* Present address: Foothills Hospital, Calgary, Alberta, Canada.

FERTILITY AND STERILITY Copyright 1993 The American Fertility Society Vol. 59, No. 6, June 1993 Printed on acid-free paper in U.S.A. A randomized trial of in vitro fertilization versus conventional treatment for infertility Samuel Soliman, M.D. Salim Daya, M.B., M.Sc.*t John Collins, M.D. John Jarrell, M.D.:j: Department of Obstetrics and Gynaecology, McMaster University, Hamilton, Ontario, Canada Objective: To evaluate the effectiveness of IVF in couples with infertility. Design: Two hundred forty-five consecutive couples with infertility were randomized to receive one cycle of IVF treatment (experimental group) or to wait for a period of 6 months before receiving IVF treatment, during which time other infertility treatments could have been undertaken (control group). Setting: Patients were referred to the Fertility Clinic at Chedoke-McMaster Hospitals, a university-associated institution in Hamilton, Ontario, Canada, in which IVF has been offered to couples since 1984. Patients: Couples with infertility (mean duration of 65 months) not corrected by conventional treatment. They came from all socioeconomic classes, and the costs of IVF treatment, except medication, were covered by the Ontario Health Insurance Plan. Main Outcome Measure: Pregnancy was confirmed by ultrasound documentation of a gestational sac or histologic examination of tissue. Outcomes included livebirth, spontaneous abortion, and ectopic pregnancy. The overall pregnancy rate (PR) and the interval-to-pregnancy duration were compared in each group. Results: Univariate analysis demonstrated a significant beneficial effect of IVF treatment in patients with bilateral severe tubal disease. Although in other diagnostic categories the crude and cumulative PRs in the experimental group were higher than in the control group, the differences did not reach statistical significance. Among the early IVF group, those with endometriosis had significantly more pregnancies when compared with other diagnostic categories. Although IVF increases the likelihood of pregnancy by 4% with severe tubal disease, the overall 31% increase associated with IVF was not statistically significant. Conclusions: There was a significant difference in favor of treatment in patients with severe bilateral tubal disease. For couples with other causes of infertility, the confidence limits around the treatment effect included unity. To reject the null hypothesis of no treatment effect, a larger sample size or a meta-analysis to combine the results of similar trials is required. Fertil Steril1993;59:1239-44 Key Words: IVF, infertility, randomized controlled trial, effectiveness In vitro fertilization has become widely accepted for the treatment of infertility since its clinical Received November 23, 1992; revised and accepted February 24,1993. * Career Scientist, Ontario Ministry of Health, Canada. t Reprint requests: Salim Daya, M.B., M.Sc., Department of Obstetrics and Gynaecology, McMaster University, 12 Main Street West, Hamilton, Ontario, Canada L8N 3Z5. * Present address: Foothills Hospital, Calgary, Alberta, Canada. application was first reported in 1978. Pregnancy rates (PRs) per cycle vary from 15% to 25%, and livebirth rates range from 1% to 15% in different reports (1, 2). Although it may be assumed that the PR is better with IVF, several reports have demonstrated that spontaneous pregnancies can occur in infertile couples even when bilateral tubal occlusion is present (3). The effectiveness of IVF in other causes of infertility is debatable in view of the frequency of treatment-independent pregnancies in Vol. 59, No.6, June 1993 Soliman et al. Randomized trial of IVF for infertility 1239

these groups (4-6). To our knowledge, no randomized clinical trials in which the effectiveness of IVF treatment has been evaluated have been reported. Such trials are indicated with any new therapy, particularly with IVF, which has the highest cost of all infertility treatments. Trials designed to evaluate effectiveness, in contrast to efficacy of a treatment, take into account other aspects of therapy including compliance, withdrawals because of side effects, as well as efficacy (7). The purpose of this study is to evaluate the effectiveness from an infertility clinic's point of view, where the issue is relevant particularly to those clinics that have long waiting lists for treatment. MATERIALS AND METHODS Patients and Setting The study was conducted in the Infertility Clinic of Chedoke-McMaster hospitals, a regional referral center. The majority of couples