. We decided against balancing
for the expected poorer embryo morphology in the NIR groups

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Number of ongoing pregnancies or live births

Secondary efcacy variables

95

NIR fails to predict pregnancy: an RCT

could be argued that the ability to select the best Day 5 embryo compared with one of the cohort of good quality Day 5 embryos does,
however, provide benet. It is apparent from Table II that even
though morphologically lower quality embryos were selected in the
NIR group, pregnancy rates were not signicantly altered. This again
draws attention to the need for secondary measures in addition to
morphology that will provide precise physiological information about
an individual embryo to facilitate selection for SET. The premise of
any technology examining metabolic, metabolomic or proteomic differences between embryos is that this information provides a
greater ability to establish which is the viable embryo within a
cohort of GQEs. From our results it would appear that the current
technology was unable to improve on selection of the most viable
embryo when compared with blastocyst morphology. It therefore
appears that the morphological criteria used to perform SET on
Day 5 are superior compared with a combination of morphology
and metabolomics. On the contrary, the same may not be true for
selection of a SET on Day 2 and the extra information supplied by
this platform may be benecial.
In conclusion, this study has shown that adding NIR spectroscopy to
embryo morphology in its present form does not improve the chance
of a viable pregnancy. The NIR technology appears to need further development before being used as an objective marker of embryo
viability.

Authors roles
All authors played a role in study conception and design. T.H., A.A.
and L.R. took part in collection and assembly of data. T.H., A.A.,
D.S. and M.W. carried out data analysis, interpreted the ndings
and drafted the manuscript. All authors critically reviewed and
approved the nal version of the manuscript.

Acknowledgements
The authors thank Nils-Gunnar Phersson for his and his colleagues invaluable help with the statistical part of the study. The authors also
thank staff members of Molecular Biometrics Inc. for their assistance.

Conict of interest
D.S. has shares in and consults for Molecular Biometrics Inc.

Funding
This study was supported by an unconditional grant from Molecular
Biometrics Inc., USA.

References
Ahlstrom A, Wikland M, Rogberg L, Barnett JS, Tucker M, Hardarson T.
Cross-validation and predictive value of near-infrared spectroscopy
algorithms for day-5 blastocyst transfer. Reprod Biomed Online 2011;
27:477 483.
Botros L, Sakkas D, Seli E. Metabolomics and its application for
non-invasive embryo assessment in IVF. Mol Hum Reprod 2008;
14:679 690.

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

mainly because we would have introduced a positive patient bias for

the NIR group, i.e. the patients in the NIR group would have had a
better embryo cohort, aside from the embryo transferred. Also, previous studies using the NIR technique did not show any correlation
between embryo morphology and ongoing pregnancy, in a good
cohort of embryos. Finally, we decided to use the same set-up as
many other similar RCTs have used when selecting embryos using a
method other than morphology to make it easier to compare these
studies (Staessen et al., 2004; Mastenbroek et al., 2007; Hardarson
et al., 2008). It is important to note that although the morphology
of the embryo that was chosen for transfer was poorer in the NIR
group, the cohort of embryos in both study groups were exactly the
same. Therefore, from a morphological point of view, the study
groups were the same and the NIR technique simply chose differently
from the human eye. If techniques that are being introduced to aid in
the selection of embryos for transfer do not choose differently from
the embryologist, then they will not become clinically useful.
Another limitation of the study is the possibility of bias introduction
due to the fact that only one in three eligible patients were randomized, which could raise questions about our ability to generalize
from the study ndings. The largest group (n 439) that was not randomized were patients that did not full the inclusion criteria of having
at least two GQE. We recognize that through this we have selected a
special group of good patients for the study and may have inadvertently set the bar higher. However, we were forced to design our study in
this manner as we would otherwise have had to include embryos for
possible embryo transfer that would have otherwise been discarded
(i.e. an embryo of poor morphological quality). That design,
however attractive it might have been, was unethical as we did not
beforehand know if the NIR technique would prove benecial. Thus,
our chosen criteria for randomization in this study has to be taken
into consideration when generalizing the benets of this technology
to other groups of IVF patients.
A major concern of the NIR method for selecting embryos for
transfer is the absence of a clear cause and effect, with regard to
changes in the functional groups measured and the underlying biological processes. If the method had been clinically useful, one might have
overlooked this limitation. However, considering the results in this
study, it seems even more important to identify the factors being measured and their underlying biological mechanisms in order to improve
the method. Some efforts have been made to measure and identify
more distinct markers using Raman and mass spectroscopy (Seli
et al., 2008). However, further studies are still required if we are to
accurately and reliably use the NIR or any other spectroscopic
method to measure molecular changes and extrapolate on how
those changes reect embryo viability. It is beyond question that
markers do exist in the spent embryo culture medium indicative of viability. Recently, a further proof of principle study by Gardner et al.
(2011) has revalidated the importance of glucose uptake as a predictive marker. The major benets of a non-invasive type of technology is
the fact that the technology can be used on spent media and the time
taken to assess the samples is very short, making it possible to
perform the analysis just prior to embryo transfer.
The design of this study also inadvertently showed the importance
of morphology for selecting the most viable embryo. Selection of the
single best morphological embryo compared with selection of one of
the GQEs did not show any signicant benet for Day 2 transfers. It

