a Department of Obstetrics, Gynecology and Reproductive Biology, Brigham & Women s Hospital, Harvard Medical School,

Transcription

1 National collection of embryo morphology data into Society for Assisted Reproductive Technology Clinic Outcomes Reporting System: associations among day 3 cell number, fragmentation and blastomere asymmetry, and live birth rate Catherine Racowsky, Ph.D., a Judy E. Stern, Ph.D., b William E. Gibbons, M.D., c Barry Behr, Ph.D., d Kimball O. Pomeroy, Ph.D., e and John D. Biggers, D.Sc., Ph.D. f a Department of Obstetrics, Gynecology and Reproductive Biology, Brigham & Women s Hospital, Harvard Medical School, Boston, Massachusetts; b Department of Obstetrics and Gynecology, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire; c Department of Obstetrics and Gynecology, Baylor College of Medicine, Houston, Texas; d Department of Obstetrics and Gynecology, Stanford University, Palo Alto, California; e Arizona Reproductive Medicine Specialists, Phoenix, Arizona; and f Department of Cell Biology, Harvard Medical School, Boston, Massachusetts Objective: To evaluate the validity of collecting day 3 embryo morphology variables into the Society for Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS). Design: Retrospective. Setting: National database SART CORS. Patient(s): Fresh autologous assisted reproductive technology (ART) cycles from in which embryos were transferred singly (n ¼ 1,020) or in pairs (n ¼ 6,508) and embryo morphology was collected. Intervention(s): None. Main Outcome Measure(s): Relationship between live birth, maternal age, and morphology of transferred day 3 embryos as defined by cell number, fragmentation, and blastomere symmetry. Logistic multiple regressions and receiver operating characteristic curve analyses were applied to determine specificity and sensitivity for correctly classifying embryos as either failures or successes. Result(s): Live birth rate was positively associated with increasing cell number up to eight cells (<6 cells: 2.9%; 6 cells: 9.6%; 7 cells: 15.5%; 8 cells: 24.3%; and >8 cells: 16.2%), but was negatively associated with maternal age, increasing fragmentation, and asymmetry scores. An area under the receiver operating curve of (95% confidence interval ) was derived, with a sensitivity of 45.0%, a specificity of 83.2%, and 76.4% of embryos being correctly classified with a cutoff probability of 0.3. Conclusion(s): This analysis provides support for the validity of collecting morphology fields for day 3 embryos into SART CORS. Standardization of morphology collections will assist in controlling for embryo quality in future database analyses. (Fertil Steril Ò 2011;95: Ó2011 by American Society for Reproductive Medicine.) Key Words: Day 3 embryo morphology, evaluation, standardization, SART CORS, embryo fragmentation, embryo asymmetry, live birth rate The national Society for Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS) online database (1) and the companion National Assisted Reproductive Technology (ART) Surveillance System database (2) maintained by the Centers for Disease Control and Prevention have been successfully used for many studies of ART practice and ART outcome. These databases allow for studies that include a large number of ART cycles from multiple clinics, thus increasing the statistical validity and generalization of observations. Previous studies have included observations Received October 22, 2010; revised and accepted February 2, 2011; published online March 16, C.R. is a member of the scientific advisory board for Origio. J.E.S. has nothing to disclose. W.E.G. has nothing to disclose. B.B. has nothing to disclose. K.O.P. has nothing to disclose. J.D.B. has nothing to disclose. The cost of data collection was supported by SART. Reprint requests: Catherine Racowsky, Ph.D., Brigham & Women s Hospital, 75 Francis Street, ASB 1þ3, Room 082, Boston MA ( cracowsky@partners.org). on the health of ART offspring (3 5), effects of early fetal loss on ART outcome (6, 7), optimization of outcome of ART treatment as a function of number of embryos transferred (8, 9), and effects of race (10), basal metabolic index (11, 12), and single ET (13) on live birth rate, birth weight, and gestational age. Despite the usefulness of these datasets, some key information that could enhance and improve database studies has been missing. In recognition of one such omission, in 2007, the Centers for Disease Control and Prevention recently required the addition of weight and height for calculation of basal metabolic index. An additional, essential, critical missing component has been embryo morphology. Embryo morphology is known to be associated with implantation potential (14 16) and attaining live birth (17, 18) and its absence in the dataset has hampered studies of ART success rate and outcome. Collection of embryo morphology data has been complicated by the fact that morphology assessments are subjective and there are multiple available systems for assessing embryo quality (19 22). In 2007, embryo morphology fields of 2006 data were added to the SART CORS database in response to a request by SART member /$36.00 Fertility and Sterility â Vol. 95, No. 6, May doi: /j.fertnstert Copyright ª2011 American Society for Reproductive Medicine, Published by Elsevier Inc.

