Sage Journals: Discover world-class research

Abstract

Time-lapse (TL) embryo monitoring is the latest technology that is proposed for embryo evaluation and selection for transfer. TL technology enables us to collect significantly more information about the in vitro development of the embryos that can be obtained through the daily-once evaluation under the light microscope. In addition, the embryos do not need to be removed from the culture environment for this. The extra morphokinetic information and the undisturbed culture conditions could both be beneficial for the cultured embryo cohort. Many morphokinetic parameters have been tested in relation to variety of laboratory (e.g. blastocyst development) and clinical (implantation and live-birth rate) outcomes. Most of these studies are retrospective in nature and suffer from methodological problems (heterogeneous patient population, culture conditions not standardized, and small sample size). Several groups attempted to build algorithms, however, have not yet been confirmed externally as attempts so far could not reproduce the expected predictive abilities. Therefore, these algorithms cannot be universally accepted. The latest algorithm proposed for embryo selection was developed based on data from 24 clinics using local stimulation and laboratory procedures. It groups embryos into five categories (KIDScore) based on in and out of range kinetic events. The algorithm was tested in subsets of patients using various fertilization methods or culture conditions and its predictive ability remained the same. The authors, therefore, feel comfortable to recommend it for routine use in any laboratory using TL technology. There is, however, still limited prospective, randomized trial data testing the algorithms. This article reviews TL technology, retrospective and prospective reports on various morphokinetic parameters, and the benefits and shortcomings of currently available algorithms.

Keywords

Time-lapse technology morphokinetic parameters algorithm known implantation data in vitro fertilization

Introduction

The step limiting the success of in vitro fertilization (IVF) the most is implantation. A healthy embryo, a properly built-up endometrium and appropriate synchronization in between them are required for this step to proceed successfully. In a typical IVF cycle, a cohort of embryos are created in vitro and typically 1 or 2 are selected for transfer. Depending primarily on the age of the woman, on average, 5–40% of them implant and therefore around 1/3 of the cycles result in a pregnancy (European IVF-Monitoring Consortium (EIM), 2016).

It has been realized for some time that the daily-once evaluation of the actual morphology of the developing embryos provides us the limited information about their overall status and therefore does not allow proper selection for transfer. In order to improve our ability to identify the embryo(s) with the highest implantation potential, various methods (extended culture to the blastocyst stage (Glujovsky, 2016), metabolomics, proteomics, granulosa cell gene expression profiling, preimplantation genetic screening (PGS) (Montag et al., 2013)) have been tested. Among them, PGS seemed to be the most effective as the identification of the embryo with a normal, euploid chromosome content has been shown to improve pregnancy rate in younger, good prognosis patients (Dahdouh et al., 2015).

Time-lapse (TL) embryo monitoring is the latest tool that may aid the embryologist’s assessment of the developing embryos. TL technology allows us to continuously monitor embryonic development without the need to remove the embryos from the optimal culture conditions. The extra information obtained through TL monitoring gives us a more detailed knowledge about the kinetic and morphologic changes/abnormalities an embryo undergoes in vitro. The kinetic events can be precisely timed and these timings/intervals can be correlated with various stages of embryonic development, implantation, and live birth (Kovacs, 2014). Ultimately, the kinetic and morphologic parameters obtained through TL could be used to build algorithms that can help to choose the fittest embryo for transfer.

TL technology

There are various TL equipment in use (Primo Vision, Vitrolife AB, Sweden (PV) time-lapse system, embryoscope (ES), early embryo viability assessment (EEVA)) (Kovacs, 2014). They are all built around a similar concept. A TL unit is made up of a camera that takes a picture of the developing embryos at preset (10–20 min) intervals and is connected to a microscope system. This complex either has to be placed into a standard incubator (PV, EEVA) or is part of an incubator already (ES). Embryos in the TL system need to be identified and followed up individually. This can be achieved by culturing the embryos in individual microwells with no connection between the embryos (Embryoslide) or by the use of so called multiwell dishes (Primo Vision Culture Dish, EEVA dish) when each embryo is placed in individual microwells (9–16-well dishes) and they are cultured under a single drop of culture medium. This latter system allows individual embryo tracking but also provides the benefit of group culture via enhanced auto- and paracrine effects (Vajta et al., 2008).

The pictures captured by the camera are then processed by the appropriate software. This way a short film is created by connecting the pictures that can be rewound and fast-forwarded, and in the case of the PV or ES systems, embryos can be evaluated in several focal planes (Kovacs, 2014).

TL parameters

TL monitoring allows us to closely follow the embryos from fertilization up until the transfer. Events of the pronuclear phase, precise timing of cell divisions, duration and synchrony of the cell cycles, the sometimes transient changes of morphology (e.g. fragmentation), timing of compaction, blastocyst formation, and expansion and blastocyst dynamics can be precisely timed (Figures 1 and 2). In addition to the normal events of embryo development, abnormal cellular events can be detected as well that otherwise easily could be missed by relying on the daily-once observation. It has been shown that direct cleavage from 1 to 3 cells, multinucleation, and uneven blastomere size at the 2- to 4-cell stage are strong negative predictors of implantation (Meseguer et al., 2011).

Figure 1.

Embryo development from 2PN to blastocyst stage and the various terminology used in the different papers for certain developmental events (with permission from Reproductive Biology and Endocrinology: Kovacs (2014)).

Figure 2.

Definitions of kinetic TL parameters.

Heterogeneity in the markers identified as predictors of various outcomes

As TL technology became available, several groups have started to use it and started to correlate certain kinetic events with laboratory and clinical outcomes. These studies are mostly retrospective and rely on building databases of various kinetic markers that can subsequently be correlated with blastocyst formation, implantation, clinical pregnancy or live birth. In order to associate these parameters with a given clinical outcome, the analysis has to be limited to embryos with known implantation data (KID embryos). This is straightforward when a single embryo is transferred (SET); in cycles with a double embryo transfer (DET), only those can be considered for analysis when either both embryos implant or neither of them do.

