Prediction of embryo developmental potential and pregnancy based on early stage morphological characteristics

Transcription

1 Prediction of embryo developmental potential and pregnancy based on early stage morphological characteristics Peter Sjöblom, Ph.D., a,c Judith Menezes, Ph.D., b Lisa Cummins, B.Sc., b Bagyalakshmi Mathiyalagan, Ph.D., b and Michael F. Costello, M.B.B.S., M.Med., FRANZCOG, CREI a,b a School of Women s and Children s Health, University of New South Wales, b Department of Reproductive Medicine and IVF Australia, Royal Hospital for Women, Randwick, New South Wales, Australia; and c Fertility Centre Scandinavia Stockholm, Stockholm, Sweden Objective: To analyze the association between morphological details at different stages of culture with blastocyst development, with an aim to improve selection for transfer. Design: Retrospective audit of data. Setting: Tertiary referral center and university hospital. Patient(s): Two hundred sixty-eight couples underwent 357 treatment cycles. Intervention(s): Oocyte pickups for IVF or intracytoplasmic sperm injection (ICSI) after ovarian stimulation. Embryos were individually cultured and examined on days 0 2 for morphological details and developmental characteristics, and selected for transfer, freezing, or further culture. Main Outcome Measures: The association of blastocyst development and pregnancy with morphological characteristics. Result(s): Five morphological characteristics (appearance of the cytoplasm, pronuclei and nucleoli, cytoplasmic deficit, and developmental rate) showed the strongest association with blastocyst development. By combining information from all days of culture into a cumulative score, prediction was greatly improved, compared to only using day 2 morphology. Cytoplasmic dysmorphisms of the oocyte, including accumulation of smooth endoplasmic reticulum, were associated with poor developmental performance. Differential weighting of these characteristics was included in a new embryo scoring system, which showed a strong correlation with implantation. Conclusion(s): Weighting individual morphological characteristics of zygotes and embryos and combining them into a cumulative embryo score can improve selection of embryos for transfer. (Fertil Steril 2006;86: by American Society for Reproductive Medicine.) Key Words: Embryo score, blastocyst, oocyte, zygote, morphology A reduction of the number of embryos transferred after IVF/intracytoplasmic sperm injection (ICSI) is highly desirable in view of the burden associated with multiple gestation (1). The selection of embryos for transfer into the uterus is a critical step in the IVF/ICSI treatment to reduce the risk of multiple gestation and to maximize the probability of pregnancy, especially when only one embryo is transferred. Although some clinics replace embryos into the uterus on day 2 of development, others prefer replacement on day 3 or, even later, at the blastocyst stage. One advantage of choosing a later stage of development for replacement is that it allows for selection of embryos that have shown competency to reach developmental milestones (e.g., the maternal embryonic switch of developmental control) (2). A theoretical disadvantage of later stage transfers is that, because of deficiencies in the present culture systems, Received May 26, 2005; revised and accepted March 5, Reprint requests: Peter Sjöblom, Ph.D., Fertility Centre Scandinavia Stockholm, Storangsvagen 10, SE Stockholm, Sweden (FAX: ; peter.sjoblom@fcivf.com). some embryos that could have survived in vivo may be lost through attrition in vitro. Also, blastocyst culture places more demand on the resources of the laboratory, as the embryos occupy space in the incubators for longer periods of time and require additional handling. Randomized controlled trials involving unselected patients, comparing transfer at the cleavage stage versus at the blastocyst stage, do not suggest that either alternative is superior in terms of births per oocyte pickup (3). Against this background, there seems to be a case for improving the selection of cleavage stage embryos. Some patients may still request blastocyst transfer and a cleavage stage scoring system is also potentially useful to help counsel such patients on day 2 as to what the chances are for their embryos to develop to blastocyst stage in vitro, and hence to choose between day 2 3 (cleavage stage) embryo transfer versus day 5 6 (blastocyst stage) embryo transfer. Hundreds of articles have been published on embryo morphology, and there are numerous articles indicating the usefulness of multistage scoring systems (see Ebner et al. [4] for review). However, crude embryo scoring systems are still 848 Fertility and Sterility Vol. 86, No. 4, October /06/$32.00 Copyright 2006 American Society for Reproductive Medicine, Published by Elsevier Inc. doi: /j.fertnstert

2 common and further information needs to be disseminated to improve practices. The aim of the present retrospective study was to analyze how a range of morphological characteristics, used in an initial scoring system, are interrelated and related to the developmental potential of embryos, as well as to determine their usefulness for predicting embryonic development when combined into a summary score. Based on these findings a new, differentially weighted scoring system was developed and its relationship to implantation rates was analyzed. MATERIALS AND METHODS Patients and Treatments During the period January to December 2001, 268 couples underwent 357 oocyte pickups for IVF (n 170) or ICSI (n 187) treatment. Embryo transfer was usually done 2 days after oocyte pickup (D2). In the majority of the treatment cycles, a hormonal treatment regimen designated OCP- Long protocol was used. In this protocol, the patients started oral contraceptive (OC) pills (Brevinor 21; Pharmacia Australia, Rydalmere, NSW, Australia) 3 5 days after the onset of menses and continued for 21 days. After taking the OCs for 21 days, administration of GnRH agonist (GnRH-a) was started, either using nafarelin (Synarel, Pharmacia Australia) 200 g intranasal spray twice a day, or leuprolide acetate (LA) injections (Lucrin; Abbot Australasia, Cronulla, NSW, Australia) 1 mg SC daily. After 14 days of GnRH-a treatment, ovarian down-regulation was confirmed by measurement of serum