Prediction of in vitro fertilization rates from semen variables

FERTILITY AND STERILITY Copyright il'1 1993 The American Fertility Society Printed on acid~ free paper in U.S.A. Prediction of in vitro fertilization rates from semen variables William W. Duncan Mary J. Glew Xin-Jun Wang, Ph.D.* Sean P. Flaherty, Ph.D. Colin D. Matthews, M.D. Department of Obstetrics and Gynaecology, The University of Adelaide, The Queen Elizabeth Hospital, Woodville, South Australia, Australia Objective: To assess the value of semen variables for predicting fertilization rates. Design: Measures of the fresh semen and the motile sperm fraction used for insemination were related to the fertilization rate by multiple regression analysis. The regression model was then used to construct a two-dimensional clinical chart. Setting: University-affiliated reproductive medicine unit. Patients: The results of 294 IVF cycles were analyzed retrospectively. Selection criteria were: [1] first cycle ofivf; [2] tubal and/or male factor infertility; and [3] four or more oocytes inseminated. Interventions: None. Main Outcome Measures: The fertilization rate was related to measured variables of the fresh semen and the motile sperm fraction used for insemination. Fertilization rate was categorized as poor (<35%) or acceptable (~35%). Results: Multiple regression analysis demonstrated a strong correlation between the fertilization rate and the combined indexes of percentage normal morphology and grade of motility in the fresh semen and percentage progressive motility in the motile sperm fraction. A two-dimensional chart that expressed these relationships was constructed. Its accuracy of prediction was 77% for poor fertilization and 95% for acceptable fertilization. Conclusions: The fertilization rate is strongly correlated with percentage normal sperm morphology in the fresh semen and the percentage progressive motility in the motile sperm fraction used for insemination. The clinical chart provides a simple but powerful tool for predicting fertilization outcome. Fertil Steril 1993;59:1233-8 Key Words: IVF, semen variables, multiple regression analysis, predictive power Received September 1, 1992; revised and accepted February 4, 1993. *Reprint requests: Xin-Jun Wang, Ph.D., Department of Obstetrics and Gynaecology, The University of Adelaide, The Queen Elizabeth Hospital, Woodville, South Australia 5011, Australia. The general relationship between the usual measures of fresh semen and the expectation of conception in vivo has been acknowledged, and certain measures have been recognized as distinguishing between fertile and infertile samples (1-3). The advent of IVF, however, has provided a direct means of measuring the fertility potential of a semen sample without interference from other infertility factors. In addition, IVF has become an important treatment option for male infertility, and although some studies have shown that IVF can be an efficacious treatment for male infertility, it is also recognized that fertilization rates are often lower than for normospermic males, resulting in fewer embryos for transfer and, consequently, a lowered expectation of conception (4, 5). It has thus become important to differentiate between male infertility patients who may benefit from routine IVF treatment and those who will not. Various studies have reported correlations between the fertilization rate and semen variables (sperm concentration, motility, morphology) in the fresh semen or motile sperm fractions (6, 7). More recently, there have been studies on the relationship between fertilization rates and specific tests of sperm function such as the ability of sperm to acrosome react or bind to the zona pellucida (8, 9). Overall, the results of these different studies show that Duncan et al. Prediction of fertilization rates 1233