2 programs in This request was made because there were membership concerns that despite SART discouraging comparison of success rates among programs such comparisons are made, and there was no direct way to benchmark differences in patient populations. As embryo morphology is associated with implantation potential, it was decided that collection of morphology fields could provide a way to help assess patient reproductive potential. In addition, it was believed that such collection would help standardize grading systems among clinics, thereby helping not only in quality assurance and quality improvement activities, but also in reducing the number of embryos transferred. By establishing a standardized grading system and with associated live birth rates from a large number of embryos, clinics would have more assurance that pregnancy rates (PR) may not be compromised after transfer of fewer embryos. Voluntary collections by SART-affiliated clinics began in June The present study is one of two designed to evaluate the validity of these collections for day 3 embryos. In the present study we have focused on assessing the predictive value of variables collected for overall embryo morphology (i.e., cell number, fragmentation, and symmetry), as well as maternal age, according to whether their transfer resulted in a live born. MATERIALS AND METHODS Development of Collection System In 2004, a subcommittee of eight experienced embryologists was appointed by the SART Executive Council to develop a system for collecting morphology fields into SART CORS. This committee developed a system based on the guiding principles that the collection system must be [1] simple, [2] comprised of fields that have a basis in scientific inquiry with some proven predictive value; and [3] easily adopted in laboratories not routinely capturing these parameters. A system incorporating embryo cell number (1-cell embryos to those having >8 cells), fragmentation (0%, 1% 10%, 11% 25%, >25%, scored as 0, 1, 2, 3), and symmetry scores (perfect, moderately asymmetric, and severely asymmetric in size and shape of the blastomeres, scored as 1, 2, 3) was developed (Table 1) (23, 24). The system was implemented for collection of values for transferred embryos in fresh autologous cycles. Implementation of the System A letter was sent to program and laboratory directors in spring 2007 notifying them that the collection system was being launched on June 1, It asked them to encourage voluntary input of data for transferred embryos. TABLE 1 SART grading system for components of day 3 embryo morphology. SART parameters Cell no. 1, 2, 3, 4, 5, 6, 7, 8, >8 Scores for analysis Fragmentation 0% 0 <10% 1 11% 25% 2 >25% 3 Symmetry Perfect 1 Moderately asymmetric 2 Severely asymmetric 3 Note: SART ¼ Society for Assisted Reproductive Technology. Dataset for Assessing Validity Institutional Review Board approval was sought for this project but was deemed exempt by the Partner s Institutional Review Board on the grounds that already collected, de-identified data were to be analyzed. The SART CORS was queried to identify all fresh, autologous cycles that reported embryo morphology with embryos transferred on day 3 between June 2006 and December Additional inclusion criteria were: [1] transfer of the embryos either singly or in pairs, and [2] no clinical sac or a singleton in cases of single ET, or no clinical sac or dizygotic twins in cases of double ET. Those embryos transferred in pairs that resulted in singleton live births were not included in our dataset as we would not know which of the two transferred embryos resulted in the birth. Biometrical Considerations The models fit to the data were based on a model whose predictive value was previously demonstrated by us in a study using data from only one clinic (16). Univariate Analyses Preliminary univariate analyses were undertaken to determine the independent effects of maternal age, cell number, fragmentation score, and symmetry score on the presence or absence of a live born. The effect of maternal age was examined with logistic regression analysis. The significance of categorical data was tested using Fisher s exact test. Spearman correlation coefficients between the dependent variable (live born) and the covariates were also