E 2 level. If the level of E 2 was 120 pmol/l, GnRH-a treatment continued for another week and the E 2 level was assessed again. Estrogen-producing cysts were aspirated. When E 2 was 120 pmol/l, stimulation with recombinant FSH (Gonal-F; Serono Laboratories, Frenchs Forest, NSW, Australia; or Puregon; Organon Laboratories, Lane Cove, NSW, Australia) was initiated, commonly with a starting dose of IU injected SC daily. The FSH dose was adjusted according to the ovarian response, and serial E 2 and transvaginal ultrasound scans were performed every 2 3 days from day 6 of FSH stimulation onward. When one or more follicles had reached a size 18 mm, final oocyte maturation was induced with an SC injection of 10,000 IU of hcg (Profasi; Serono Laboratories or Pregnyl; Organon Laboratories). Oocyte pickup under general anesthesia followed 36 hours later. Embryo transfer was commonly performed on D2. Fourteen days after embryo transfer, serum -hcg was determined as a pregnancy test (considered positive if 20 IU/L) and a transvaginal ultrasound scan of the uterus was done after 6 7 weeks of amenorrhea to determine whether a clinical pregnancy had been established (intrauterine gestational sac visible). Because this study was a retrospective audit of already existing clinical practices and collected data, it was not deemed necessary to obtain approval from the University of New South Wales ethical committee (the equivalent of an Institutional Review Board). Culture Conditions and Scoring The gametes and embryos were handled and cultured in the Quinn s Advantage (QA) range of media and supplements and additives, all from Sage Biopharma (Gytech, Melbourne, Australia). All media were supplemented with 10 mg/ml of human serum albumin. Egg collection, sperm preparation, and ICSI were done in QA medium with HEPES, whereas fertilization in vitro was done in QA fertilization medium in four-well dishes, with up to six eggs per dish. The ICSI oocytes were cultured individually in 10- L droplets under oil. From D1 onward, all zygotes/embryos were individually cultured in droplets. On D1, at about 18 hours after insemination or ICSI, zygotes were moved to 10- L droplets of QA cleavage medium under oil and scored. Scoring on D2 took place at about 42 hours after insemination. On D3, the embryos were transferred to 10- L droplets of QA blastocyst culture medium under oil. Embryos were cultured at 37 C either in large incubators in 5% CO 2 in air, or in MINC incubators (Cook IVF, Eight Miles Plain, QLD, Australia) in 6% CO 2,5%O 2, and 89% N 2. All tissue culture plastics (all Becton Dickinson, North Ryde, NSW, Australia) were prewarmed to 37 C, aspirates were kept in tube warmers at all times, work bench surfaces were heated to 38.5 C to maintain the temperature in the culture dishes at 37 C, the room temperature of the laboratory was C, and laboratory lighting was dimmed. Oocytes for ICSI were denuded using hyaluronidase (Sage Biopharma), 10 IU/mL, in QA medium with HEPES. Using the initial scoring system, they were scored for characteristics shown in Table 1A, some of which are illustrated in Figure 1. Oocyte cumulus complexes for IVF were not scored, but anything remarkable about their appearance, such as absence of matrix expansion, was noted on the embryology work sheet. Because oocytes obtained for IVF treatment were not denuded, they could not be scored in the same way as oocytes obtained for ICSI, and hence were not assigned a D0 score. On D1, at hours after insemination, zygotes from IVF were denuded of their granulosa cells (GC) using Flexipets (Cook IVF) to facilitate inspection, and moved to cleavage medium. Later in the day, at 25 hours after insemination, the embryos were inspected for occurrence of syngamy. The zygotes were scored for the characteristics described in Table 1B, some of which are illustrated in Figure 1. The next day, D2, at 42 hours after insemination embryos were, in most cases, selected for transfer and freezing. Scoring criteria for D2/3 are shown in Table 1C, some of which are illustrated in Figure 1. The morphological characteristics for the D2 score were derived from that described by Mohr et al. (5). The corrected D2 score is the D2 score corrected for multinucleated blastomeres (MNBs), therefore the score Fertility and Sterility 849

3 TABLE 1 Morphological characteristics used in the initial embryo scoring system. Characteristic Description Score A D0 (oocyte) characteristics Polar body Round/oval, unfragmented 1 Cytoplasm Smooth, finely granulated 1 Coarse, vacuoles, dark patch 0 Membrane Smooth 1 Jagged 0 Zona pellucida Normal thickness and color, no debris 1 Maximum score 4 Aggregation of SER Pancake-like diffuse structure Deselected from transfer or freezing B D1 (zygote) characteristics Polar bodies 90 apart, 45 with axis of pronuclei 1 Cytoplasmic halo Present, normal cytoplasm 1 Absent 0 Membrane Smooth 1 Jagged 0 Nucleoli Equal numbers and sizes, 3 1 Pronuclei Equal size, central, apposed 1 Other Zona pellucida Normal thickness and color, no debris 1 Maximum score 6 Syngamy Breakdown of pronuclear membranes; cleavage at 25 hpi Noted and preferred when selecting embryos for ET C D2/D3 (embryo) characteristics Cell numbers Number of blastomeres Noted Zona pellucida Normal thickness and color, no debris 1 Cytoplasm Smooth, honey-colored 1 Membrane Smooth 1 Jagged 0 Cell size Equal if 2 n blastomeres, otherwise in 1 accord with cleavage stage Cell shape Spherical 1 Perivitelline space Blastomeres fill the space under the 1 zona Fragmentation 10% fragmented 2 10% 30% fragmented 1 30% fragmented 0 Developmental rate 4 cells 42 hpi; 8 cells 66 hpi 2 2,3or 4 cells 42 hpi; 6 7 or 8 cells 66 hpi Sjöblom et al. Embryo score and blastocyst development Vol. 86, No. 4, October 2006