variables such as the percentage normal morphology (10), some aspects of sperm motility (6, 11), the ability of sperm to undergo the acrosome reaction (8), and zona binding ratios (9) are strongly correlated with the fertilization rate. Although it is well recognized that these measures correlate with fertility, it is more important to know their predictive power and accuracy of prediction (12). Since male factor patients have a normal chance of conception once embryos have been transferred (5), a reliable method for predicting the fertilization rate during routine IVF would be invaluable to the clinician for counseling infertile couples with respect to the appropriateness of IVF and its possible outcome. With the advent of micromanipulative procedures such as zona cutting and subzonal sperm microinjection for the treatment of severe male infertility (13, 14), it is even more important to be able to predict the chances of success with routine IVF so that patients can be advised as to whether these latter options should be considered. In this study, we have constructed a simple clinical chart that can be used to predict fertilization rates, and we present an evaluation of its predictive accuracy. Patient Selection MATERIALS AND METHODS This retrospective study was based on 294 IVF cycles performed between 1987 and 1991. To reduce the influence of female and unexplained factors and to focus on the contribution of semen to the success or failure of fertilization, we only analyzed IVF cycles in which [1] it was the first cycle of treatment, [2] the infertility factors were tubal and/or male factor infertility, and [3] four or more oocytes were inseminated. Cycles in which a total of <50,000 motile sperm were recovered were excluded from the analysis because this represented a bias toward fertilization failure because of suboptimal numbers of sperm at insemination, rather than necessarily reflecting the fertilizing potential of the sperm. Oocyte Collection Follicular growth was stimulated with either the combined use of clomiphene citrate (Clomid; Merrill Dow Pharmaceuticals, North Sydney, New South Wales, Australia) and hmg (Humagon; Organon, Lane Cove, New South Wales, Australia and Pergonal; Serono, Frenchs Forrest, New South Wales, Australia) or with hmg used in concert with leuprolide acetate (Lucrin; Abbott Australasia, Kurnell, New South Wales, Australia) (15). Ovulation was induced with 5,000 IU ofhcg (Profasi; Serono), and oocytes were collected 34 to 36 hours later using transvaginal ultrasound-directed follicle aspiration or laparoscopic retrieval. Oocytes were matured for 4 to 6 hours in vitro in human tubal fluid (HTF) medium (16) supplemented with 7.5% maternal serum. Semen Preparation Semen was collected by masturbation after at least 2 days of sexual abstinence. After complete liquefaction, samples were analyzed according to World Health Organization (WHO) standards (17). Sperm morphology was assessed on Papanicolaou-stained smears using WHO guidelines for normal sperm morphology. For the purpose of this study, the following variables were selected: sperm concentration (X10 6 /ml), percentage normal morphology, percentage progressive motility, and grade of motility. The motility grade was subjectively assessed on a scale of 0 to 4 based on the vigor and degree of progression of the sperm. Semen was prepared for insemination by collecting a motile sperm fraction using either a swim-up procedure (n = 236) or discontinuous Percoll gradients (n =58) {18). Samples that had a sperm concentration of <20 X 10 6 /ml and/or <30% progressive motility, and grade< 2 were prepared on Percoll gradients. The variables analyzed for the isolated motile sperm fraction were as follows: sperm concentration, percentage progressive motility, total number of motile sperm, and the motile sperm concentration (motile sperm/ml). IVF Procedures Each oocyte was placed in 1 ml of HTF medium containing 7.5% maternal serum and inseminated with 30,000 to 50,000 motile sperm. Up to 100,000 motile sperm were inseminated in cases of severe male factor infertility in which the risk of fertilization failure was high. The gametes were cultured together at 37 C under an atmosphere of 5% C0 2, 5% 0 2, and 90% N 2 The oocytes were examined 17 hours after insemination for evidence of fertilization {presence of 2 polar bodies and 2 pronuclei), and embryos were checked for cleavage approximately 42 hours after insemination. The fertilization rate was calculated as the ratio of oocytes fertilized to oocytes inseminated, expressed as a percentage. 1234 Duncan et al. Prediction of fertilization rates Fertility and Sterility