computed. P<.05 was considered to be statistically significant. Multiple Logistic Regression Analyses Logistic multiple regressions of the form Y ¼ logit P ¼ P b i x i were fit to the data using both maximum likelihood and bootstrap methods. The results obtained with the two methods of estimation were almost identical. The first term in the equation is a constant and the remaining terms are covariates determined by the results previously published on the data from the Brigham and Women s Hospital (16). The covariates were patient age, patient age squared, cell number, deviations of cell number from the mode, fragmentation score, and symmetry score. The fragmentation and symmetry scores were treated as factor variables. In one additional model, embryos having a fragmentation score of 3 (i.e., >25% fragmentation) were grouped with those with a fragmentation score of 2 (i.e., 11% 25% fragmentation) as previously described (16). In two other models, a fragmentation symmetry interaction term was included because of the observed high correlation coefficient between these factors, with and without grouping embryos having fragmentation scores of 2 and 3. The goodness of fit models was tested with the Hosmer-Lemeshow test, as previously described (24). The areas under the receiver operating curves (AUC) for each of the four derived models were then computed and their confidence limits determined (P¼.05) (25). The predictive values of the model with the highest AUC was assessed by calculating the sensitivity (i.e., the true positive rate) and the specificity (i.e., the true negative rate), assuming a cutoff probability equal to 0.3. All computations were performed using STATA10 (Stata Corp. LP, College Station, TX). RESULTS The final dataset comprised 7,528 embryos 1,020 embryos from single ET and 6,508 embryos from double ETs. Within this dataset, the fate of every embryo was defined as either resulting in or not resulting in a live born. There were 167 singleton live births from single ETs and 1,166 twin births from double ETs. Preliminary Univariate Analyses Maternal age Logistic regression was used to determine the effect of maternal age on the occurrence of live births. The following quadratic equation, rather than a linear equation was needed to adequately fit the data: logit (P) ¼ þ 1.059t t 2, where P is the probability of a live birth and t is the maternal age (16). The probability of a live birth increased to a maximum at about 29 years 1986 Racowsky et al. Validity of SART embryo grading system Vol. 95, No. 6, May 2011

3 FIGURE 1 Relationship between live birth rate (per embryo transferred) and (A) cell number, P<.0001; (B) fragmentation, P<.0001; and (C) blastomere symmetry, P<.001. of age and then decreased to almost zero by 40 years. This curved relationship was allowed for in the multiple regression analysis by including age squared in addition to age as a covariate. Number of cells on day 3 The range of the number of cells in the transferred embryos was one to more than eight. There was a significant quadratic association between cell number and incidence of live births (P<.0001) with the maximum number of live births developing from embryos comprised of eight cells (Fig. 1A). This nonlinear relationship was allowed for in the multivariate analysis by including the absolute deviation of the counts from the maximum [Celldev ¼ abs (8 - cells)] in the model. In SART CORS, the number of cells more than eight is not recorded exactly. Racowsky et al. (16) showed that reduced PRs are associated with these higher cell counts. In the current calculations a cell number more than 8 was replaced by a score of 10, which is the weighted mean of the cell counts more than 8 found in the previous study. Fragmentation on day 3 There was an overall significant decrease in the proportion of live births as the fragmentation score increased (P<.0001), with a less marked, albeit still significant, association observed between scores 0 and 1 (i.e., 0 and 1% 10%; P¼.038) (Fig. 1B). The nonlinearity of the effect of fragmentation was allowed for by treating fragmentation as a factor variable in the multivariate analysis. Symmetry on day 3 The proportion of live births decreased significantly, and approximately linearly, as asymmetry increased (P<.0001; Fig. 