Cumulative pregnancy rates for in vitro fertilization

Transcription

1 FERTLTY AND STERLTY Copyright 1986 The American Fertility Society Printed in U.8A. Cmlative pregnancy rates for in vitro fertilization David S. Gzick, M.D., Ph.D.* Charles ilkes, M.D.t Hoard. Jones, Jr., M.D.t Sothestern Medical School, Dallas, Texas, and The Hoard and Georgeanna Jones nstitte for Reprodctive Medicine, Eastern Virginia Medical School, Norfolk, Virginia Data on 575 coples ndergoing 1057 consective cycles of in vitro fertilization (lvf) ere sed to calclate cmlative pregnancy rates for repeated NF cycles. Exclding preclinical abortions and coples in hom the male partner had poor semen parameters, calclated cmlative pregnancy rates for cycles 1 to 6 ere 13.6%, 24.8%,37.2%,47.8%,52.2%, and 59.6%, respectively. A parametric model sed to fit these data yielded a strong correlation beteen observed and predicted pregnancy rates (r = 0.99, P < 0.001). Predicted cmlative pregnancy rates after 9 and 12 cycles ere 75% and 84%, respectively. Exclding preclinical abortions, the pregnancy rate per cycle as approximately constant, at approximately 15% over repeated cycles. As the cost of VF declines and as treatment cycles become more easily tolerated, persistence in VF can lead to sccessfl pregnancy for a large proportion of coples. Fertil Steril 46:663, 1986 Estimation of the likelihood of pregnancy for coples ndergoing in vitro fertilization (lvf) often is based on the pregnancy rate per cycle. 1-3 From the viepoint of the infertile cople at the start of the VF process, hoever, the relevant estimate is the pregnancy rate that can be expected after a specified nmber of cycles. n the early development of VF, pregnancy rates per cycle ere virtally identical to pregnancy rates per cople, becase fe coples had more than one cycle of treatment. As VF programs have matred and become more established, hoever, Received Febrary 3, 1986; revised and accepted Jne 2, *Reprint reqests: David S. Gzick, M.D., Ph.D., Department of Obstetrics and Gynecology, University of Texas Health Science Center at Dallas, 5323 Harry Hines Bolevard, Dallas, Texas tthe Hoard and Georgeanna Jones nstitte for Reprodctive Medicine, Eastern Virginia Medical School, Norfolk, Virginia. many coples enter the VF process ith the expectation of ndergoing repeated cycles if necessary. Ths it is important to obtain information on the probability of pregnancy per cople after specified nmbers of VF cycles. Sch information is important for to reasons. First, infertile coples ho are considering VF treatment can be conseled more accrately abot their overall chance of pregnancy in the VF program. This estimate of the probability of pregnancy can be evalated in an absolte sense, or in comparison ith an alternative treatment (e.g., tbal srgery for omen ith distal tbal disease). Second, the stdy of cmlative pregnancy rates over repeated VF cycles may provide some insight into the nderlying biology and technology of VF. For example, if the first VF cycle does not reslt in pregnancy, does this represent a level of biologic compromise that implies a redced probability of pregnancy in sbseqent VF cycles? Or if the first cycle fails for a given Gzick et al. Cmlative pregnancy rates for VF 663

2 cople, can e learn enogh from this experience to alter the protocol in sch a ay as to improve the probability of pregnancy? Or, finally, does the probability of pregnancy remain abot the same for a given cople from one cycle to the next? The prpose of this stdy as to estimate the cmlative probability of pregnancy reslting from repeated VF cycles. Althogh there is a groing consenss abot the se of cmlative pregnancy crves4-6 for the evalation ofinfertility treatment, this approach has had only limited se in the evalation of VF data. The only pblished description of cmlative pregnancy rates for VF is contained in a recent report from Norfolk. 