Table 1 Stepwise Multiple Linear Regression Analysis of the Eight Semen Variables and the Fertilization Rate Variables in model Semen variable Regression coefficient SE T-value Probability F-enter Fresh semen Motility grade Percentage normal morphology Motile fraction Percentage motility 2.72 0.398 0.349 1.31 0.112 0.101 2.07 <0.05 4.28 3.54 <0.001 12.5 3.46 <0.001 12.0 Variables not in model Fresh semen Concentration Percentage motility Motile fraction Concentration Total motile Motile/mL Regression coefficient 0.094 0.036 0.003 0.004 0.108 F-enter 1.27 0.185 0.0013 0.0025 1.70 Statistical Analysis Statistical analysis was performed on an IBM compatible computer using the Statgraphics program (STSC Inc., Rockville, MD). Pearson's correlation analysis and forward stepwise multiple linear regression were undertaken to identify any relationships between the fertilization rate and specific semen variables. Statistical differences between the three fertilization groups (0% to 40%, 40% to 60%, 60% to 100%) with respect to each variable (morphology, motility, and motility in the isolated sperm fraction) were tested by ANOV A with a Scheff range test as the post hoc test (P < 0.05 considered significant). After plotting the percentage normal morphology (fresh semen) and percentage motility (motile fraction), the multiple regression formula was used to draw the line separating the low ( <35%) and acceptable (;:::35%) fertilization rate areas. Accuracy levels and 95% confidence intervals (CI) based on this predictive chart were assessed for the present data set as well as a new data set of 55 IVF cycles performed from January to March 1992. RESULTS Analysis revealed no significant differences between the linear regression lines of the swim-up and Percoll fractions for each variable (not shown), so the results were combined and analyzed together. The variables measured in the semen and the motile sperm fraction were significantly correlated with the fertilization rate with the exception of the sperm concentration and motile sperm concentration in the motile fraction. The correlations for the fresh semen variables were as follows: sperm concentration (r = 0.15, P < 0.02), percentage progressive motility (r = 0.36, P < 0.001), motility grade (r = 0.3, P < 0.001), and percentage normal morphology (r = 0.38, P < 0.001). The correlations for the motile fraction were: sperm concentration (r = 0.07, P = 0.27), percentage motility (r = 0.32, P < 0.001), total number of motile sperm (r = 0.15, P < 0.02), and motile sperm concentration (r = -0.12, P = 0.06). Since the positive correlations were relatively weak and therefore had little predictive power (and hence clinical value), forward stepwise multiple linear regression analysis was performed using the same eight variables (Table 1). The combined indexes of percentage normal morphology and grade of motility in the fresh semen together with percentage motility in the motile sperm fraction were selected by this analysis as having the strongest relationship with the fertilization rate (R 2 = 0.906). The percentage normal morphology, percentage motility (fresh semen), and the percentage motility (motile fraction) were each plotted against the fertilization rate (at 10% intervals). Three fertilization groups (0% to 40%,40% to 60%, and 60% to 100%) could be distinguished; AN OVA indicated statistically significant differences (P < 0.001) between the three groups in all three variables. However, because Duncan et al. Prediction of fertilization rates 1235

there was some variation in the cutoffs for the three groups with respect to the different variables and our aim was to identify semen values that characterized cycles with poor fertilization, a final cutoff of 35% was selected for the poor fertilization group and the two higher fertilization groups (35% to 60% and 60% to 100%) were combined. In Figure 1, we have plotted the percentage normal morphology (fresh semen) and percentage motility (motile sperm fraction) with the fertilization rate represented by different symbols to visually represent the multiple linear regression model. An area that contained the majority of cycles with <35% fertilization rate was delineated by the model, and this group was termed the poor fertilization group. The remaining area contained the majority of cycles in which fertilization was ~35% and was termed the acceptable fertilization group. Attempts to delineate three separate fertilization rate groups were less successful. The accuracy level and CI for the prediction of fertilization using these classifications were calculated from the retrospective data set (Table 2). The model accurately predicted 77% of poor fertilization cycles and 95% of acceptable fertilization cycles. A new set of data was then obtained from 55 IVF cycles performed between January and March 1992, and these data were used to determine the accuracy of prediction of the chart in a prospective manner. Comparable results were obtained with a 75% accuracy in predicting poor fertilization cycles and 88% accuracy for acceptable fertilization cycles (Table 2). 