1C). Correlations All the Spearman correlation coefficients were highly significant. With the exception of one, all were small and the significance was due to the large number of observations on which they were based. One coefficient stands out, namely that between the fragmentation and symmetry scores (Table 2). Multiple Logistic Regression Analyses The computed AUCs obtained from the four models are summarized in Table 3. Neither grouping the fragmentation scores as previously done by Racowsky et al. (16), nor inclusion of the fragmentation symmetry interaction term had any significant effect on the AUCs and the overall classification rates. The percentage of embryos classified correctly was very similar for all four models, with an overall classification rate of approximately 75%. DISCUSSION The results presented in this study clearly show that the standardized morphology parameters recently adopted by SART for day 3 embryos have value for predicting live birth. When day 3 embryos are transferred, live birth was associated with increasing cell number to the 8-cell stage and was negatively associated with a greater percent fragmentation and increasing asymmetry. The effect of maternal age was also highly significant. Together, these observations suggest a robust significant correlation between these morphologic parameters and live birth. The current study was undertaken to demonstrate that the standardized system that SART adopted for day 3 embryos has validity at the national level. Data collected by laboratories across the TABLE 2 Spearman correlation coefficients between the dependent variable (live born) and the covariates. Age Cells Fragmentation Symmetry Live born Age 1 Cells Fragmentation Symmetry Live born Fertility and Sterility â 1987

4 TABLE 3 The goodness of fit, AUC estimates percentage classified correctly from the SART data. Model No. of observations P value a AUC Confidence limits (P[.05) Correctly classified (%) Fragmentation ungrouped 7, Fragmentation grouped 7, Interaction model 7, Interaction model 7, (fragmentation grouped) Fragmentation grouped b 1, Note: AUC ¼ area under the receiver operating curve. a Hosmer-Lemeshow goodness of fit test. b From Racowsky et al. (16). United States with embryos assessed by multiple embryologists having different training and experience must yield high predictive value for the system to have merit. In the present study we used a previously validated method for this assessment using the AUC (16). In this previous study, models were built using maternal age and characteristics late on day 1 (cell number and number of pronuclei), on day 2 (cell number, fragmentation and symmetry scores, and number of nuclei in blastomeres), and on day 3 (cell number, fragmentation and symmetry scores) to assess which characteristics should be recorded on which days to provide the greatest discrimination for predicting the probability of an embryo resulting in a viable fetus (to at least 12 weeks of gestation). We found that a single evaluation, on either day 2 or day 3, provides a similar predictive value to multiday scoring. This approach, having been used to show the strength of these three parameters as well as age for predicting implantation and development to a fetus of more than 12 weeks, was used in the current study with day 3 evaluations using a greater number of embryos and the end point of live birth per embryo. The goodness of fit results support the conclusion that the combined data from the different clinics was homogeneous. The high estimates of the AUCs and the classification rates show the general applicability of the previously derived model, which was obtained from data from only one clinic. Accurate collection of embryo morphology to the national database has a number of distinct advantages that were recognized by SART in developing this system. A robust national collection system will enable use of embryo morphology in studies of live birth rate and outcome using the national database. Such will supercede use of cryopreserved supernumerary embryos as a surrogate marker for embryo quality (13, 26, 27), as this surrogate is known to be a less than optimal marker of good morphology. Development of the SART morphology system has already led to increased national