7 n this paper, e expand pon the previos analysis to inclde (1) an pdate of the overall cmlative pregnancy crve for the Norfolk VF Program; (2) estimates of the ltimate probability of pregnancy per cople that cold be achieved ith present VF technology if cycles ere repeated a large nmber of times; (3) a comparison of cmlative pregnancy crves according to case of infertility; and (4) an evalation of the trend in the probability of pregnancy per VF cycle as cycles are repeated. MATERALS AND METHODS Since the inception ofthe VF program in Norfolk, data from each cycle have been entered into a data management system ritten for a personal compter by one of the athors (C. A..). Cases for the present report ere obtained from this data base and consist of all cycles condcted beteen September 1, 1981 and March 30, 1985 (series 2 to 18). Only cycles that reached the point of an egg-retrieval procedre ere inclded in the analysis. Protocols for ovarian stimlation, oocyte retrieval, gamete processing, embryo transfer, and management of the lteal phase and early pregnancy have been described.8-13 To perform the analysis of cmlative pregnancy, e sed the data management system to rite a ra data otpt file, hich as then sed as inpt into statistical packages on a V AX 11/780 minicompter (Digital Eqipment Corporation, Marlboro, MA). Observed cmlative pregnancy rates ere calclated according to the method of Berkson and Gage.14 For prposes of this analysis, all cycles in hich there as an attempted egg harvest ere inclded, hether or not eggs ere ob- tained. Also, preclinical abortions15 ere not conted as pregnancies. The observed vales of cmlative pregnancy rates for each cycle ere then fitted by a parametric model described previosly,5 in hich the cmlative probability of pregnancy after a nmber of cycles (P n) is related to the "cre rate" (c) and the probability of pregnancy per cycle among those cred (>.) as follos: P n = C (1 - e kn ) (1) n this model, c represents the proportion of coples in hom a pregnancy old be established by VF if cycles ere repeated many times. Ths the model takes into accont the possiblity that some coples ill not be "cred" of their infertility problem by VF-that is, they ill never achieve a pregnancy regardless of ho many times VF is repeated. Case of infertility as categorized into five grops: (1) tbal disease and obstrction; (2) endometriosis; (3) male factor; (4) nexplained infertility; and (5) miscellaneos, inclding immnologic problems, anovlation, and diethylstilbestrol exposre. Differences beteen cmlative pregnancy crves for different cases of infertility ere tested by comparing cmlative pregnancy crves ith the se of a Lee-Des statistic.16 t old have been preferable to compare the parametric estimates of these crves,6 bt sample sizes for the different grops ere too small to do this in a reliable manner. The trend in the probability of pregnancy per cycle as lvf cycles ere repeated as analyzed by performing a linear regression of monthly pregnancy rate on cycle nmber. One methodologic problem that arose ith this approach as that the sample size sed to estimate the probability of pregnancy for each cycle declined markedly as treatment cycles ere repeated. This reslted in increased variance, and hence redced reliability, of the estimate of the monthly probability of pregnancy. * To adjst for this, a eighted least sqares regression.ll as performed in hich the estimated variance of the pregnancy rate for each cycle as sed as the eights. "Proc Reg" of SAS18 as sed to obtain the eighted least sqares estimates. *The variance ofa sample proportion ifp(l-p)/n, here n = sample size and p = proportion. Ths the variance of an estimate based on small sample size is mch higher. 664 Gzick et a. Cmlative pregnancy rates for VF Fertility and Sterility

3 Table 1. Life Table Estimates of Cmlative Pregnancy Rates by Case of nfertility Case of infertility 1 2 Tbal No. completing cycles Pregnancies Cmlative pregnancy rate (%) Endometriosis No. completing cycles Pregnancies 8 2 Cmlative pregnancy rate (%) Male factor No. completing cycles Pregnancies 4 1 Cmlative pregnancy rate (%) Unexplained No. completing cycles Pregnancies 8 1 Cmlative pregnancy rate (%) Miscellaneos No. completing cycles Pregnancies 3 3 Cmlative pregnancy rate (%) All except male factor No. completing cycles Pregnancies Cmlative pregnancy rate (%) VF treatment cycle RESULTS Dring the stdy period, 575 patients had 1057 VF cycles in hich an egg retrieval as attempted. An abridged life table is shon for each case of infertility in Table 1, and cmlative pregnancy crves are plotted by case in Figre 1. The reslts for later cycles are often based on a small nmber of observations and are therefore less reliable than estimates for earlier cycles in hich the sample size is larger. There ere no significant differences beteen the cmlative pregnancy crves of coples ith different cases of infertility (Lee-Des statistic = 3.78; P = 