enrolled in the study that had been approved by the hospital's ethics committee had previously been treated for their infertility by conventional means. All couples had a laparoscopy, two semen analyses, and an endocrine assessment that included measurement of PRL, P, FSH, and LH levels in the female. Tubal disease and endometriosis were identified by laparoscopy. Male factor diagnosis was assigned if either the sperm count was <2 X 1 6 ml or the progressive motility was <4%. Ovulation disorders were defined as the presence of irregular periods, oligomenorrhea, amenorrhea, or a midluteal phase serum P level < 15 nmol/l in at least three separate cycles. IVF Protocol A vaginal ultrasound (US) examination was performed on day 3 of the cycle for assessment of ovarian size and to exclude patients with large ovarian cysts. If a cyst >2 mm was present, the cycle was canceled. Commencing on day 1 of the menstrual cycle, patients self-administered.5 mg of leuprolide acetate (Lupron; Abbott Laboratories Limited, Montreal, Quebec, Canada) two times per day by subcutaneous injection. Patients attended a teaching session for self-injection before commencement of treatment. At an initial dose of 15 IU, hmg (Pergonal; Pharmascience Inc., Montreal, Quebec, Canada) was given daily by IM injection commencing on day 3 of the cycle. The dose was adjusted as required depending on the E 2 level. Daily endovaginal US examinations for follicle growth assessment were started when the E. level reached 275 pg/ml (1, pmoljl). When at least two follicles had reached 18 mm in diameter, 5, IU of hcg (Profasi HP; Pharmascience Inc.) was administered intramuscularly. Oocyte retrieval was performed vaginally under US guidance 36 hours later. Sperm for insemination of oocytes was prepared by a standard swimup technique using Percol (Pharmacia, Montreal, Quebec, Canada). Three hours after oocyte retrieval, 5, sperm were added to each culture dish containing an oocyte. The policy at our institution allowed the insemination of up to six oocytes per patient and required that all fertilized oocytes be transferred back to the patient. Embryo cryopreservation was not available during the study period. Uterine transfer of embryos was carried out approximately 48 hours after oocyte retrieval using a Tom Cat catheter (Sherwood Medical, St. Louis, MO). Design and Randomization Couples presenting with infertility of> 12 months' duration were included in the study. They were grouped into the following diagnostic categories: male factor, tubal disease, endometriosis, unexplained infertility, and ovulation defect. Women with a history of recurrent abortion and those who were pregnant at the time of randomization were excluded from the study. Couples were assigned, by means of a computer-generated list of random numbers, to either early IVF treatment within 6 months (treatment group) or to late IVF treatment after 6 months (control group). During the 6 months' waiting period for the control group, treatment other than IVF was permitted. Thus, treatment with one cycle of IVF was compared with 6 months of waiting for IVF treatment, during which time any other infertility treatment could have been undertaken. Randomization was done in batches consisting of approximately 8 couples each, in April and November 199 and April1991. Follow-up The trial ran from April 199 to October 1991. After the 6-month trial period, during which the treatment group received only one cycle of IVF treatment, both groups were offered IVF treatment on a first-come, first-served basis. Each couple had at least one IVF treatment cycle during this followup period unless they had withdrawn from the study or a pregnancy had occurred. This additional period of observation ended on December 31, 1991. Reasons for cancellation of IVF cycles were recorded; 124 Soliman et al. Randomized trial of IVF for infertility Fertility and Sterility

Table 1 Characteristics of Patients in Each Group IVF Control (n = 127) (n = 118) Probability Withdrew 28 (22)* 18 (15.2).17 Female age (mean ± SO) 33.4 ± 3.1 3.7 ± 3..1 Infertility duration (mean± SO) 7.2 ± 35.9 65. ± 3..23 Primary infertility 78 (61.4) 62 (52.5).16 * Values in parentheses are percents. included were pregnancy, poor ovarian stimulation with low E 2 levels, development of a single follicle, ovarian cysts, premature LH surge, and patient preference. Statistical Analysis The groups were compared with respect to age of the female partner, duration of infertility, parity, gravidity, and the distribution of the infertility diagnoses. Significance testing was performed with the X 2 test