0:.. -~ u.!::: 0 0 e <35% FERT.RATE 0 ~35% FERT.RATE Normal Morphology(%)(Fresh semen) Figure 1 Two-dimensional chart showing the relationship between percentage normal morphology in fresh semen, percentage motility in the isolated motile fraction, and the fertilization rate. Cycles with fertilization rates of <35% or ~35% are indicated by different symbols. The curved line was predicted by the regression model; most cycles with <35% fertilization are below the line, whereas most cycles with ~35% fertilization are above it. Table 2 Accuracy of Predicting the Correct Fertilization Rate Using the Regression Model No. of No. correctly Fertilization rate cycles classified 95% CI Retrospective data (1987 to 1991) <35% 65 50 (77.0)* 65 to 86 ~35% 229 217 (94.8) 91 to 97 Prospective data (1992) <35% 12 9 (75.0) 43 to 95 ~35% 43 38 (88.0) 75 to 96 * Values in parentheses are percents. DISCUSSION In this study we have assessed the value of standard semen analysis measures for predicting fertilization rates during human IVF. Initially, we assessed the correlation between individual semen variables and fertilization rates. We confirmed previous studies showing that the percentage normal morphology in the fresh semen has the strongest positive correlation with the fertilization rate (6, 7, 10), although by itself it was not a useful predictor of IVF outcome (19). The percentage progressive motility in the fresh semen also correlated with the fertilization rate, though at a slightly lower rate than normal morphology. Some previous studies found motility to have the strongest relationship with fertilization rate (6, 20), whereas others have reported it to be less important than morphology (7). The sperm concentration in fresh semen and in the motile fraction correlated very weakly or not at all with the fertilization rate as reported by others ( 6, 7). An isolated motile sperm fraction is used for insemination of oocytes, so analysis of the motile fraction rather than the fresh semen sample could be expected to provide more information on the fertilizing potential of the sample (20, 21). In this study, however, the percentage progressive motility of fresh semen had a slightly stronger relationship with fertilization rate than the motility of the isolated motile fraction (r = 0.32 and 0.36, respectively), though neither was useful for predicting fertilization rates as a single variable. Bongso et al. (21) suggested that the "intensity" of motility (i.e., grade) of the swimup fraction reflected the real fertilizing capacity of semen. In the present study, the grade of motility was only assessed in the fresh semen, and because it is a subjective assessment, it is difficult to compare motility grades or their influence between studies. Certainly, our findings do suggest that the lowest 1236 Duncan et al. Prediction of fertilization rates Fertility and Sterility

grade of motility of the fresh sample correlated with a poor fertilization outcome. Because fertilization is likely to be influenced by a number of factors and the correlation coefficients for single variables were low, we employed forward stepwise multiple linear regression to investigate the combined relationships of various semen variables to the fertilization rate. This analysis resulted in a strong correlation (R 2 = 0.906, P < 0.0001) when percentage normal morphology and grade of motility in the fresh semen and percentage motility in the motile fraction were included in the model. The other measures did not add further to the predictive power. Barlow et al. (20) analyzed fertilization outcomes using multiple regression and also found a reasonable correlation of the percentage normal morphology and swim-up motility with the fertilization rate but gave no