4 TABLE 1 Continued. Characteristic Description Score Maximum score 10 MNB Multiple (equal sized) or fragmented (varying sizes) nuclei Score is decreased by the percentage of blastomeres with MNBs Note: D0 day of oocyte pick-up; D1 day after oocyte pick-up; D2/3 2/3 days after oocyte pick-up; hpi hours post insemination or ICSI; SER smooth endoplasmic reticulum; MNB multinucleated blastomeres. is reduced by the proportion of blastomeres having multiple or fragmented nuclei. For example, if the D2 score of a four-cell embryo is 12 and one blastomere is multinucleated, the corrected D2 score for that embryo is 9 (i.e., 12 1/4 of 12). If two blastomeres were multinucleated, the corrected D2 score of that embryo would be 6. The rationale for this correction is that blastomeres with fragmented or multiple nuclei would have a severely disturbed chromosomal complement and thus cause a reduction of the developmental capacity of the embryo (6 9). An indication of the validity of this reasoning can be seen in a study by Edgar et al. (10) that showed that the implantation rate of frozen and thawed embryos was reduced proportionally to the fraction of blastomeres not surviving thawing. Embryos deemed unsuitable for transfer or freezing on that day were cultured further, and frozen, if appropriate, on D3 or D5 or D6. However, embryos that were considered terminally arrested or completely fragmented were discarded. Embryos that developed to blastocysts on D5 or D6 were scored according to the criteria used by Gardner et al. (11) and described in Table 2. Good quality blastocysts (score 3AA) were frozen. For embryo selection, the sum of the scores from all individual days of culture was used. The D0, D1, and D2 scores are the sums of the scores for individual characteristics observed on the respective days. A cumulative score was used, which was made up from the sum of the scores of the separate culture days. The corrected cumulative D2 score is the sum of the D0, D1, and the corrected D2 scores. However, because the IVF oocytes are not scored on D0, the corrected cumulative D2 score for embryos derived from IVF is the sum of the D1 and corrected D2 scores, whereas the corrected cumulative D2 score for ICSI-derived embryos is the sum of the D0, D1, and corrected D2 scores. To facilitate comparisons between embryos derived from ICSI and IVF, the percentage score was constructed. The percentage score is the corrected cumulative D2 score divided by the maximum score for that embryo (16 for IVFderived embryos, 20 for ICSI-derived embryos). Thus, an IVF embryo scoring 16 and an ICSI embryo scoring 20 both have a percentage score of 100. To evaluate the interobserver variability, videotaped images of 16 D2 embryos were analyzed by three embryologists separately and the coefficient of variation (CV) for the D2 score was calculated. A further estimate of concordance between observers was obtained from the proportion of observations of individual characteristics where there was no unanimity between the embryologists with regard to scoring. After analyzing the association between individual characteristics and blastocyst development, a new scoring system was designed, placing more emphasis on characteristics that were significantly associated with developmental potential. Differential weighting of the individual characteristics was assigned as shown in Table 3 (for further discussion of the scoring system, see Menezes et al. [12]). This scoring system was applied clinically, and the association between the D2 percentage score of transferred embryos and implantation was analyzed. During the period January to December 2003, a total of 249 embryos were transferred in 156 cycles, in which either all embryos implanted or no embryo implanted, resulting in 49 implantation events. Statistics The association between blastocyst (blastocoel) development and individual morphological characteristics, scores on particular days, or summary scores, was analyzed using the Spearman rank order correlation. The significance of differences between proportions was estimated with the 2 test or Fisher s exact test, as appropriate. The concordance between observers was assessed by analysis of variance and intraclass coefficient of correlation. RESULTS Patient and cycle characteristics are shown in Table 4, as are the number of oocytes collected, fertilized, cleaved, and studied. The frequency of occurrence of normal morphological characteristics of oocytes, zygotes, and embryos are Fertility and Sterility 851

5 FIGURE 1 Examples of normal and abnormal morphological characteristics of oocytes, zygotes, and embryos. (A) Oocyte with a large perivitelline space (solid arrow) and a dark patch (dashed arrow); (B) oocyte with a vacuole (dashed arrow) and a necrotic patch (solid arrow); (C) oocyte accumulation of smooth endoplasmic reticulum; (D) normal zygote with halo (arrow), central pronuclei and polarized and coalesced nucleoli; (E) embryo with much fragments (out of focus, fragmentation score 1) uneven-sized and irregularshaped (oval) blastomeres; (F) highly fragmented embryo (fragmentation score 0) and abnormal cytoplasm in one of the blastomeres (clear zone at arrow); (G) four-cell embryo with slightly irregularshaped (oval) blastomeres, one of which jagged membrane; (H) four-cell embryo with slightly irregularshaped and uneven sized blastomeres, and a gap between the blastomeres and the zona pellucida, thus classified as Does not fill the space under the zona. listed in Table 5. In this table, data for all oocytes and embryos cultured (not just those cultured to D5 or D6) were included, to improve the validity of the analysis. Some characteristics, notably syngamy status and pronuclear and nucleolar development that occur along the developmental time line are different between embryos derived from IVF and ICSI. The relative CV for the D2 score among the embryologists was 18%. For obvious reasons, the relative CV is greater for low-scoring embryos. It appears that the scoring is more consistent for some characteristics than for others, in that the proportion of unanimity in the scoring of a particular feature varied between 94% (whether there are MNBs) and 55% (whether the membrane is smooth or jagged). The scoring of D2 morphology showed an intraclass correlation coefficient of 0.71, and there was no significant difference between observers in their scoring (F 2.38, P.13, not significant). There is a significant correlation between many variables. Thus, of the D0 characteristics, the appearance of the cytoplasm is significantly and negatively associated with the appearance of the membrane (r 0.09; P.001) and the polar body (PB) (r 0.13; P.001). Of the D1 characteristics, significant association of pronuclear morphology with PB appearance (r 0.13; P.001), cytoplasm (R 0.9; P.001), as well as nucleolar organization (r 0.13; P.001) was seen. Fragmentation on D2 was associated TABLE 2 D5/6 (blastocyst) Characteristics. Characteristic Blastocoel score 1 Early blastocyst, blastocoel 50% of the volume 2 Early blastocyst, blastocoel 50% 80% of the volume 3 Nonexpanded blastocyst, large blastocoel 4 Expanded blastocyst 5 Beginning to hatch 6 Fully hatched blastocyst Inner cell mass score A Tightly packed, many cells B Loosely grouped, few cells C Very few cells Trophectoderm score A Many cells B Few cells C Very few cells Note: Gardner et al., 2000 (11). 852 Sjöblom et al. Embryo score and blastocyst development Vol. 86, No. 4, October 2006