When the outcomes of these studies are evaluated, their heterogeneity has to be considered as well. Some studies included fresh cycles using only own oocytes; others included cycles with fresh autologous or donated oocytes, while others included frozen oocytes or cryopreserved fertilized oocytes as well. The culture conditions are not uniform; different culture media are used by the various research groups, the amount of out-of-incubator handling is likely to differ, and the oxygen (O₂) concentration is not standardized either. The day of transfer varies from day 2 to day 5 and there are reports in which embryos are not transferred just cultured and observed to a certain developmental stage. Obviously, with a shorter culture period, we cannot enjoy the full benefit of the TL technology, as less morphokinetic data is generated and the out-of-incubator handling of embryos is similar to the standard methods (for details of the studies, see Tables 1 –3.).

Table 1.

Results of RCTs using TL technology.

Author	Study type	Participants	Embryo selection	Outcome	TL unit; O₂ concentration
Kahraman et al. (2013)	RCT; randomization at retrieval to TL culture vs. conventional culture; eSET on d5	Young good responder patients (N = 64)	Selection: TL group d5 morphology + t ₅, S ₂, CC₂ (Meseguer hierarchical model (2011)) vs. d5 morphology in controls	PR and miscarriage rates similar	Embryoscope; 5% O₂
Rubio et al. (2014)	RCT: randomization the day before retrieval; TL culture vs. conventional culture	Patients using own oocytes (20–38 years, 1st, second cycle), or egg donation (fresh + frozen; (N = 930)	Selection: TL group t ₅, S ₂, CC₂ (Meseguer hierarchical model (2011) vs. d3 or d5 morphology	OPR higher in TL group (51.4% vs. 41.7%); early pregnancy loss rate lower in TL (16.6% vs 25.8%)	Embryoscope; 21% O₂
Park et al. (2015)	RCT: randomization (2:1) after retrieval, TL incubator vs. conventional culture	Patients under 40 years, 1st and second cycle (N = 364)	Day 2 transfer, selection based on morphology in both groups	Primary endpoint no. of good quality embryos on day 2: no significant difference; secondary outcome: PR similar, pregnancy loss rate higher in TL (33.3% vs. 10.2%), OPR lower in TL (20% vs. 28.2%)	Embryoscope; 21% O₂
Goodman et al. (2016)	RCT: randomization at retrieval to morphologic evaluation vs. morphologic evaluation + morphokinetic parameters, TL culture both groups	Patients 18–43 years, own oocytes (N = 300)	Selection: standard morphology vs. standard morphology and then rank based on CC₂, t ₅, S ₂, S ₃, tSB	PR and IR similar per patient; embryos with known implantation tSB predictive of implantation	Embryoscope; 5.5% O₂
Matyas et al. (2015)	RCT: randomization prior to stimulation start, conventional culture + morphologic evaluation vs. TL observation (same type of incubator/incubation) + morphokinetic evaluation; third nonrandomized arm two BC ET	Patients good prognosis <36 years, 1st and second cycle (N = 161 + 257 in 2BC-ET group)	Selection: d5 morphology vs. CC₁, CC₂, S ₁, S ₂, t ₅ + BC morphology; 2 BC-ET: d5 morphology	PR: eSET TL vs. DET similar (46.2% vs. 52.1%, p = NS), DET vs. eSET control (52.1% vs. 34.6%, p = 0.039); eSET TL and eSET control significantly better perinatal outcome (GA [gestational age] at delivery, birth weight) when compared to DET)	Primo Vision TL system; 5% O₂

RCT: randomized controlled trial; TL: time-lapse; eSET: elective single embryo transfer; d5: day 5; PR: pregnancy rate; OPR: ongoing pregnancy rate; O₂: oxygen; BC: blastocyst; ET: embryo transfer; DET: double embryo transfer; tSB: time to start of blastulation.

Table 2.

Results of prospective and retrospective cohort studies evaluating TL technology.