0.43). Coples ith male factor infertility had a clinical pregnancy rate per cycle (5170 = 7.14%) that as abot one-half the rate of coples ith other types of infertility factors ( = 14.27%). Althogh the cmlative pregnancy crve for male factor coples as not statistically redced, clinical experience ith redced fertilization and transfer rates among male factor coples sggest that reslts based on a larger sample size old bear ot a redced pregnancy rate. Therefore, a "combined" cmlative pregnancy crve as calclated for all bt male factor cases. These "combined" estimates are shon in the last ro of Table 1. The reslts for cycles 7, 8, and 9 ere based on less than ten observations and ere exclded from estimation of or mathematical model of cmlative pregnancy. Estimation of this model for the combined grop based on the first six cycles yielded a cre rate of (standard error [SE] = 0.048; P < 0.01) and a probability ofpregnancy per cycle among those cred of (SE = >- Z 100 «z 90 0 [t: D Z 60 U 50 [t: 40 D.. 30 > «10...J :J 0 ::i :J U 3'ET,bl1 "ifili!cuonplli.. j / Figre 1 Observed cmlative pregnancy crves for VF, according to case of infertility. # ti--~--- ~~ /'J // / / VF TREATMENT CYCLES ~1a1. Gzick et al. Cmlative pregnancy rates for VF 665

4 >- Z { Z () 90 t: f-- Z 60 U 50 0 t: > - 20 f-- 00{ 10 -.J ::J 0 ::;; ::J U VF TREATMENT CYCLES Figre 2 Observed cmlative pregnancy crves for combined grop (sqares) and predicted crve estimated from mathematical model (smooth line) ; P < 0.01). Ths the predicted cmlative pregnancy crve (see eqation 1) as as follos: P n = (1 - e -O.153n) A plot of the observed and predicted cmlative clinical pregnancy crves for the combined grop is shon in Figre 2. The crve predicted by the model provides a close fit of the observed data. The predicted cmlative probability of pregnancy at 3,6,9, and 12 cycles is 37%,60%,75%, and 84%, respectively. A eighted least sqares regression of pregnancy rate per cycle (PPC) on cycle nmber (n) yielded the folloing eqation PPC = (n), (4.22) (0.18) here t statistics are shon belo the estimated coefficient. Ths the probability of clinical pregnancy per cycle as estimated to be essentially constant at 14.8% as treatment cycles ere repeated over time. The observed clinical pregnancy rate per cycle and the regression line throgh these points are shon in Figre 3. DSCUSSON For some time folloing the initial description of sccessfl hman VF, 19 there ere only scat- 666 Gzick et a!. Cmlative pregnancy rates for VF tered reports of live births reslting from VF. Crrently, becase a consistent record of sccess has been established in many clinics arond the orld, VF has been accepted generally by the medical commnity as a realistic treatment for many forms of infertility. Nonetheless, there is still some skepticism and misnderstanding abot the tre probability of a viable pregnancy reslting from VF. This misnderstanding has been fostered in part by reports in the literatre that focs on pregnancy rates per cycle. For a,cople faced ith the decision abot VF verss tbal srgery or VF verss adoption, the appropriate estimate of their probability of pregnancy is not the pregnancy rate per cycle, bt the expected likelihood of pregnancy after a specified nmber of cycles. There is a great financial and emotional brden sffered as a conseqence of each nsccessfl VF cycle. Ths if the first fe cycles are nsccessfl, the cople may not ish to ndergo additional cycles to achieve some theoretic probability of pregnancy. f, hoever, the cople can accept a particlar time horizon (e.g., three cycles), then ith the methods described here an estimate can be provided of the probability of viable pregnancy at the end of that time (e.g., 37% at the end of three cycles). Exclding male factor coples, a model of cmlative pregnancy that fit the observed data ell as predictive of a clinical pregnancy rate indistingishable from 100% if a large nmber of cycles ere repeated. Althogh extrapolation be- '".J U 0 30 > '" ~ 0 24 ~ 0 21 z ~ 0 18 o o ~o 151 ~n-----= n o > f J m 0 03 < m ~o OO~----._ ~ n. VF TREATMENT CYCLES Figre 3 Probability of pregnancy per cycle for combined grop (sqares) and eighted least sqares regression of probability of pregnancy per cycle on nmber of cycles. o Fertility and Sterility