6 TABLE 3 Description of the weighted scoring system. Characteristic Description Score A D1 (zygote) characteristics (day after oocyte pick-up) Polar body orientation 45 with axis of pronuclei 3 Cytoplasmic halo Present 3 Absent 0 Cytoplasmic texture Normal 6 Slightly granular 3 Vacuoles, dark patch, very granular 0 Membrane Smooth 3 Jagged 0 Nucleoli Equal number ( 8) and polarized 18 Equal number and scattered 6 Pronuclei, size Equal 3 Unequal 0 Pronuclei, position Central 3 Eccentric 0 Pronuclei, apposition Apposed 3 Apart 0 Syngamy at 25 hpi Cleavage to the 2-cell stage 8 Breakdown of pronuclear membranes 4 Intact PNs 0 Maximum score 50 Aggregation of smooth No freezing or transfer, total score 0 endoplasmic reticulum on D0 B D2 (embryo) characteristics at 42 hpi (2 days after oocyte pick-up) Zona pellucida thickness Variable 3 Uniform 0 Cytoplasm Clear 3 Granular, vacuoles 0 Membrane Smooth 3 Jagged 0 Blastomere size Equal if 2 n blastomeres, otherwise in 3 accord with cleavage stage Cell shape Spherical, regular 3 Perivitelline space Blastomeres fill the space under the 5 zona Large space between cells and zona 0 Fragmentation 10% fragmented 10 10% 30% fragmented 5 30% fragmented 0 Developmental rate 4 cells 20 2,3or 4 cells 10 Maximum score 50 Fertility and Sterility 853

7 TABLE 3 Continued. Characteristic Description Score Multinucleated blastomeres (MNBs) Note: hpi hours post insemination. Multiple (equal sized) or fragmented (varying sizes) nuclei The score is decreased by the percentage of blastomeres with MNBs with cell size ((r 0.29; P 001), shape (r 0.25; P.001) and whether blastomeres fill the space under the zona (r 0.21; P.001). Furthermore, the morphology of the zona pellucida (ZP) is judged similarly on D0, D1, and D2, and this is true also for cytoplasmic appearance (data not shown). Finally, there was a strong association between the occurrence of syngamy at 25 hours after insemination and the cell numbers at 42 hours after insemination (r 0.49; P.001). The occurrence of aggregation of smooth endoplasmic reticulum in oocytes seems to indicate a very poor developmental performance, in that no oocyte with smooth endoplasmic reticulum aggregation developed into any sort of TABLE 4 Patient and cycle characteristics, and causes of infertility. No. of patients 268 Median age (range) (y) 37 (22 46) Proportion of cycles in patients 30% aged 40 y Median (range) number of 2 (1 9) completed IVF/ICSI cycles No. of oocyte pickups 357 No. of embryo transfers (ETs) 340 No. of embryos to ET 724 (2.1/ET) No. of positive hcg 89 (26.2%/ET) No. of clinical pregnancies 81 (23.8%/ET) Implantation rate 14.2% Tubal infertility 124 (35%) Male factor infertility 42 (12%) Other female infertility 59 (16%) Combined female and male 46 (13%) infertility Unexplained infertility 86 (24%) No. of oocytes collected 3,237 No. of oocytes fertilized (2PN) 1,961 No. of cleaved zygotes 1,876 No. of oocytes cultured to 431 day 6 blastocyst. Table 6 shows the association between morphological characteristics on the various days of observation and blastocyst development. A normal cytoplasm of the oocyte is significantly associated with blastocyst development (P.035). It appears that pronuclear (P.031) and nucleolar (P.003) morphology are strong determinants of blastocyst development. In the present material, zygotes that TABLE 5 Frequency of occurrence of morphological normal characteristics on D0 D2, and differences between ICSI and IVF. Treatment: Characteristic ICSI IVF n % n % P value D0 polar body 1, D0 cytoplasm 1, D0 membrane 1, D0 zona pellucida 1, D1 polar bodies 1, D1 cytoplasm 1, D1 membrane 1, D1 nucleoli 1, D1 pronuclei 1, D1 zona pellucida 1, Syngamy D2 zona pellucida D2 cell size D2 cell shape D2 perivitelline space D2 membrane D2 cytoplasm D2 fragments 1, D2 MNBs Note: n number of observations; NA not applicable; NS not significant; D0 day of oocyte pickup; D1 day after oocyte pick-up; D2 2 days after oocyte pick-up. Data for all oocytes/embryos cultured are included. 854 Sjöblom et al. Embryo score and blastocyst development Vol. 86, No. 4, October 2006

8 TABLE 6 Association between morphological characteristics and blastocoel development in embryos cultured to D5 or D6, estimated by the Spearman rank order coefficient of correlation, r. FIGURE 2 The impact of the number of blastomeres on D2 on the percentage of embryos developing into blastocysts. Blastocoel score vs. n r P value D0 polar body NS D0 cytoplasm D0 membrane NS D0 zona pellucida NS D1 polar bodies NS D1 cytoplasm NS D1 membrane NS D1 nucleoli D1 pronuclei D1 zona pellucida NS Syngamy NS D2 zona pellucida NS D2 cell size NS D2 cell shape NS D2 perivitelline space D2 membrane NS D2 cytoplasm D2 fragments NS D2 developmental rate D2 cell number Note: n number of observations; NS not significant; D0 day of oocyte pick-up; D1 day after oocyte pick-up; D2 2 days after oocyte pick-up. had entered syngamy or first cleavage by 25 hours after insemination were not more likely to develop to the blastocyst stage than those that had not. The appearance of the cytoplasm on D2 is also significantly associated with blastocyst development (P.038). A large cytoplasmic deficit, seen as blastomeres not filling the space under the zona, seemed detrimental to blastocyst development (P.044). The appearance of the membrane and the extent of fragmentation were observed to not be significantly associated with blastocyst development. There was a significant negative association between the presence of MNBs and cell numbers on D2 (R 0.21; P.001) and D3 (r 0.11; P.05). A negative association between D2 MNBs and blastocyst development was also seen, albeit not statistically significant. The developmental rate is