for proportions and the Student's t-test for means. Pregnancy was confirmed by the presence of a gestational sac on US examination or by pathological examination of products of conception in the event of spontaneous abortion or ectopic pregnancy (EP). Crude PRs were calculated and compared. Because pregnancies occurred over time and because the time to study termination was different for each patient, cumulative PRs were calculated using lifetable analysis (8). In this analysis, the date of randomization was taken as the starting time of the study. For patients who became pregnant in the early group, the last menstrual period (LMP) was used to calculate the time to pregnancy; for patients who did not continue in the study, the date of withdrawal was used as the time of censoring for the analysis. In the late IVF group, the termination date was either 6 months from randomization or the LMP date for those patients who became pregnant during this period. Univariate analysis was performed to identify variables to be included in the regression analysis. Cox proportional hazard analysis was used because it is a multiple regression technique particularly suited to evaluating events that occur over time and when several variables that may be inter-related affect the outcome of interest (9). This analysis was carried out using the identified covariates to determine which of these variables had an independent effect on the time to pregnancy. RESULTS Of the 245 couples who were enrolled, 199 completed the trial and were kept under observation until December 31, 1991; 99 couples were in the early group (treatment) and 1 couples in the late group (control). Forty-six couples (18.7%) (28 in the early group and 18 in the late group) withdrew from the study for personal or social reasons. The mean (±SD) duration of infertility for the study population was 65 ± 3 months. The mean delay in receiving IVF treatment between the early (treatment) group and the late (control) group was 6 ±.5 months. The mean age of the female partner for the study population was 32 ± 3. years. Despite randomization, there was an imbalance in some potentially important covariates, namely, female age (Table 1) and the proportion with endometriosis (Table 2). In the early group of patients, 16 (16.2%) who started the treatment cycle were canceled for various reasons, 18 (18.2%) patients had no fertilization, and 65 (65.6%) patients had embryos transferred. In the control group, 27 (27%) had no treatment, and 73 (73%) had one or more treatments for their infertility. Treatments in the control group included lui in conjunction with US scanning to identify the optimal time by measuring follicle size in conjunction with superovulation using hmg or clomiphene citrate (n = 35), donor insemination (n = 8), and surgery for tubal disease before randomization (n = 22) or after randomization (n = 8). The number of pregnancies occurring in each diagnostic category and the outcome of these pregnancies are shown in Tables 3 and 4, respectively. During the study period (6 months from randomization), 17 (17.2%) patients in the early group and 8 (8.%) patients in the control group became pregnant (P =.5). However, after excluding the chemical pregnancies, the respective rates were 8% and 13.1% (P =.24). Thus, the observed difference in the overall clinical PR between the two groups was not statistically Table 2 Diagnostic Categories in Each Group Diagnosis IVF Control x' Probability Male factor 26 (26.2)* 34 (34.) 1.41.23 Tubal factor 35 (35.3) 42 (42.).93.34 Endometriosis 15 (15.1) 6 (6.) 3.58.6 Unexplained infertility 21 (21.2) 14 (14.) 1.79.18 Ovulation defect 2 (2.) 4 (4.).6.41 Total 99 1 * Values in parentheses are percents. Vol. 59, No.6, June 1993 Soliman et al. Randomized trial of IVF for infertility 1241

Table 3 Clinical Pregnancies in Each Diagnostic Category for the Two Groups During the Study Period Diagnosis IVF Control x' Probability Male factor 4/26 (15.3)* Tubal factor 3/35 (8.6) Bilateral tubal obstruction 2/15 (13.) Endometriosis 5/15 (33.3) Unexplained infertility 1/21 (4.7) Ovulation defect /2 () Total 13 3/34 (8.8).62.43 3/42 (7.1).5.82 /3 () 4.19.4 /6 () 2.63.11 2/14 (14.2).97.32 /4 () 8 1.39.24 * Values in parentheses are percents. significant. Although there was a significant difference in favor of IVF treatment among patients with severe bilateral tubal disease, the numbers are too small to draw any definite conclusions. Within the early IVF group, the best results were obtained in patients with endometriosis. The PR was significantly better in this group than in patients with