	Type of study	Outcome studied	Parameter studied	Outcome	TL system
Wong at el. (2010)	Cohort study of frozen-thawed embryos (cryopreservation at zygote stage)	BC development (embryos not transferred)	S ₁, CC₂, S ₂	Studied parameters when in optimal range predictive of BC development	EEVA
Meseguer et al. (2011)	Retrospective analysis, normal responder patients using own oocytes and donor oocyte cycles; 247 embryos with known implantation	Implantation rate	CC₂ <5 h, multinucleation and uneven blastomere size excluded, t ₅, CC₂, S ₂ (Meseguer decision tree based on in and out of range kinetic parameters)	Studied parameters when in optimal range predictive of implantation	Embryoscope; 21% O₂
Rubio et al. (2012)	Retrospective study, donor oocytes and own oocytes (N = 1659 transferred embryos)	Implantation rate	Direct cleavage: CC₂ <5 h	Embryos with CC₂ <5 h min. chance (1.2%) to implant	Embryoscope; 21%O₂
Cruz et al. (2012)	Retrospective analysis, donor oocyte treatments, ET on d5	BC development; implantation rate (embryos with known implantation)	t ₅, S ₂ (four categories based on in and out of range values); uneven blastomeres at 2-cell stage, direct cleavage 1–3 cells	Symmetric blastomeres and no direct cleavage predictive of BC development, t ₅/S ₂ categories not predictive of implantation	Embryoscope; 21% O₂
Conaghan et al. (2013)	Prospective study to predict usable blastocyst formation by d3	Blastocyst development rate	S ₁, CC₂, S ₂	CC₂, S ₂ when in optimal range predictive of usable blastocyst development	EEVA, O₂ concentration not specified
Meseguer et al. (2012)	Retrospective study, own and donated oocytes, TL incubation vs. standard incubation	Clinical pregnancy rate	Standard morphology vs. t ₅, CC₂, S ₂ (Meseguer decision tree) in TL cycles	20.1% average improvement in CPR	Embryoscope; 21% O₂
Azzarello et al. (2012)	Prospective cohort study of 159 zygotes from women under 39 years	Live birth (embryos with known implantation data)	Pronuclear fading, PNB	PNB is higher when transfer results in live birth; no live birth when PNB <20 h:45 min	Embryoscope; 5% O₂
Chamayou et al. (2013)	Retrospective analysis, patients <40 years	Implantation, clinical pregnancy	Various kinetic parameters	Predictive of BC development: t ₁, t ₂, t ₄, t ₇, t ₈, time to visible pronuclei, S ₃; parameter predictive of implantation: CC₃	Embryoscope; 5% O₂; different parameters predict BC development and implantation
Dal Canto et al. (2012)	Retrospective analysis of TL data, women 27–42 years, d3 and d5 ET (n = 71 cycles)	BC development	Cleavage times	Up to 6-cells no difference; t ₇ and t ₈ are shorter in embryos that reached BC stage; t ₈–t ₄ and t ₈−t ₅ are shorter in embryos that reached BC stage; t ₈ shorter in embryos that implanted but t ₅ did not differ	Embryoscope; 5% O₂
Basile et al. (2015)	Two phase study (I: algorithm building N = 765 cycles), II: algorithm testing (N = 885 cycles); donor oocytes, own oocytes; d3 ET	Implantation	Direct cleavage, multinucleation, uneven blastomeres; t ₂, t ₃, t ₄, t ₅, CC₂, S ₂, only embryos with known implantation data	Phase I: t ₃, CC₂ and t ₅ most relevant (based on in and out of range eight categories created); phase II: significant decline in implantation as moving from all three parameters in range to none in range	Embryo scope; 21% O₂
Siristatidis et al. (2015)	Prospective cohort study of 239 ICSI cycles (169 standard culture and morphology based selection vs. 70 cycles using TL culture and selection based on kinetic markers	Clinical and ongoing PR; live-birth rate	t ₂, CC₂, t ₃, S ₂, t ₄, CC₃, t ₅, S ₃, t ₈ (in TL group the embryo(s) with the most in range parameters was selected for transfer	Clinical pregnancy rate TL vs. control: 65.7% vs. 39.0% (p < 0.001); ongoing pregnancy TL vs. control: 55.7% vs. 31.3% (p < 0.001); live-birth TL vs. control: 45.7% vs. 28.4% (p = 0.01)	Primo Vision; atmospheric O₂
Motato et al. (2016)	Retrospective analysis, three phases: (1) algorithm building to predict BC, (2) algorithm building to predict implantation, (3) validation of implantation algorithm; own and donated oocytes	BC development	t ₂, t ₃, t ₄, t ₅, t ₆, t ₇, t ₈, t ₉, tM, tBC, t expand BC, t hatching BC, t ₃−t ₂, t ₅−t ₃, t ₅−t ₂, t ₈−t ₅	For BC development best model is based on tM and t ₈−t ₅ (four categories based on in and out of range) BUT little utility to predict BC; for IR best model is based on tEB and t ₈−t ₅ (four categories based on in and out of range) poor performance.	Embryoscope; O₂ not specified
VerMilyea et al. (2014)	Retrospective analysis based on data from six clinics, 331 embryos with known implantation data; fresh IVF, ICSI cycles	Clinical pregnancy, implantation	CC₂, S ₂; CC₂: 9.33≤ and ≤11.45 h and S ₂:≤1.73 h; two category results: EEVA high when both in range and EEVA low when one or both out of range; three category output: high: CC₂ and S ₂ in range; EEVA medium: CC₂ 9.33≤ and ≤12.65 h and S ₂: ≤4 h; EEVA low: out of the above ranges	Two category results: EEVA high vs. low IR: 37% vs. 23% (p = 0.003); three category results: high vs. medium vs. low IR: 37% vs. 35% vs. 15% (p: significance between high vs. low and medium vs. low)	EEVA system; different clinic specific protocols, culture medium; O₂ concentration
Milewski et al. (2015)	Retrospective analysis of embryos that developed to BC stage (n = 156) vs. those that did not (n = 276)	BC development	t ₂,₃,_4,5; CC₂; S ₂	A score created based on t ₂, t ₅ and CC₂ is predictive of BC development	Embryoscope; 5% O₂
Petersen et al. (2016)	Retrospective analysis; data from 24 clinics	Implantation	t ₃-tPNf; t ₃; (t ₅−t ₃)/(t ₅−t ₂); cell count at 66 h	5 scores assigned based on TL parameters and cell count at 66 h; sevenfold increase in implantation rates from score 1 to score 5; model is predictive regardless of IVF vs. ICSI or low O₂ vs. atmospheric O₂	TL system not specified; both 5% and atmospheric O₂ concentration

BC: blastocyst; ET: embryo transfer; d5: day 5; d3: day 3; CPR: clinical pregnancy rate; PNB: pronuclear break down; IVG: in vitro fertilization; ICSI: intra-cytoplasmic sperm injection; TL: time-lapse; EEVA: early embryo viability assessment.

Table 3.

Results of studies evaluation aneuploidy and TL technology.

	Type of study	Outcome studied	Parameter studied	Outcome	TL system
Campbell et al. (2013a: 26)	Retrospective analysis, patients undergoing ICSI-PGS	Aneuploidy	Initiation of compaction (tSC), start of blastulation (tSB), full blastocyst development (tB)	All parameters delayed in aneuploidy embryos	Embryoscope; 5% O₂
Campbell et al. (2013b: 27)	Retrospective analysis using aneuploidy risk model based on tSB and tB	Implantation (embryos with known implantation)	Low risk: tSB <96.2 h, tB <122.9 h; medium risk: tSB ≥96.2 h, tB <122.9 h; high risk: tB ≥122.9 h	High risk: none implanted; low risk: 74% increase in implantation rate when compared to all three risk categories	Embryoscope; 5% O₂
Basile and del Carmen Nogales (2014)	Retrospective analysis of patients undergoing PGS	Euploidy–aneuploidy risk based in TL parameters	t ₅−t ₂ and CC₃; four categories established based on in and out of range values	The proportion of euploid embryos decreases across the categories (highest when t ₅−t ₂ and CC₃ are in range and lowest when both out of range)	Embryoscope
Chavez et al. (2012)	Cohort of frozen-thawed zygotes cultured to d2	Aneuploidy	S ₁, CC₂, S ₂	Majority of aneuploidy embryos display time intervals outside the normal range; aneuploid embryos more likely to be fragmented	EEVA; 5% O₂

EEVA: early embryo viability assessment; TL: time-lapse; PGS: preimplantation genetic screening.