5 yond 6 cycles is based on limited data, and mst be vieed as tentative, the predicted cmlative probability of pregnancy at the end of 12 cycles as 84%. This finding may become more salient as the cost and morbidiy of VF cycles declines, de to economies of scale and innovations sch as ltrasond aspiration of oocytes. 20, 21 Althogh a pregnancy rate of abot 15% per cycle may seem loer than that previosly reported by this center,7 preclinical abortions ere exclded from the cont of pregnancies in the nmerator, and all cycles in hich oocyte recovery as attempted(and not only those in hich embryos ere transferred) ere conted in the denominator. For male factor coples, data beyond to cycles are limited, so it is prematre to dra firm conclsions. t appears, hoever, that the probability of pregnancy per cycle is redced for these coples. The problem seems to be one of a redced fertilization rate, becase pregnancy rates per transfer are no different for male factor coples than for coples ith other cases of infertility.7 t shold be emphasized, hoever, that none of the differences observed beteen the cmlative pregnancy crves of the different cases of infertility reached statistical significance. The message in these data appears to be persistence. As the cost of VF declines, and as treatment cycles become more easily tolerated by the cople, persistence in VF can lead to sccessfl pregnancy for a large proportion of coples. REFERENCES 1. Edards RG, Fishel SB, Cohen J, Fehilly CG, Prdy JM, Slater JM, Steptoe PC, ebster JM: Factors inflencing the sccess of in vitro fertilization for alleviating hman infertility. J n Vitro Fert Embryo Transfer 1:3, Garcia J, Acosta A, Andres MC, Jones GS, Jones H Jr, Mantzavinos T, Mayer J, McDoell J, Sando B, Veeck L, hibley T, ilkes C, right GJr: n vitro fertilization in Norfolk, Virginia, J n Vitro Fert Embryo Transfer 1:24, Tronson A, ood C: n vitro fertilization reslts, , at Monash University Qeen Victoria, and Eporth Medical Centers. J n Vitro Fert Embryo Transfer 1:42, Cramer D, alker AM, Schiff : Statistical methods in evalating the otcome ofinfertility therapy. Fertil Steril 32:80, Gzick DS, Rock JA: Estimation of a model of cmlative pregnancy folloing infertility therapy. Am J Obstet Gynecol 140:573, Gzick DS, Bross DS, Rock JA: A parametric method for comparing cmlative pregnancy crves folloing infertility therapy. Fertil Steril 37:503, ilkes CA, Rosenaks Z, Jones D, Jones H Jr: Pregnancy related to infertility diagnosis, nmber of attempts, and age in a program of in vitro fertilization. Obstet Gynecol 66:350, Jones H Jr, Jones GS, Andres MC, Acosta A, Bndren C, Garcia J, Sando B, Veeck L, ilkes C, itmyer J, ortham JE, right G Jr: The program for in vitro fertilization at Norfolk. Fertil Steril 38:14, Garcia JE, Jones GS, Acosta AA, right GL Jr: Hman menopasal gonadotropinlhman chorionic gonadotropin folliclar matration for oocyte aspiration: Phase 1,1981. Fertil Steril 39:167, Garcia JE, Jones GS, Acosta AA, right G Jr: Hman menopasal gonadotropinlhman chorionic gonadotropin folliclar matration for oocyte aspiration: Phase, Fertil Steril 39:174, Jones H Jr, Acosta AA, Garcia JE, Sando BA, Veeck L: On the transfer of conceptses from oocytes fertilized in vitro. Fertil Steril 39:241, Jones GS: Update on in vitro fertilization. Endocr Rev 5:62, Masher S, Acosta AA, Garcia JE, Jones GS, Jones H Jr: Lteal phase serm estradiol and progesterone in in vitro fertilization. Fertil Steril 41:838, Berkson J, Gage RP: Calclation of srvival rates for cancer. Mayo Clin Proc 25:270, Jones H Jr, Acosta AA, Andres MC, Garcia JE, Jones GS, Mantzavinos T, McDoell J, Sando BA, Veeck L, hibley T, ilkes CA, right GL Jr: hat is a pregnancy? A qestion for programs of in vitro fertilization. Fertil Steril 40:728, Lee E, Des M: A compter program for comparing k samples ith right-censored data. Comp Prog Biomed 2: 315, onnacott RJ, onnacott TH: Econometrics. Ne York, John iley & Sons, 1970, p SAS nstitte nc. SAS Users Gide: Statistics, Version 5 Edition. Cary, NC; SAS nstitte nc., 1985, p Edards RG, Steptoe PC, Prdy JM: Establishing fllterm hman pregnancies sing cleaving embryos gron in vitro. Br J Obstet Gynaecol 87:737, Parsons J, Booker M, Gosamy R, Akkermans J, Riddle A, Sharma V, ilson L, hitehead M, Campbell S: Oocyte retrieval for in-vitro fertilization by ltrasonigraphically gided needle aspiration via the rethra. Lancet 1:1076, Lenz S, Laristen JG, Kjellon M: Collection of hman oocytes for in vitro fertilization by ltrasonically gided folliclar pnctre. Lancet 1:1163, 1981 Gzick et a. Cmlative pregnancy rates for VF 667