tubal factor infertility (P =.3) and unexplained infertility (P =.5) (Table 3). Life-table analysis showed no significant difference between the early IVF and the control groups during the study period. The higher cumulative PR in the control group at the end of 21 months of observation (29%) compared with the early IVF group (2%) is not statistically significant. The observed difference could have resulted from more IVF cycles received by the control group because there was no limit on the number of IVF cycles a patient may undergo in this follow-up period. Using proportional hazards analysis, no variables had a significant effect on the time required to achieve a pregnancy (global X 2 = 1.14, P =.51) (Table 5). The following variables were included in the analysis: age of the female partner, duration of infertility, previous pregnancy, semen parameters, IVF treatment, and the presence of tubal disease, endometriosis, or ovulation disorder. In the follow-up period ending December 31, 1991, all patients had had at least one IVF treatment cycle, and there were 12 (12.1%) pregnancies in the early group and 14 (14.%) pregnancies in the control group, giving a total PRof 13%. The study period and the follow-up period were combined, and the life-table analysis presented graphically as cumulative PRs in Figure 1 shows no significant difference. DISCUSSION In vitro fertilization treatment was initially used for patients with bilateral tubal obstruction. Today IVF is being offered to almost any couple with infertility, particularly, those who have failed to become pregnant with conventional therapy. Several studies have reported good PRs with IVF, but no evaluation of its efficacy or effectiveness has been performed in randomized groups of infertile patients. Thus, the question of whether IVF is better than other therapies for infertility remains unanswered. This study was conducted to gain insight into this problem by attempting to answer the question of effectiveness. Within the study period in this trial, patients in the early IVF group were 1.31 times more Table 4 Pregnancy Outcome in Each Group for the Study Period IVF (n = 99) Control (n = 1) x2 Probability Total pregnancies Livebirth Abortion Chemical EP ET Mean time to pregnancy (d) Pregnancy per cycle (%) 17 (17.4)* 12 4 1 65 92 12 8 (8.) 6 2 82 1 3.81 2.27 2. 4.12 1.2 9.95.5.13.16.4.31.85.1 *Values in parentheses are percents. 1242 Soliman et al. Randomized trial of IVF for infertility Fertility and Sterility

Table 5 Proportional Hazards Analysis for the Trial Period Variable Infertility duration (y) Female age (y) Previous pregnancy in female Endometriosis Tubal factor Unexplained infertility Sperm density (X1 6 /ml) Sperm progressive motility (%) Sperm normal morphology (%) IVF treatment * EXP, exponent. t Values are the coefficients ± SE. Coefficient -.49 ±.8t -.5 ±.7.53 ±.47.9 ±.52 -.25 ±.59 -.5 ±.66 ±.1 ±.1.1 ±.1.27 ±.52 EXP* (Coefficient).95.94 1.7 2.47.77.6 1. 1. 1.1 1.31 likely to conceive than those in the control group who had to wait for 6 months. This observation suggests that early IVF treatment may be desirable. However, because the confidence interval (.27 to 2.35) around this odds ratio includes unity, the potential effectiveness of therapy remains unproven. If IVF treatment doubled the likelihood of pregnancy, our study had 8% power to show that with one-tailed a=.5. However, because the observed difference in this study was smaller, the power was only 45% at the same a level. In vitro fertilization appears to be effective in patients with bilateral tubal obstruction, suggesting that patient selection may be warranted, especially in programs in which resources are limited. In spite of randomization in our study, a difference between the two groups was observed in female age. Thus, the effectiveness of a single cycle of IVF may have been masked by this negative prognostic factor in the treatment group. The age of the female partner has previously been shown to be an important prognostic factor in couples with infertility (1, 11). Such potential biases can be reduced by conducting trials in one diagnostic group at a time or by stratifying the groups before randomization. Although the patients in this trial had a long duration of infertility (65 months) and a higher proportion with male factor infertility than those published in the American Fertility Society report (12), the PR with IVF was similar to the published rates. To address the question of whether IVF is more efficacious than conventional therapy, one could argue that because pregnancy is a terminal event that can occur