highly predictive of blastocyst development, with embryos being in the four-cell stage at 42 hours after insemination, having the highest blastocyst rate (Fig. 2). Of the D3 features, only developmental rate and blastomere shape were associated with blastocyst development in the univariate analysis (data not shown). The combined assessment of these variables is expressed in the D0, D1, D2, D2 corrected cumulative score, and D2 percentage scores. It appears from Table 7 that, although the D1 score is significantly associated with the degree of blastocoel development in vitro, neither the D0 score nor the D2 score are. However, if data from all days of culture are combined, and the D2 score is corrected for the presence of MNBs (i.e., the corrected cumulative D2 score), a statistically significant relationship with blastocoel development emerges. The D2 Percentage Score is also strongly associated with blastocoel development. In our system, the D3 scores were not superior to the D2 scores for predicting blastocyst development (data not shown). To summarize, the characteristics that showed the strongest association with blastocyst development were: cytoplas- TABLE 7 The association between different embryo scores and blastocoel development in embryos cultured to D5 or D6, estimated by the Spearman rank order coefficient of correlation, r. Blastocoel score vs. n r P value D0 score D1 score D2 score D2 corrected cumulative score D2 percentage score Fertility and Sterility 855

9 FIGURE 3 The association between implantation rate and the D2 weighted score (based on differential weighting of morphological characteristics) of transferred embryos. Numbers in parentheses above the bars are the number of embryos transferred in each category. mic appearance on D0, pronuclear and nucleolar appearance, whether the blastomeres on D2 filled the space under the zona, and the developmental rate. Furthermore, prediction of blastocyst development using data from all culture days was superior to using information from D2 only. To investigate the clinical usefulness of these findings, a new scoring system was developed, using differential weighting of characteristics (Table 3). The D2 weighted score of transferred embryos showed a strong correlation (r 0.96) with implantation rate (Fig. 3). DISCUSSION The present study analyses the usefulness of various morphological characteristics for predicting the developmental potential of an oocyte, zygote, or embryo, and it appears that five of these characteristics (appearance of the cytoplasm, pronuclei, and nucleoli, cytoplasmic deficit, and developmental rate) were better predictors than others. Furthermore, it is clear that by considering information from all developmental stages (D2 percentage score) a better prediction can be achieved, compared to only using D2 morphology, which is currently a common basis for embryo grading. Based on the present results, a new scoring system was designed, in which the morphological characteristics were differentially weighted. The D2 weighted score showed a strong correlation between the score of transferred embryo and implantation rate. The differential weighting of the characteristics for the D2 weighted score was based on the univariate statistical analysis, and those characteristics that were significantly correlated to blastocyst development were assigned more weight. For example, of zygote characteristics, pronuclear morphology accounts for more than 50% of the zygote score, in line with them being the only characteristics observed on D1 that were significantly associated with blastocoel development. Similarly, of the characteristics observed on D2 the developmental rate, the perivitelline space and the cytoplasmic appearance account for more than 50% of the score. In the legacy initial scoring system these variables accounted for less than 50% of the total score. A more valid weighting of variables could be obtained by using multiple regression techniques that are appropriate for the properties of the different variables. This requires appropriate definition and classification of characteristics, followed by suitable coding. Large numbers of observations are also required to make statistical evaluation feasible, and the present study is too small to permit such an analysis, although it offers valuable data to develop such a scoring system. A scoring system may be useful in three ways: 1. for selecting embryos for embryo transfer and freezing, 2. for training the eye to observe details in oocyte, zygote, or embryo morphology, and 3. for monitoring the performance of the IVF program, with changes in the frequency of occurrence of individual characteristics signaling changes in the system. Although much emphasis has been placed on the first item, the other two have been less discussed (see, however, Mortimer [13]). Judging from publications on IVF, it appears that many scoring systems currently used are rather crude, considering only two to three characteristics, such as fragmentation, cell numbers, and general appearance of the blastomeres, and only on the day of embryo transfer. It seems that traditional scoring systems do not tap the full potential of the observations of embryos. Previous reports (14 16) suggest that cleavage stage scoring systems have a limited ability to predict blastocyst development. This is also illustrated in the present article, where we cultured only embryos that were deemed unsuitable for freezing or transfer, and still obtained a number of good quality blastocysts. Hardarson et al. (17) also reported that good quality blastocysts can develop from embryos classified as suboptimal. Such embryos may be identified by sequential scoring systems that take into account morphological or biochemical observations during a longer period of time, rather than just a snapshot from one time point (4, 18 21). It has been suggested that, by using an effective scoring system, D3 transfers can achieve results on par with blastocyst culture (19). A disadvantage of a detailed, multistage scoring system is that it necessitates culture of single embryos in droplets, which is more time consuming. However, it does not seem detrimental in terms of embryo development (22, 23). In some clinics, transfer on D3 is preferable to transfer on D2. The rationale is that selection is improved as more developmental milestones have been passed on D3 (e.g., the shift from maternal to embryonic genomic control and com- 856 Sjöblom et al. Embryo score and blastocyst development Vol. 86, No. 4, October 2006