Finally, the patient populations studied are rather heterogeneous as well. The significance of this may be limited though. Ultimately, we wish to differentiate healthy, ready to implant embryos from the unhealthy cohort. Healthy embryos (when cultured under similar conditions) should be more likely to follow a strict, normal kinetic developmental pattern regardless of parental characteristics.

Due to the heterogeneity of the studies, it should not be surprising that a wide variety of kinetic markers have been associated with the various clinical outcomes (Tables 1 –3). Sometimes, even the same group identified various markers as “predictive” when a different outcome was studied (e.g. blastocyst development vs. implantation). One of the earliest studies in which an algorithm was proposed to identify the embryos with a better chance to implant incorporated t ₅, CC₂, and S ₂ into the hierarchical model and showed that those embryos that had kinetic parameters in the optimal ranges were more likely to implant (Meseguer et al., 2011). In a later study, Cruz et al. (2012) categorized embryos based on in or out of range t ₅ and S ₂ values but found no significant difference in the implantation rates across the four kinetic categories of embryos. Basile et al. (2015) proposed a different hierarchical model in which S ₂ was replaced by time to the 3-cell stage (t ₃). This final hierarchical model was built using in and out of range t ₃, CC₂, and t ₅ parameters to predict implantation. These studies were published by the same group and were carried out in the same chain of clinics. The patient population involved is rather heterogeneous as fresh autologous and donated oocytes as well as the use of frozen oocytes were included to build the database. The sample size of the studies (from a few hundred to the thousands) is likely to influence the predictive value of the various studied markers. Very importantly though, these authors have identified markers that are associated with minimal chance of success and these parameters are therefore proposed as deselection markers. Rubio et al. (2012) have shown that embryos with very short interval between the 1- and 3-cell stage (<5 h) have a minimal chance to implant. Furthermore, it has been shown that uneven blastomere size at the 2-cell stage and multinucletaion at the 4-cell stage are strong negative predictors of implantation (Meseguer et al., 2011). These deselection markers should be part of an algorithm built.

The parameters identified as significant are influenced by the technology itself. The EEVA system using dark field technology and an automated software analysis can reliably follow embryos up to the second and third cell cycle and therefore their algorithm depends on early markers (Conaghan et al., 2013; Wong et al., 2010). Wong et al. (2010), using frozen-thawed fertilized oocytes, found S ₁, CC₂, and S ₂ to be predictive of blastocyst formation. More relevant clinical outcomes could not be studied, as the embryos were not transferred. Later on, Conaghan et al. (2013) using the EEVA system have confirmed the improved ability to predict development to the blastocyst stage using the same early kinetic markers. Clinical outcome was not discussed in this study either.

There are groups that evaluated morphokinetic markers to predict embryonic aneuploidy. Campbell et al. (2013a, 2013b) published two papers on this topic. In their first study, they showed that late kinetic events (time to start of compaction, time to start of blastocyst formation, and time to blastocyst formation) were all delayed in aneuploid embryos. Early markers, however, failed to correlate with genetic health (Campbell et al., 2013a). In a subsequent study, they tested the predictive ability of time intervals (time to start of blastulation (96.2 h) and time to full blastocyst formation (122.9 h)) and proposed low-, medium-, and high-risk categories for aneuploidy based on them (Campbell et al., 2013b). Chavez et al. (2012) studied embryos obtained from frozen-thawed fertilized oocytes and found that euploid embryos followed a much tighter early development pattern as S ₁, S ₂, and CC₂ parameters were more homogeneous when compared to aneuploid embryos. Basile and del Carmen Nogales (2014) found t ₅–t ₂ and t ₅–t ₃ as the most reliable parameters to predict euploidy and built a model using in and out of ranges of these parameters to predict healthy chromosome content.

Algorithms predicting implantation

Once those morphokinetic markers that correlate with a given clinical outcome have been identified they need to be built into an algorithm that could improve our ability to identify the embryos with a higher chance to implant and result in a pregnancy/live birth (Table 4).

Table 4.

TL algorithms predicting clinical outcome (implantation or pregnancy rate).