only once in any given cycle, the best way to establish efficacy of IVF is to design a trial involving one cycle. Using one treatment cycle as the denominator, the IVF PR per cycle is significantly better (12%) than conventional therapy (1%) (P =.42); however, our study does not address this simple efficacy question. Consequently, one has to take into account patient withdrawals, delays in treatment, and other factors that may affect a program. The logistics of providing IVF treatment include waiting periods between treatment cycles, cancellations, and the likelihood of other external events that may interfere with arrangements for the IVF cycle. Thus, we believe that the evaluation of effectiveness should take time into account. Obviously, those patients who are destined to become pregnant in an IVF cycle will do so approximately 14 days from the date of IVF retrieval. In contrast, patients in the control group can become pregnant at any time during the 6-month trial period. Thus, the time to pregnancy is dependent on scheduling of cycles in the treatment group. Also, the question of cointervention may be raised because IVF patients may have had several cycles before and between treatments in which their risk of pregnancy per cycle is at least equal to that in the control group, especially if they had conventional treatment; hence, the first cycle rather than multiple cycles of treatment is the intervention tested. Further clinical trials with stratification by diagnostic category and other important variables are needed before the effectiveness of IVF can be evaluated in patients with infertility. The difficulty inherent in these trials cannot be ignored, especially because withholding treatment for prolonged periods in the control group may become ethically unacceptable. Nevertheless, CUMULATIVE PREGNANCY RATE (%) 3~------~------------------~===+ 1 1Z 14 16 18 ZO HONTH I EARLY IVF +LATE IVF Figure 1 Cumulative PRs for the study and follow-up periods in the IVF treatment and control groups. No significant difference (P =.13) was observed in cumulative PR between the IVF treatment(- e-) and control (+)groups at the end of the 6-month study period. Vol. 59, No.6, June 1993 Soliman et al. Randomized trial of IVF for infertility 1243

more clinical trials of IVF are urgently required so that consistency of treatment effect can be ascertained by meta-analysis; moreover, there should be a definitive study with a large sample size and adequate power. Acknowledgments. We thank Susan Ward, R.N., Ms. Karlee Dockstator, and all members of the IVF team at Chedoke McMaster Hospitals, Hamilton, Ontario, Canada, for their continuous efforts in making this study possible. REFERENCES 1. Beral V, Doyle P, Tan SL, Mason BA, Campbell S. Outcome of pregnancies resulting from assisted conception. Br Med Bull 199;46:753-68. 2. Medical Research International, Society for Assisted Reproductive Technology, The American Fertility Society. In vitro fertilization-embryo transfer (IVF-ET) in the United States: 1989 results from the IVF-ET registry. Fertil Steril 1991;55: 14-23. 3. Gomel V, McComb P. Unexpected pregnancies in women afflicted by occlusive tubal disease. Fertil Steril1981;36:529-3. 4. Collins JA, Wrixon W, Janes LB, Wilson EH. Treatmentindependent pregnancy among infertile couples. N Eng! J Med 1983;39:121-6. 5. Bernstein D, Levin S, Amsterdam E, Insler V. Is conception in infertility couples treatment related? A survey of 39 pregnancies. Int J Fertil1979;24:65-7. 6. Roh SI, Awadalla SG, Friedman CI, Park JM, ChinN, Dodds WG, et al. In vitro fertilization and embryo transfer: treatment-dependent versus -independent pregnancies. Fertil Steril 1987;48:982-6. 7. Robert HF, Suzanne WF, Edward HW. Clinical epidemiology the essentials. 2nd ed. Baltimore: Williams and Wilkins, 1988: 132-3. 8. Kalbfleisch JP, Prentice RL. The statistical analysis of failure time data. Toronto: John Wiley and Sons, Inc., 198:1-2. 9. Cox DR. Regression models and life tables. J Royal Stat Soc [B]1972;22:719-48. 1. Dor J, Romberg R, Rabau E. An evaluation of etiologic factors and therapy in 665 infertile couples. Fertil Steril 1977;28: 718-22. 11. Federation CECOS, Schwartz D, Mayaux MJ. Female fecundity as a function of age. N Eng! J Med 1982;36:44-6. 12. Medical Research International, Society of Assisted Reproductive Technology, and The American Fertility Society. In vitro fertilization-embryo transfer (IVF-ET) in the United States: 199 results from the IVF-ET registry. Fertil Steril 1992;57:15-24. 1244 Soliman et al. Randomized trial of JVF for infertility Fertility and Sterility