standardization of morphology assessments. Until this system was developed, there were no proficiency tests for morphology available as part of ART laboratory accreditation. These are now beginning to be developed. The American Association of Bioanalysis began distributing a twice yearly embryo morphology assessment challenge in April 2009 ( This challenge uses the SART collection standards for day 3 embryos as described in the present study, as well as for day 5 embryos, which will be evaluated in a subsequent study. Adoption of this challenge by the American Association of Bioanalysis represents an encouraging first step to improving national consistency in these assessments. Validation of the standardized system for morphology variables reported in the present article allows us to move forward using these collections to assist in further improving success for our patients with a history of infertility. Confirmation of the utility of the previously derived regression coefficients for day 3 morphology evaluations (16) justifies launching one of the models in Table 3 as a Webbased tool to enable available day 3 embryos to be prospectively ranked for each patient before transfer. The four models in Table 3 are so similar, additional research will be needed to determine which is superior. An ultimate goal would be for IVF laboratories of SART member clinics to be able to enter maternal age (i.e., oocyte age), cell stage, fragmentation, and symmetry, into the online tool that, in turn, would instantly rank selection of embryos for transfer. Data generated from individual clinics could then be used not only for in-house quality control and quality assurance appraisals, but also for optimizing the number of embryos to transfer to specific patient populations. As such, we expect that the use of this tool would result in fewer embryos being transferred, an increased utilization of elective single ETand maintenance, or even enhancement of national live birth rates with a reduction in multiple pregnancies resulting from IVF. We further expect that with increasing collection of morphology data into SART CORS, identification of even more accurate regression coefficients will be possible, which optimally will adjust for other variables such as body mass index (BMI), infertility diagnosis, parity, and gravidity. The value of collecting and standardizing embryo morphology is also now supported by the international community. At a recent special meeting of Alpha, an international association of embryologists, participants chose to adopt and implement a morphology system based largely on the one developed for SART (23, 24). This exciting and encouraging development will mean that embryologists around the world will soon be able to compare results and do collaborative studies using a standardized system. In addition, this collection system has already provided a platform for establishing the first proficiency test for morphology as part of ART laboratory accreditation. These collective efforts should improve national consistency in these assessments, and enable international collaborations in which embryo morphology is controlled in a standardized fashion. Acknowledgment: SART thanks all of its members for providing clinical information to the SART CORS database for use by patients and researchers. Without the efforts of SART members, this research would not have been possible Racowsky et al. Validity of SART embryo grading system Vol. 95, No. 6, May 2011

5 REFERENCES 1. Society for Assisted Reproductive Technology. Available at: wwwsartorg Last accessed March 14, Centers for Disease Control and Prevention. Available at: wwwcdcgov/art/2010. Last accessed March 14, Schieve LA, Meikle SF, Ferre C, Peterson HB, Jeng G, Wilcox LS. Low and very low birth weight in infants conceived with use of assisted reproductive technology. N Engl J Med 2002;346: Schieve LA, Ferre C, Peterson HB, Macaluso M, Reynolds MA, Wright VC. Perinatal outcome among singleton infants conceived through assisted reproductive technology in the United States. Obstet Gynecol 2004;103: Boulet SL, Schieve LA, Nannini A, Ferre C, Devine O, Cohen B, et al. Perinatal outcomes of twin births conceived using assisted reproduction technology: a population-based study. Hum Reprod 2008;23: Luke B, Brown MB, Grainger DA, Stern JE, Klein NA, Cedars MI. The effect of early fetal losses on singleton assisted-conception pregnancy outcomes. Fertil Steril 2009;91: Luke B, Brown MB, Grainger DA, Stern JE, Klein NA, Cedars MI. The effect of early fetal losses on twin assisted-conception pregnancy outcomes. Fertil Steril 2009;91: Stern JE, Goldman MB, Hatasaka H, MacKenzie TA, Racowsky C, Surrey ES, et al. Optimizing the number of blastocyst stage embryos to transfer on day 5 or 6 in women 38 years of age and older: a Society for Assisted Reproductive Technology database study. Fertil Steril 2009;91: Stern JE, Goldman MB, Hatasaka H, MacKenzie TA, Surrey ES, Racowsky C. Optimizing the number of cleavage stage embryos to transfer on day 3 in women 38 and older. Fertil Steril 2009;91: Fujimoto VY, Luke B, Brown MB, Jain T, Armstrong A, Grainger DA, et al. Racial and ethnic disparities in assisted reproductive technology outcomes in the United States. Fertil Steril 2010;93: Luke B, Brown MB, Stern JE, Missmer SA, Fujimoto VY, Leach R. Effect of maternal body mass index (BMI) on assisted reproductive technology (ART) pregnancy rates and obstetric outcomes. Fertil Steril 2009;92(Suppl):S Luke B, Brown MB, Stern JE, Missmer SA, Fujimoto VY, Leach R. Maternal obesity adversely affects assisted reproductive technology (ART) pregnancy rates and obstetric outcomes. Fertil Steril 2009;92(Suppl):S Luke B, Brown MB, Stern JE, Grainger DA, Klein N, Cedars M. Effect of embryo transfer number on singleton andtwinimplantationpregnancy outcomes after assisted reproductive technology. J Reprod Med 2010;55: Tesarik J, Junca AM, Hazout A, Aubriot FX, Nathan C, Cohen-Barcrie P, et al. Embryos with high implantation potential after intracytoplasmic sperm injection can be recognized by a simple, noninvasive examination of pronuclear morphology. Hum Reprod 2000;154: Van Royen E, Mangelshots K, De Neubourg D, Laureys I, Ryckaert G, Gerris J. Calculating the implantation potential of day 3 embryos in women younger than 38 years of age: a new model. Hum Reprod 2001;16: Racowsky C, Ohno-Machado L, Kim J, Biggers JD. Is there an advantage to scoring embryos on more than 1 day? Hum Reprod 2009;24: Ceyhan ST, Jackson KV, Racowsky C. Selecting the most competent embryo. In: Carrell DT, Racowsky C, Schlegel PN, Van Voorhis BJ, eds. Biennial review of infertility, Vol. 1., New York: Humana Press; p Cummins J, Breen T, Harrison K, Shaw JM, Wilson LM, Hennessey JF. A formula for scoring human embryo growth rates in in vitro fertilization: its value in predicting pregnancy in comparison with visual estimates of embryo quality. J In Vitro Fert Embryo Transf 1986;3: Rienzi L. Significance ofmorphologicalattributes of the early embryo. Repro Biomedicine Online 2005;10: Fisch J, Sher G, Adamowicz M, Keskintepe L. The graduated embryo score predicts the outcome of assisted reproductive technologies better than a single day 3 evaluation and achieves results associated with blastocyst transfer from day 3 embryo transfer. Fertil Steril 2003;80: Desai NN, Goldstein J, Rowland DY, Goldfarb JM. Morphological evaluation of human embryos and derivation of an embryo quality scoring system specific for day 3 embryos: a preliminary study. Hum Reprod 2000;15: Sioblom P, Menezes J, Cummins L, Mathiyalagan B, Costello MF. Prediction of embryo developmental potential and pregnancy based on early stage morphological characteristics. Fertil Steril 2006;86: Racowsky C, Vernon M, Meyer J, Ball GD, Behr B, Pomeroy KO, et al. Standardization of grading embryo morphology. Fertil Steril 2010;94: Racowsky C, Vernon M, Meyer J, Ball GD, Behr B, Pomeroy KO, et al. Standardization of grading embryo morphology. J Assis Reprod Genetics 2010;27: Cleves MA. From the help desk: comparing areas under the receiver operating characteristic curves from two or more probit or logit models. The Stata Journal 2002;2: Clayton HB, Schieve LA, Peterson HB, Jamieson DJ, Reynolds MA, Wright VC. A comparison of heterotopic and intrauterine-only pregnancy outcomes after assisted reproductive technologies in the United States from Fertil Steril 2007; 87: Kissin DM, Schieve LA, Reynolds MA. Multiple-- birth risk associated with IVF and extended embryo culture: USA, Hum Reprod 2005;20: Fertility and Sterility â 1989