	Type of study	Outcome studied	Parameter studied	Outcome
Meseguer et al. (2011)	Retrospective analysis (embryos with known implantation data)	Implantation	t ₅:48.8–56.6 h; CC₂: 11.9 h; S ₂ ≤0.76 h (in and out of range “Meseguer hierarchical model”); embryo selection: abnormal morphology exclude (F)/direct cleavage, multinucleation, and uneven blastomere size exclude (category E); A+: t ₅, CC₂, S ₂ all in range; A−: t ₅, S ₂ in CC₂ out of range; B+: t ₅ in, S ₂ out and CC₂ in range; B−: t ₅ in, S ₂ out, CC₂ out of range;C+: t ₅ out, S ₂ and CC₂ in range; C−: t ₅ out, S ₂ in, CC₂ out of range; D+: t ₅ out, S ₂ out, CC₂ in range; D−: t ₅, S ₂, CC₂ all out of range	Studied parameters when in optimal range predictive of implantation
Campbell et al. (2013: 27)	Retrospective analysis (embryos with known implantation)	Implantation	tSB, tB; low risk: tSB <96.2 h, tB <122.9 h; medium risk: tSB ≥96.2 h, tB <122.9 h; high risk: tB ≥122.9 h	High risk: none implanted; low risk: 74% increase in implantation rate when compared to all three risk categories
Rubio et al. (2014)	RCT	Ongoing pregnancy; pregnancy loss	t ₅, S ₂, CC₂ selection: TL group t ₅, S ₂, CC₂ (“Meseguer hierarchical model” (2011) vs. control: d3 or d5 morphology	OPR higher in TL group (51.4% vs. 41.7%); early pregnancy loss rate lower in TL group (16.6% vs. 25.8%)
VerMilyea et al. (2014)	Retrospective analysis (multiple IVF clinics, embryos with known implantation data)	Clinical pregnancy, implantation	CC₂, S ₂; CC₂: 9.33≤ and ≤11.45 h and S ₂:≤1.73 h; two category results: EEVA high when both in range and EEVA low when one or both out of range; three category output: high: CC₂ and S ₂ in range; EEVA medium: CC₂ 9.33≤ and ≤12.65 h and S ₂: ≤4 h; EEVA low: out of the above ranges	Two category results: EEVA high vs low IR: 37% vs. 23% (p = 0.003); three category results: high vs. medium vs. low IR: 37% vs. 35% vs. 15% (p: significance between high vs. low and medium vs. low)
Basile et al. (2015)	Retrospective analysis (embryos with known implantation data)	Implantation	t ₃: 34–40 h, t ₅: 45–55 h, CC₂: 9–12 h (in and out of range); embryo selection: abnormal morphology exclude (F)/direct cleavage, multinucleation, and uneven blastomere size exclude (category E); A+: t ₃, CC₂, t ₅ all in range; A−: t ₃, CC₂ in t₅ out of range; B+: t ₃ in, CC₂ out and t ₅ in range; B−: t ₃ in, CC₂ out, t ₅ out of range; C+: t ₃ out, CC₂ and t ₅ in range; C−: t ₃ out, CC₂ in, t ₅ out of range; D+: t ₃ out, CC₂ out, t ₅ in range; D−: t ₃ in, CC₂ out, t ₅ out of range	Significant decline in implantation as moving from all three parameters in range to none in range
Petersen et al. (2016)	Retrospective analysis (multiple IVF clinics, embryos with known implantation data)	Implantation	tPNf, t ₃, t ₅ and equations (A and B) based on them: A: t ₃-tPNf; t _3: ≥42.91; B: (t ₅−t ₃)/(t ₅−t ₂); cell count at 66 h; Score 1: equation A <11.48; Score 2: equation A >11.48 and t ₃: ≥42.91; Score 3: equation A > 11.48; t ₃ < 42.91 and equation B < 0.3408; Score 4: A >11.48; t ₃ <42.91; equation B > 0.3408 and ≥0.5781 OR equation B >0.3408 and <0.5781 cell number at 66 h <8; Score 5: A > 11.48; t ₃ < 42.91; equation B > 0.3408 and <0.5781 and cell number >8 at 66 h	5 scores assigned based on TL parameters and cell count at 66 h; sevenfold increase in implantation rates from score 1 to score 5; model is predictive regardless of IVF vs. ICSI or low O₂ vs. atmospheric O₂

IVF: in vitro fertilization; ICSI: intra-cytoplasmic sperm injection; TL: time-lapse.

The first algorithm to predict implantation was published by Meseguer et al. (2011). 247 embryos with KID were available to build the algorithm. According to their description, those embryos with clear morphologic abnormalities should be discarded and should not be considered for transfer. Those embryos with direct cleavage 1–3 cells, multinucleation, and uneven blastomere size at the 2- to 4-cell stage are not recommended for transfer either (deselection markers). The remaining embryos were split into eight kinetic categories (A+ (highest chance to implant), A−, B+, B−, C+, C−, D+, and D− (lowest chance to implant)) first based on in and out of range of t ₅ (48.8–56.6 h), then in and out of range of S ₂ (≤0.76 h), and finally based on in and out of range of CC₂ (≤11.9 h). About 66% of the embryos in the best, A+ category implanted, while only 15% of those in the lowest D− category implanted successfully. Only 8% of those embryos that were deselected based on exclusion criteria implanted. As mentioned earlier, Campbell et al. published two papers assessing late TL kinetic markers and aneuploidy and implantation (Campbell et al., 2013a, 2013b). In their first study, time to start of blastulation and time to full blastocyst development emerged as predictive for euploidy. In a subsequent study, a retrospective analysis based on 69 cycles, they showed that none of those embryos considered high risk for aneuploidy based on time to full blastocyst formation ≥122.9 h implanted, while 72.7% of those identified as low risk (time to start of blastulation <96.2 h and time to full blastocyst formation <122.9 h) implanted (Campbell et al., 2013b). Basile et al. (2015) proposed a different algorithm based on retrospective analysis of morphokinetic data of 1137 embryos with known implantation outcome. Embryos considered nonviable based on morphology were discarded (n = 55) and embryos showing any of the deselection criteria according to the Meseguer algorithm (n = 197) were suggested to be excluded too. The remaining 885 embryos were first split based on in and out of range of t ₃ (34–40 h), then based on in and out of range of CC₂ (9–12 h), and finally based on in and out of range of t ₅ (45–55 h) (A+ (highest implantation potential, A−, B+, B−, C+, C−, D+, D− (lowest implantation potential)). About 32% of A+ embryos implanted, while only 19% of those categorized as D−. The implantation rate was 17% among those embryos identified based on deselection criteria (Basile et al., 2015). Motato et al. (2016) proposed a hierarchical model to predict blastocyst implantation based on retrospective analysis of 832 blastocysts with KID. Two markers were included in the algorithm; time to expanded blastocyst ≤112.9 h versus >113 h and t ₈₋₅: ≤5.67 versus ≥5.68 h (categories A, B, C, and D). Implantation rates decreased across the four groups from 72.7% in category A to 39.7% in category D. It is interesting to point out that this study had two parts and the other part in which markers to predict blastocyst formation were searched identified different markers as predictive of blastocyst formation. Those markers, however, were not able to predict implantation (Motato et al., 2016).