indication of the precision of their prediction. It should also be remembered that other seminal measures (not included in the regression model) may also be important. Although a regression equation to predict the fertilization rate was possible, we believed it to be of limited value and less effective than a graphic representation of the relationship. Hence, a chart was constructed to illustrate the relationships determined by the multiple regression analysis. For the purposes of the chart, the fertilization rate was classified as either poor ( <35%) or acceptable (35% to 100%). It is well recognized that IVF pregnancy rates (PRs) are directly related to the number of embryos transferred (22), and our own experience is that optimal PRs are achieved when three embryos are transferred. Given that approximately 7 oocytes were collected on average in each IVF cycle, a fertilization rate of <35% would result in less than three embryos for transfer, which in turn would lower the expectation of pregnancy for the patient. In contrast, a fertilization rate of ~35% would result in three or more embryos for transfer in most cycles, thus optimizing the chance of conception. Hence, a 35% fertilization rate represented a useful clinical cutoff value. It should be noted that up to 12 oocytes were collected in some cycles, and in these instances an adequate number of embryos may have been available for transfer despite a poor fertilization rate. We chose to generate a visual chart incorporating only two variables (percentage normal morphology in fresh semen and percentage motility in motile fraction) and the fertilization rate rather than simply setting cutoff values for each variable below which fertilization rate is greatly reduced. This has the major advantage of compensating for the effects of both variables. Indeed, no single cutoff value for either variable was able to satisfactorily distinguish poor from acceptable fertilization rates, and thus the use of single variables was inappropriate. Nevertheless, some couples with poor fertilization ( <35%) locate in the acceptable fertilization area. This indicates that although the multiple regression model accounts for much of the variability, other unaccounted factors are also involved. Defects of sperm function (8, 9) or oocyte quality (23) might be the source of this variability. Although many studies have investigated the relationship between semen variables and fertilization rates, few have established the accuracy of prediction and hence the predictive power of these correlations (19, 24). We tested the predictive power of the multiple regression model on data from the 294 IVF cycles and found that it correctly predicted 77% of poor fertilization cycles and 95% of acceptable fertilization cycles. Furthermore, comparable levels of accuracy were obtained when the model was tested in a prospective series of 55 IVF cycles. Such testing was considered an important part of validating the model. More accurate levels of prediction could be obtained when extreme grades of motility in the fresh semen (i.e., grade 1 or grade 3) were incorporated, although this was of limited value because most samples had an intermediate motility grade of 2. The model is clinically useful and should enable a more informed clinical decision as to whether or not to proceed with IVF based on the expectation of normal fertilization or otherwise. Our results contrast those of Ramsewak et al. (24) who described a poor ability to predict the fertilization rate ( 58.3%) and consequently claimed that IVF was required to test fertilization. The use of a regression model to construct a simple visual chart represents a novel approach and is useful for the clinical management of patients as the chart provides the clinician with an easy to use visual tool. However, the chart may not be directly applicable to other laboratories in its present form because of differences in semen analysis and IVF procedures. Acknowledgments. The authors thank all those members of the Reproductive Medicine Unit who participated in aspects of this study as part of their routine work and Ms. Carol Burford for assisting with preparation of the manuscript. REFERENCES 1. Aitken R.J, Best FSM, Richardson DW, Djahanbakhch 0, Lees MM. The correlates of fertilizing capacity in normal fertile men. Fertil Steril 1982;38:68-76. Duncan et al. Prediction of fertilization rates 1237