96

Scott R, Seli E, Miller K, Sakkas D, Scott K, Burns DH. Noninvasive

metabolomic proling of human embryo culture media using Raman
spectroscopy predicts embryonic reproductive potential: a
prospective blinded pilot study. Fertil Steril 2008;90:77 83.
Scott RT Jr, Treff NR. Assessing the reproductive competence of individual
embryos: a proposal for the validation of new -omics technologies.
Fertil Steril 2010;94:791 794.
Seli E, Sakkas D, Scott R, Kwok SC, Rosendahl SM, Burns DH. Noninvasive
metabolomic proling of embryo culture media using Raman and
near-infrared spectroscopy correlates with reproductive potential of
embryos in women undergoing in vitro fertilization. Fertil Steril 2007;
88:1350 1357.
Seli E, Botros L, Sakkas D, Burns DH. Noninvasive metabolomic
proling of embryo culture media using proton nuclear magnetic
resonance correlates with reproductive potential of embryos in
women undergoing in vitro fertilization. Fertil Steril 2008;90:
2183 2189.
Seli E, Vergouw CG, Morita H, Botros L, Roos P, Lambalk CB,
Yamashita N, Kato O, Sakkas D. Noninvasive metabolomic proling
as an adjunct to morphology for noninvasive embryo assessment in
women undergoing single embryo transfer. Fertil Steril 2010;
94:535 542.
Staessen C, Platteau P, Van Assche E, Michiels A, Tournaye H, Camus M,
Devroey P, Liebaers I, Van Steirteghem A. Comparison of blastocyst
transfer with or without preimplantation genetic diagnosis for
aneuploidy screening in couples with advanced maternal age: a
prospective randomized controlled trial. Hum Reprod 2004;
19:2849 2858.
Vergouw CG, Botros LL, Roos P, Lens JW, Schats R, Hompes PG,
Burns DH, Lambalk CB. Metabolomic proling by near-infrared
spectroscopy as a tool to assess embryo viability: a novel,
non-invasive method for embryo selection. Hum Reprod 2008;
23:1499 1504.
Wikland M, Hardarson T, Hillensjo T, Westin C, Westlander G, Wood M,
Wennerholm UB. Obstetric outcomes after transfer of vitried
blastocysts. Hum Reprod 2010;25:1699 1707.

Downloaded from http://humrep.oxfordjournals.org/ by guest on November 17, 2014

Brison DR, Houghton FD, Falconer D, Roberts SA, Hawkhead J,

Humpherson PG, Lieberman BA, Leese HJ. Identication of viable
embryos in IVF by non-invasive measurement of amino acid turnover.
Hum Reprod 2004;19:2319 2324.
Gardner DK, Leese HJ. Non-invasive measurement of nutrient uptake by
single cultured pre-implantation mouse embryos. Hum Reprod 1986;
1:25 27.
Gardner D, Schoolcraft W. In-vitro culture of human blastocyst. In:
Jansen R, Mortimer D (eds). Towards Reproductive Certainty: Infertility
and Genetics Beyond. Carnforth: Parthenon Press, 1999.
Gardner DK, Wale PL, Collins R, Lane M. Glucose consumption of single
post-compaction human embryos is predictive of embryo sex and live
birth outcome. Hum Reprod 2011.
Hardarson T, Hanson C, Sjogren A, Lundin K. Human embryos with
unevenly sized blastomeres have lower pregnancy and implantation
rates: indications for aneuploidy and multinucleation. Hum Reprod
2001;16:313 318.
Hardarson T, Rogberg L, Ahlstrom A, Wikland M, Rose P, Burns D.
Non-invasive assessment of embryo reproductive potential:
metabolomic proling of biomarkers of oxidative metabolism (OM)
during embryo development. Fertil Steril 2007;88:S307 S308.
Hardarson T, Hanson C, Lundin K, Hillensjo T, Nilsson L, Stevic J, Reismer E,
Borg K, Wikland M, Bergh C. Preimplantation genetic screening in women
of advanced maternal age caused a decrease in clinical pregnancy rate: a
randomized controlled trial. Hum Reprod 2008;23:28062812.
Hollywood K, Brison DR, Goodacre R. Metabolomics: current
technologies and future trends. Proteomics 2006;6:4716 4723.
Katz-Jaffe MG, McReynolds S, Gardner DK, Schoolcraft WB. The role of
proteomics in dening the human embryonic secretome. Mol Hum
Reprod 2009;15:271 277.
Mastenbroek S, Twisk M, van Echten-Arends J, Sikkema-Raddatz B,
Korevaar JC, Verhoeve HR, Vogel NE, Arts EG, de Vries JW,
Bossuyt PM et al. In vitro fertilization with preimplantation genetic
screening. N Engl J Med 2007;357:9 17.
Renard JP, Philippon A, Menezo Y. In-vitro uptake of glucose by bovine
blastocysts. J Reprod Fertil 1980;58:161 164.

Hardarson et al.