It should be mentioned here that while several groups consider multinucleation as a negative predictor of implantation/pregnancy, there are reports that disagree with this. Balakier et al. (2016), in a retrospective analysis, evaluated the impact of multinucleation on embryonic euploidy and implantation/pregnancy rates. They found that multinucleation was more frequent at the 2-cell stage (43.2%) when compared to the 4-cell stage (15%), suggesting that there are mechanisms that can correct this abnormality. In addition, they reported similar rates of multinucleation in euploid versus aneuploid embryos (40.8% vs. 46.7%). Finally, of those embryos showing multinucleation at the 2-cell stage, 61 were transferred and a 45.9% of clinical pregnancy rate obtained with them.

VerMilyea et al. (2014) used the EEVA prediction model developed by Wong et al. (2010) and Conaghan et al. (2013) to study its predictive ability for implantation and clinical pregnancy. TL recordings were obtained from six different clinics that used clinic specific stimulation and culture protocols. Embryos were selected based on morphology for transfer and the TL recordings of those embryos where implantation was known (n = 331) were analyzed. A two- (high and low implantation potential) and a three-category (high, medium, and low implantation potential) evaluation was tested based on CC₂ and S ₂ (CC₂: 9.33≤ and ≤11.45 h and S ₂:≤1.73 h; 2-category results: EEVA high when both in range and EEVA low when one or both out of range vs. 3-category results: EEVA high: CC₂ and S ₂ in range; EEVA medium: CC₂ 9.33≤ and ≤12.65 h and s ₂: ≤4 h; EEVA low: out of the above ranges). Both the two- and three-category models were predictive of implantation and clinical pregnancy.

The latest model was published by Petersen et al. (2016). This algorithm is based on five kinetic and one morphological event. The algorithm was built based on retrospectively collected data from 24 clinics using local embryology practices. The algorithm assigns five scores to the embryos and a sevenfold increase in implantation rate can be seen starting from the lowest score to the highest. This score, the KIDScore, is described in detail in the next section.

External validation

There are many potential confounders when morphokinetic parameters and clinical outcome are correlated. Patient characteristics, stimulation methods, culture conditions, O₂ concentration, and so on can all influence embryo development and as a result the kinetic markers. Therefore, when one group recommends an algorithm as a model to predict implantation or pregnancy, a different group in a different patient population using different culture conditions may not find the results helpful. In order to accept and introduce an algorithm in routine practice, it has to be tested in a different clinical setting.

Freour et al. (2015) studied the Meseguer hierarchical model (Meseguer et al., 2011) in a retrospective analysis. About 2240 embryos obtained from 450 couples were considered for the analysis. There were no exclusion criteria, any couple undergoing intra-cytoplasmic sperm injection (ICSI) treatment was considered eligible. Embryos were cultured in ES under a reduced (5%) O₂ environment and transfers were performed at cleavage as well as blastocyst stages. Embryos were selected primarily based on morphology and then based on kinetic analysis. 528 embryos with KID were analyzed. The authors observed a heterogeneous distribution of implantation rates across the kinetic categories. The correlation coefficients were lower than in the original study by Meseguer et al. (2011), were modest only, and were not significant. Based on their results, the authors did not recommend routine application of the original model but proposed each clinic to develop their own predictive algorithm.

Another study, conducted by Kirkegaard et al. (2014), tested the Conaghan model. The Conaghan model (Conaghan et al., 2013) predicts blastocyst formation using two early kinetic markers, CC₂ and S ₂. Those cleavage stage embryos that follow in range kinetics are more likely to turn into usable blastocysts. Kirkegaard et al. tested whether the same model is able to predict implantation. TL kinetic data was obtained from seven clinics. Embryos were cultured in ES up until the transfer on day 2 or 3. Selection for transfer was based on morphology. The model was applied to those 1519 embryos that had KID. In the overall cohort, the implantation rate was 17.4%. Among embryos that were considered usable by the model the implantation rate was 22.7%, while among those embryos that were considered nonusable the implantation rate was 14.2%. More importantly, half of the embryos that eventually resulted in a pregnancy were identified as nonusable by the model and potentially would have been discarded.

A different approach for “external validation” is to pull data from different clinics using different daily practices.

This approach was used by Petersen et al. (2016) and they proposed a different algorithm to predict implantation using day 3 embryos. The model was built on retrospective analysis of TL imaging of 3275 embryos with known implantation data. Data was obtained from 24 clinics using heterogeneous patient populations, culture conditions, and embryo handling. Five kinetic events and one morphological event are included in the algorithm. The first parameter is t ₃-tPNf (pronuclear fading). Optimally, this would be over 11.48 h (if shorter the embryo is given score 1). When it is >11.48 h, we proceed to the second checkpoint, t ₃ which ideally would be less than 42.91 h (if longer the embryo is given score 2). When t ₃ is <42.91 h, a third parameter is calculated by the following formula: (t ₅ − t ₃)/(t ₅ − t ₂). If the equation is <0.3408, the embryo is given score 3, if it is ≥0.5781, it is given a score of 4. If it is between 0.3408 and 0.5781 and the embryo fails to reach the 8-cell stage by 66 h and it is also given a score of 4. For those embryos that reach the 8 or more cell stage by 66 h a score of 5 is assigned. Across the five score groups, the implantation rate increased sevenfold from 5.18% (score 1) to 36.17% (score 5). The strengths of this scoring system (KIDScore) over the other algorithms have to be emphasized. The algorithm was built based on TL recordings over 3000 embryos with known implantation outcome. The data was obtained from 24 clinics treating patients with a wide range of problems and using different culture conditions. The model predicts a clinically relevant outcome rather than surrogate markers of success. The algorithm was tested in subsets of cases based on method of fertilization and O₂ concentration and very similar values were obtained when compared to the entire data set. This suggests that the algorithm works well both in low and ambient O₂ concentration and both with IVF and ICSI fertilization.