2. Aitken RJ, Best FSM, Richardson DW, Djahanbakhch 0, Templeton A, Lees MM. An analysis of semen quality and sperm function in cases of oligozoospermia. Fertil Steril 1982;38:705-11. 3. Silber SJ. The relationship of abnormal semen parameters to male fertility. Hum Reprod 1989;4:947-53. 4. Cohen J, Edwards R, Fehilly C, Fishel S, Hewitt J, Purdy J, et a!. In vitro fertilization: a treatment for male infertility. Fertil Steril 1985;43:422-32. 5. Yates CA, De Kretser DM. Male-factor infertility and in vitro fertilization. J In Vitro Fert Embryo Transf 1987;4:141-7. 6. Mahadevan MM, Trounson AO. The influence of seminal characteristics on the success rate of human in vitro fertilization. Fertil Steril 1984;42:400-5. 7. Liu DY, DuPlessis YP, Nayudu PL, Johnston WIH, Baker HWG. The use of in vitro fertilization to evaluate putative tests of human sperm function. Fertil Steril1988;49:272-7. 8. Cummins JM, Pember SM, Jequier AM, Yovich JL, Hartmann PE. A test of the human sperm acrosome reaction following ionophore challenge. Relationship to fertility and other seminal parameters. J Androl 1991;12:98-103. 9. Liu DY, Lopata A, Johnston WIH, Baker HWG. Human sperm-zona pellucida binding, sperm characteristics and invitro fertilization. Hum Reprod 1989;4:696-701. 10. Kruger TF, Acosta AA, Simmons KF, Swanson RJ, Matta JF, Oehninger S. Predictive value of abnormal sperm morphology in in vitro fertilization. Fertil Steril1988;49:112-7. 11. Liu DY, Clarke GN, Baker HWG. Relationship between sperm motility assessed with the Hamilton-Thorn motility analyzer and fertilization rates in vitro. J Androl 1991;12: 231-9. 12. Amann RP. Can the fertility potential of a seminal sample be predicted accurately? J Androl 1989;10:89-98. 13. Payne D, McLaughlin KJ, Depypere HT, Kirby CA, Warnes GM, Matthews CD. Experience with zona drilling and zona cutting to improve fertilization rates of human oocytes in vitro. Hum Reprod 1991;6:423-31. 14. Cohen J, Alikani M, Malter HE, Adler A, Talansky BE, Rosenwaks Z. Partial zona dissection or subzonal sperm insertion: microsurgical fertilization alternatives based on evaluation of sperm and embryo morphology. Fertil Steril1991;56: 696-706. 15. Sathanandan M, Warnes GM, Kirby CA, Petrucco OM, Matthews CD. Adjuvant leuprolide in normal, abnormal, and poor responders to controlled ovarian hyperstimulation for in vitro fertilization/gamete intrafallopian transfer. Fertil Steril 1989;51:998-1006. 16. Quinn P, Kerin JF, Warnes GM. Improved pregnancy rate in human in vitro fertilization with the use of a medium based on the composition of human tubal fluid. Fertil Steril 1985;44: 493-8. 17. World Health Organization. WHO Laboratory manual for the examination of human semen and semen-cervical mucus interaction. 2nd edition. Cambridge: The Press Syndicate of the University of Cambridge, 1987:5-10. 18. Kirby CA, Flaherty SP, Godfrey BM, Warnes GM, Matthews CD. A prospective trial of intrauterine insemination of motile spermatozoa versus timed intercourse. Fertil Steril 1991;56: 102-7. 19. Jeyendran RS, Schrader SM, van der Ven HH, Burg J, Perez Pelaez M, Al-Hasani S, eta!. Association of the in-vitro fertilizing capacity of human spermatozoa with sperm morphology as assessed by three classification systems. Hum Reprod 1986;1:305-8. 20. Barlow P, Delvigne A, Van Dromme J, Van Hoeck J, Vandenbosch K, Leroy F. Predictive value of classical and automated sperm analysis for in-vitro fertilization. Hum Reprod 1991;6:1119-24. 21. Bongso TA, Ng SC, Mok H, Lim MN, Teo HL, Wong PC, et a!. The influence of sperm motility on human in vitro fertilization. Arch Androl 1989;22:185-90. 22. Edwards RG, Fishel SB, Cohen J. Factors influencing the success of in vitro fertilization for alleviating human infertility. J In Vitro Fert Embryo Transf 1984;1:24-8. 23. Van Blerkom J, Henry G. Oocyte dysmorphism and aneuploidy in meiotically mature human oocytes after ovarian stimulation. Hum Reprod 1992;7:379-90. 24. Ramsewak SS, Cooke ID, Li T-C, Kumar A, MonksNJ, Lenton EA. Are factors that influence oocyte fertilization also predictive? An assessment of 148 cycles of in vitro fertilization without gonadotropin stimulation. Fertil Steril 1990;54: 470-4. 1238 Duncan et al. Prediction of fertilization rates Fertility and Sterility