10 paction later on D3). We did not include such characteristics in our scoring system and it is possible that this explains why we did not find the D3 score superior to the D2 score. Numerous articles have been published on the relationship between oocyte morphology and embryonic development (24, 25). Antczak and van Blerkom (26) demonstrated a polarized distribution in the oocyte, zygote, and the embryo of several proteins of importance for the regulation of gene expression (e.g., transcription factors). This polarization was observed in the follicle, and it persisted through the preimplantation development. These findings increase our understanding of how the loss of cytoplasm and irregular cleavage can have an impact on development. Oocytes with a dark, coarse, or pitted cytoplasm, or in which aggregation of smooth endoplasmic reticulum (the diffuse, pancake-like structure referred to in Table 1A) occurred were more likely to be aneuploid, whereas those with vacuoles or necrotic patches did not show an increased aneuploidy rate (23, 26, 27). In our study, the blastocyst rate of oocytes displaying smooth endoplasmic reticulum aggregation was nil, and poor developmental potential of oocytes displaying this particular dysmorphism has been described previously (28, 29). Otsuki et al. (30) showed that occurrence of such oocytes in a treatment cycle was associated with poor outcome. In our study, in general, abnormal cytoplasm in the oocyte was associated with a significantly reduced blastocyst rate, and therefore should be considered when selecting embryos for freezing and transfer. In our clinic, embryos with smooth endoplasmic reticulum aggregation were not transferred. Although in our investigation the PB morphology did not correlate with blastocyst development, Ebner et al. (25) found a strong association between PB morphology on D0 and blastocyst development and implantation. They observed that oocytes, which had PBs that were round or ovoid, gave rise to embryos that had better implantation rates than oocytes that had fragmented PBs, or PBs that were not round or ovoid. The diverging results could possibly be explained by a difference between our study and those of Ebner et al. (25) in the emphasis on the different characteristics of the PB. Xia (31) graded oocytes with regard to the size of the perivitelline space, and fragmentation of the PBs. They found that PB fragmentation in itself only had a minor impact on ICSI fertilization rates. However, a large perivitelline space was associated with lower fertilization rates, an effect that was compounded by the presence of an abnormal PB. In the present study, cytoplasmic abnormalities in oocytes, mainly a large perivitelline space, were associated with a lower rate of blastocyst development. In the present study, the angle between the PBs on D1 did not appear to be associated with developmental potential, which is in agreement with the study by Garello et al. (32)on the relationship between PB orientation and embryo morphology (which presumably is related to developmental potential). There was also no significant association between the angle between the PBs and the shape or equality of size of the blastomeres (data not shown). A lower developmental potential of zygotes with abnormal pronuclear (PN) formation was reported by van Blerkom (33). He implicated epigenetic sperm dysfunction as a cause of the failure of PNs to migrate into apposition. The usefulness of scoring the PN and nucleolar morphology for predicting the developmental potential of zygotes has been explored by many researchers (32 36). From these data and from time-lapse microcinematography (37) a clearer picture has emerged of what constitutes normal zygote formation and what the implications are of aberrations. Thus, normally, the two PNs appear within a short interval and rapidly migrate to the center of the cytoplasm. Nucleoli, or nucleolar precursor bodies, first appear as scattered pinhead structures, which then coalesce. During the latter phase, the nucleolar position becomes polarized at adjacent poles of the appositioned PNs. Failure of PNs to migrate into apposition and asynchrony of nucleolar dynamics are associated with a lower developmental potential. In our study, PN and nucleolar morphology seemed to be strong determinants of blastocyst development. These characteristics should therefore be considered when selecting embryos for embryo transfer (and freezing). In the present study, the morphological characteristics observed on D2 were part of a legacy embryo scoring system, derived from the study by Mohr et al. (5). The appearance of the membrane was also examined, although its biological causes or significance has not been established. In the present study, there was no significant correlation between the appearance of the membrane on any day and the developmental potential. For this reason, this characteristic should be given a low weighting in a scoring system, but may still be included as a descriptor. Blastomere appearance (size and shape) is part of most embryo scoring systems and it is generally agreed that embryos with equal sized blastomeres have a better developmental capacity than embryos with irregular blastomeres (38, 39). A higher rate of chromosomal abnormalities in unevenly cleaved embryos was reported by Hardarson et al. (17), and this is significant as several investigators have demonstrated that embryos with chromosomal abnormalities have a disturbed development (40, 41) and this was also found in the study by Hardarson et al. (17). However, in our study, unequal cell size was only a minor indicator of poor developmental capacity. The presence of multiple or fragmented nuclei in blastomeres can create genetic and developmental disturbance (6 9). Multinucleated blastomeres occur in 5% 24% of the embryos and have been shown to be associated with a reduced developmental potential (7 9, 42 44). Nevertheless, the cited studies show also that embryos with MNBs do have a finite implantation potential. In the present study, a significant negative association was seen between cell numbers and presence of MNBs. If the chromosomal disorder Fertility and Sterility 857