Randomized controlled trials using algorithm-based embryo selection

Ultimately, the true predictive value of the various markers or their combinations (algorithms) would be tested prospectively, ideally in a randomized controlled trial (RCT) when the algorithm is compared to the current standard, daily-once morphological assessment. Unfortunately, there is still limited randomized control trial data that assesses the discriminating ability of the various TL markers/algorithms prospectively. In order to test the benefit of the TL parameters alone, test and control embryos should be cultured under similar conditions. Otherwise, if a difference is noted between the randomized groups, we will not know what to contribute the benefit to if the culture conditions also differ. It is also true that if the culture conditions are identical, then the selection of the embryos in the TL group has to rely solely on predefined TL parameters/algorithm to prove the superiority of TL algorithm-based selection over standard morphological evaluation. We have to keep this in mind when RCT data is analyzed.

Kahraman et al. (2013) tested the Meseguer hierarchical model (based on t ₅, S ₂, CC₂) in a small RCT among good prognosis patients. In the TL group, embryos were cultured in ES, while in the control group, embryos were cultured in conventional incubators. The authors did not find a difference in clinical outcome between the groups. In a much larger study, Rubio et al. (2014) randomly assigned patients to conventional culture and embryo selection for transfer based on day 3 and/or day 5 morphology versus culture in ES and selection for transfer based on t ₅, S ₂, and CC₂ (Meseguer hierarchical model). Patients aged 20–38 years were eligible to participate. The use of autologous and fresh or frozen donated oocytes was allowed. The day of transfer was not standardized as both day 3 and day 5 transfers were allowed. Unfortunately, some of the patients decided to follow the other protocol (not the assigned) after randomization. The authors found lower miscarriage rate and significantly improved ongoing pregnancy rate in the TL arm. Park et al. (2015) randomly assigned patients to embryo culture in a TL system (ES) versus in a conventional incubator. They selected embryos for transfer based on day 2 morphology. Their aim was to see whether the proportion of good quality embryos was higher if embryos were cultured in the TL unit. They found no difference in the proportion of the top quality day 2 embryos with the two incubation methods. This should not necessarily be surprising as the culture conditions differed and the duration of culture was very short. If we accept that by limiting the out-of-incubator handling of the embryos we can improve their growth potential, then in this study, this benefit was not utilized as the embryos had to be removed for fertilization check and day 2 assessment in both groups. Goodman et al. (2016) randomly assigned patients to morphologic assessment versus assessment based on morphology plus TL parameters. Embryos in both groups were cultured in TL units. In the TL group, embryos were primarily selected for the transfer based on morphology and TL parameters were only considered when similar quality embryos were available for transfer. Clinical outcome was similar in the two groups. In the study by Matyas et al. (2015), young and good prognosis patients were randomly assigned to single blastocyst transfer (1) selected based on day 5 morphology versus (2) based on a predefined TL score (algorithm) comprising both kinetic and morphologic parameters. A third group of nonrandomized patients undergoing double blastocyst transfer (DET) was included as well. Clinical and perinatal outcomes were compared. Embryos in all groups were cultured under similar conditions and the amount of out-of-incubator handling was identical too. The pregnancy rate was higher in the DET group when compared to the SET group evaluated using standard morphology; however, it was similar in the TL-selected SET group. The perinatal outcome was significantly better in the SET groups when compared to DET. Based on the results, it was concluded that one can successfully compensate for the fewer embryos transferred if the selection is based on a complex morphokinetic TL score and with this approach the perinatal outcome can be improved (Matyas et al., 2015).

Conclusions

TL technology is relatively new but it is being introduced in everyday embryology laboratory use for multiple reasons (undisturbed culture, detailed morphokinetic data, quality control, ease of workload, and improved documentation). As data has been accumulated, numerous papers were published primarily using retrospective data analysis to test the predictive abilities of different morphokinetic events. As a logical next step, algorithms built based on multiple morphokinetic parameters were proposed. The initial optimism regarding the benefits of these hierarchical models was overshadowed by the lack of ability to externally validate them. It was realized that there are multiple factors that could affect a model’s ability to predict a clinical outcome. Most recently, a new algorithm was introduced that was built based on data collected from 24 clinics using different laboratory technologies in a heterogeneous patient population. The model seems to predict implantation well as from the lowest to the highest score a sevenfold increase in implantation rate can be seen. For the first time, it seems that we have a model that is ready for introduction into routine daily care. This is particularly helpful for smaller clinics performing fewer cycles where it could take a very long time to collect enough data to build their own model especially if different algorithms have to be built for different clinical settings. Eventually, the model by Petersen et al. will also need to be tested prospectively.

It seems that there are competing technologies for the selection of the embryo with the highest chance to implant. It may be time for a paradigm shift, and rather than testing the methods against each other, we should combine them to explore their full benefit. The noninvasive TL technology could be used to “prescreen” embryos and the more expensive, invasive PGS could be applied to a smaller set of embryos to reduce costs and intervention. They together could improve IVF success in a patient-friendly way.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Azzarello

Hoest

Mikkelsen

(2012) The impact of pronuclei morphology and dynamicity on live birth outcome after time-lapse culture. Human Reproduction 27(9): 2649–2657.

Balakier

Sojecki

Motamedi

. (2016) Impact of multinucleated blastomeres on embryo developmental competence, morphokinetics, and aneuploidy. Fertility and Sterility 106: 608–614.

Basile

del Carmen Nogales

(2014) Increasing the probability of selecting chromosomally normal embryos by time-lapse morphokinetics analysis. Fertility and Sterility 101: 699–704.

Basile

Vime

Florensa

. (2015) The use of morphokinetics as a predictor of implantation: a multicentric study to define and validate an algorithm for embryo selection. Human Reproduction 30(2): 276–283

Campbell

Fishel

Bowman

. (2013a) Modelling a risk classification of aneuploidy in human embryos using non-invasive morphokinetics. Reproductive BioMedicine Online 26: 477–485.

Campbell

Fishel

Bowman

. (2013b) Retrospective analysis of outcomes after IVF using an aneuploidy risk model derived from time-lapse imaging without PGS. Reproductive BioMedicine Online 27: 140–146.

Chamayou

Patrizio

Storaci

. (2013) The use of morphokinetic parameters to select all embryos with full capacity to implant. Journal of Assisted Reproduction and Genetics 30(5): 703–710

Chavez

Loewke

Han

. (2012) Dynamic blastomere behavior reflect human embryo ploidy by the four-cell stage. Nature Communications 3: 1251. DOI: 10.1038/ncomms2249.

Conaghan

Chen

Willman

. (2013) Improving embryo selection using a computer-automated time-lapse image analysis test plus day 3 morphology: results from a prospective multicenter trial. Fertility and Sterility 100(2): 412–419.e5.

10.

Cruz

Garrido

Herrero

. (2012) Timing of cell division in human cleavage-stage embryos is linked with blastocyst formation and quality. Reproductive BioMedicine Online 25: 371–381.

11.

Dahdouh

Balayla

García-Velasco

(2015) Impact of blastocyst biopsy and comprehensive chromosome screening technology on preimplantation genetic screening: a systematic review of randomized controlled trials. Reproductive BioMedicine Online 30(3): 281–289.

12.

Dal Canto

Coticchio

Mignini Renzini

. (2012) Cleavage kinetics analysis of human embryos predicts development to blastocyst and implantation. Reproductive BioMedicine Online 25(5): 474–480.

13.

European IVF-Monitoring Consortium (EIM); European Society of Human Reproduction and Embryology (ESHRE), Kupka

D’Hooghe

Ferraretti

. (2016) Assisted reproductive technology in Europe, 2011: results generated from European registers by ESHRE. Human Reproduction 31(2): 233–248.

14.

Freour

Le Fleuter

Lammers

. (2015) External validation of a time-lapse prediction model. Fertility and Sterility 103: 917–922.

15.

Glujovsky

Farquhar

Quinteiro Retamar

. (2016) Cleavage stage versus blastocyst stage embryo transfer in assisted reproductive technology. Cochrane Database of Systematic Reviews 30(6): CD002118.

16.

Goodman

Goldberg

Falcone

. (2016) Does the addition of time-lapse morphokinetics in the selection of embryos for transfer improve pregnancy rates? A randomized controlled trial. Fertility and Sterility 105(2): 275–285.

17.

Kahraman

Cetinkaya

Pirkevi

. (2013) Comparison of blastocyst development and cycle outcome in patients with eSET using either conventional or time lapse incubators. A prospective study of good prognosis patients. Journal of Reproductive and Stem Cell Biotechnology 3(2): 55–61.

18.

Kirkegaard

Campbell

Agerholm

. (2014) Limitations of a time-lapse blastocyst prediction model: a large multicentre outcome analysis. Reproductive BioMedicine Online 29: 156–158.

19.

Kovacs

(2014) Embryo selection: the role of time-lapse monitoring. Reproductive Biology and Endocrinology 12: 124.

20.

Matyas

Kovacs

Forgacs

. (2015) Selection of single blastocyst for transfer using time-lapse monitoring during in vitro fertilization in good prognosis patients: a randomized controlled trial. Human Reproduction 30(Suppl 1): i119.

21.

Meseguer

Herrero

Tejera

. (2011) The use of morphokinetics as a predictor of embryo implantation. Human Reproduction 26(10): 2658–2671.

22.

Meseguer

Rubio

Cruz

. (2012) Embryo incubation and selection in a time-lapse monitoring system improves pregnancy outcome compared with a standard incubator: a retrospective cohort study. Fertility and Sterility 98(6): 1481–1489.

23.

Milewski

Kuc

Kuczynska

. (2015) A predictive model for blastocyst formation based on morphokinetic parameters in time-lapse monitoring of embryo development. Journal of Assisted Reproduction and Genetics 32: 571–579.

24.

Montag

Toth

Strowitzki

(2013) New approaches to embryo selection. Reproductive BioMedicine Online 27(5): 539–546.

25.

Motato

de los Santos

Escribe

. (2016) Morphokinetic analysis and embryonic prediction for blastocyst formation through an integrated time-lapse system. Fertility and Sterility 105: 376–384.

26.

Park

Bergh

Selleskog

. (2015) No benefit of culturing embryos in a closed system compared with a conventional incubator in terms of number of good quality embryos: results from an RCT. Human Reproduction 30(2): 268–275.

27.

Petersen

Boel

Montag

. (2016) Development of a generally applicable morphokinetic algorithm capable of predicting the implantation potential of embryos transferred on Day 3. Human Reproduction 2016; 31(10): 2231–2244. DOI: 10.1093/humrep/dew188.

28.

Rubio

Galán

Larreategui

. (2014) Clinical validation of embryo culture and selection by morphokinetic analysis: a randomized, controlled trial of the EmbryoScope. Fertility and Sterility 102(5): 1287–1294.

29.

Rubio

Kuhlmann

Agerholm

. (2012) Limited implantation success of direct-cleaved human zygotes: a time-lapse study. Fertility and Sterility 98: 1458–1463.

30.

Siristatidis

Komitopoulou

Makris

. (2015) Moprhokinetic parameters of early embryo development via time lapse monitoring and their effect on embryo selection and ICSI outcomes: a prospective cohort study. Journal of Assisted Reproduction and Genetics 32: 563–570.

31.

Vajta

Korosi

. (2008) The Well-of the-Well system: an efficient approach to improve embryo development. Reproductive BioMedicine Online 17(1): 73–81.

32.

VerMilyea

Tan

Anthony

. (2014) Computer-automated time-lapse analysis results correlate with embryo implantation and clinical pregnancy: a blinded, multi-centre study. Reproductive BioMedicine Online 29: 729–736.

33.

Wong

Loewke

Bossert

. (2010) Non-invasive imaging of human embryos before embryonic genome activation predicts development to the blastocyst stage. Nature Biotechnology 28(10): 1115–1121.

Time-lapse embryoscopy

Abstract

Keywords

Introduction

TL technology

TL parameters

Heterogeneity in the markers identified as predictors of various outcomes

Algorithms predicting implantation

External validation

Randomized controlled trials using algorithm-based embryo selection

Conclusions

Footnotes

Declaration of Conflicting Interests

Funding

References