11 caused by multinucleation or fragmentation of a nucleus of a blastomere leads to cleavage arrest in that blastomere, this would explain the reduced cell numbers in embryos with MNBs. However, we did not observe a significant association between blastocyst development rate and the number of MNBs, and this might be because we did not differentiate between multinucleation and nuclear fragmentation, believing that the impact on development would be similar. However, it has recently been shown that there is a difference between fragmented and multiple nuclei with regard to impact on development (44). The rationale for reducing the score proportionally to the fraction of multinucleated blastomeres was based on the assumption that a reduction in the number of healthy blastomeres would reduce the developmental potential as seen in cryopreserved embryos, in which blastomeres had been lysed with thawing. In the present study, no embryo with two MNBs developed into a good blastocyst (blastocoel score 3AA), whereas embryos with one MNB did, although at a reduced rate, which could justify the gradual reduction of the score with increasing numbers of MNBs. However, this may be of less clinical relevance, as transfer of embryos with MNBs is avoided when possible. The appearance of the ZP has been indicated as a predictor of development (45, 46). It is suggested that embryos with a zona of uneven thickness are more likely to implant, possibly because of hatching being easier. In the present study, the thickness of the ZP was not measured, but its degree of thickness variation and appearance was graded as normal or abnormal. Although measurement would have been desirable, it was not logistically feasible. The microscope was not equipped with image analysis software, and manual measurements would have been too time consuming. It is possible that the effect of an abnormal zona could have been significant, had we used image analysis software to accurately measure the zona thickness, instead of making estimates from the microscope image. The presence of cytoplasmic fragments is a commonly used characteristic in embryo scoring, as it is widely held that embryos showing excessive fragmentation of blastomeres have a poorer developmental capacity. However, in the present investigation, fragmentation was not a strong predictor of blastocyst development. It is probably more appropriate to evaluate the fragmentation patterns (47 49), as these have been suggested to be more prognostic than the mere presence or the extent of fragmentation. Early on, the cleavage rate was identified as a major determinant of development (38, 50, 51). This has since been confirmed (39), and it appears from the present study and others that embryos at the four-cell stage at 42 hours after insemination have a better developmental capacity than slower or faster embryos. Early cleavage or entry into syngamy at hours after insemination has also been shown to correlate well with developmental potential (52, 53). However, in the present study, involving smaller numbers, we could not demonstrate that syngamy or cleavage at 25 hours after insemination was significantly associated with blastocyst development, and the cause of this discrepancy is not clear. Lundin et al. (53) suggested that, because of the difference between embryos derived from IVF and ICSI with regard to when along the developmental time line they are observed, embryos generated through these treatments need to be assessed at different times after insemination/injection. Lundin et al. (53) also reported that embryos that were slow at the syngamy check also were retarded at the time of selection for embryo transfer. In keeping with this, we found in the present study a strong association between syngamy or cleavage at 25 hours after insemination and cleavage at 42 hours after insemination, indicating that slow embryos do not catch up. We also found a significant difference between IVF- and ICSI-derived embryos in the rate of syngamy or cleavage 25 hours after insemination. In the scoring systems used by Erenus et al. (54) and Grillo et al. (55), there seemed to be no interaction between the variables in the scoring system. In contrast, in the present study, we found that the different morphological features were not independent variables. The cell biological background of the observed features is likely to be common for several of them, and the actual evaluation of one feature is influenced by the presence of another. As it can be difficult to distinguish between a large anuclear cytoplasmic fragment and a small blastomere, it will be difficult to determine whether the blastomeres fill the space under the zona and to determine the number of blastomeres. However, it has recently been shown that, in D2 embryos, blastomeres 45 m rarely contain DNA (56). Such more precise classification of characteristics is important to further enhance the predictive power of embryo assessments. However, frequent discussions and comparisons between observers will reduce the variability in scoring. Morphological characteristics are notoriously hard to describe and unambiguously define, therefore complete agreement about classification can hardly be attained. This is one source of error or variability in the scoring system. Furthermore, perception of fine detail is obviously influenced by the vision of the observer, the quality and alignment of the optical equipment, and, although the latter is controllable, the former is another source of error in the scoring system. We are not aware of any publications about interobserver variability of embryo scoring (apart from objective measurements of ZP thickness by Loret de Mola et al. [57]), therefore it is not possible, at this stage, to form a judgment as to whether the interobserver CV of 18% seen in the present study is acceptable. However, the intraclass correlation coefficient was acceptable ( 0.70), and there was no statistically significant difference between the raters. As mentioned, the relative CV is (must be) greater for lower scores, and this inflates the size of the relative CV. 858 Sjöblom et al. Embryo score and blastocyst development Vol. 86, No. 4, October 2006

12 There is no doubt that well-designed, prospective studies yield information that is more valid than retrospective audits, which at best can only provide level II evidence. We firmly believe, however, that careful observation and documentation should be a part of clinical practice, and that it is imperative to analyze the findings with a view to improving practices. Decisions should be based on the best available evidence, which may well be of lower quality than the ideal level I. Findings from less robust studies are useful for the design of prospective trials. The present study is too small to permit a multivariate linear regression analysis of all of the relevant features, therefore the weighting of the individual characteristics is statistically less valid. However, it does offer valuable data for designing studies to further develop weighted scoring systems using such statistical techniques. Furthermore, we are well aware of the statistical difficulties caused by regarding individual embryos as independent observations, and by the fact that many of the observed characteristics are interrelated. In addition, with the study being performed in a clinical setting with continuous performance improvement, conditions for statistical inference could be further distorted. However, it is not easy to envisage how these difficulties could be avoided, considering the ethical limitations on research on human embryos. Most studies of embryo scoring use pregnancy as outcome measure rather than blastocyst development rate. Although pregnancy is clinically more relevant, this outcome is clearly also dependent on other variables apart from embryo quality, such as the endometrial receptivity and the skills of the operator at embryo transfer (58), which can confound an analysis of the effect of embryo quality on IVF success rates. Furthermore, it would be preferable to base estimates of predictive power on only the cases where all transferred embryos have the same characteristics, as in single embryo transfer, or where all embryos implant. In this study, we assume that the conclusions drawn from our highly selected population of oocytes, zygotes, and embryos (essentially the discards) are valid also for embryos that are deemed suitable for embryo transfer and freezing. We also assume that the ability of an embryo to develop into a blastocyst in vitro is related to the ability to develop similarly in vivo and to give rise to a healthy pregnancy (see similar discussion in Bolton et al [51]). The present researchers are well aware that not all blastocysts are developmentally normal, but embryos developing to the blastocyst stage are certainly more likely to be normal than embryos that have become arrested at an earlier stage. The validity of these assumptions is illustrated by the development of the weighted scoring system, which is based on observations of blastocysts development. A strong correlation (r 0.96) was seen between the D2 weighted embryo score and implantation potential, indicating its clinical utility and the validity of using blastocyst development as an indicator of developmental potential. The coefficient of determination (r ) suggests that 92% of the variation in implantation rate is explained by the variation in embryo quality described by the D2 weighted embryo score. The present study demonstrates the value of systematic observation and analysis of data collected from clinical practice. Although the quality of such data may be more variable than data obtained from experimental studies or controlled trials, judicious application of such information may steer a clinic to a path of improvement. To summarize, we have identified morphological characteristics from different points of time along the developmental time line that seem significantly associated with the ability of the embryo to develop to the blastocyst stage. Degenerative signs in the cytoplasm of the oocyte, such as accumulation of smooth endoplasmic reticulum or other cytoplasmic abnormalities, as well as a disordered formation of PN and nucleoli are associated with a poor developmental performance. There seems to be an optimum rate of cell division, and embryos that are slower or faster are less likely to develop into blastocysts. These findings create the foundations for weighting of individual characteristics when combined into a score, thus improving the predictive power of the scoring system. The present results suggest that by considering information from all stages of culture, selection of embryos for embryo transfer and freezing can be enhanced. We hope that this article will stimulate discussion and further improvement of embryo assessment in IVF/ICSI. Acknowledgments: The authors acknowledge the support and assistance of Dr. Stephen J. Steigrad, FRCSEd, FRCOG, FRANZCOG, head of the Department of Reproductive Medicine, Royal Hospital for Women; Dr. Graeme J. Hughes, FRCOG, FRANZCOG, School of Women s and Children s Health, University of New South Wales, both IVF Australia; and Dr. Julie Lukic, FRANZCOG, CREI, MMed, Department of Reproductive Medicine, Royal Hospital for Women, Sydney, Australia. REFERENCES 1. Lambert RD. Safety issues in assisted reproduction technology: the children of assisted reproduction confront the responsible conduct of assisted reproductive technologies. Hum Reprod 2002;17: Braude P, Bolton V, Moore S. Human gene expression first occurs between the four- and eight-cell stages of preimplantation development. Nature 1998;332: Blake DA, Proctor M, Johnson NP. The merits of blastocyst versus cleavage stage embryo transfer: a Cochrane review. Hum Reprod 2004;19; Ebner T, Moser M, Sommergruber M, Tews G. Selection based on morphological assessment of oocytes and embryos at different stages of preimplantation. Hum Reprod Update 2003;9: Mohr LR, Trounson A, Freemann, L. Deep-freezing and transfer of human embryos. J In Vitro Fert Embryo Transfer 1985;2: Munne S, Cohen J. Unsuitability of multinucleated human blastomeres for preimplantation genetic diagnosis. Hum Reprod 1993;8; Pickering SJ, Taylor A, Johnson MH, Braude PR. An analysis of multinucleated blastomere formation in human embryos. Hum Reprod 1995;10: Kligman I, Benadiva C, Alikani M, Munne S. The presence of multinucleated blastomeres in human embryos is correlated with chromosomal abnormalities. Hum Reprod 1996;11: Fertility and Sterility 859