- Email: [email protected]
Calculating cumulative live-birth rates from linked cycles of assisted reproductive technology (ART): data from the Massachusetts SART CORS
Calculating cumulative live-birth rates from linked cycles of assisted reproductive technology (ART): data from the Massachusetts SART CORS Judy E. Stern, Ph.D.,a Morton B. Brown, Ph.D.,b Barbara Luke, Sc.D., M.P.H.,c Ethan Wantman, M.B.A.,d Avi Lederman,d Stacey A. Missmer, Sc.D.,e and Mark D. Hornstein, M.D.f a Department of Obstetrics and Gynecology, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire; b Department of Biostatistics, School of Public Health, University of Michigan, Ann Arbor, Michigan; c Department of Obstetrics, Gynecology, and Reproductive Biology and Department of Epidemiology, Michigan State University, East Lansing, Michigan; d Redshift Technologies, New York, New York; e Department of Obstetrics and Gynecology, Channing Laboratory, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, and Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts; and f Center for Reproductive Medicine, Department of Obstetrics, Gynecology and Reproductive Biology, Brigham and Women’s Hospital, Boston, Massachusetts
Objective: To determine the feasibility of linking assisted reproductive technology (ART) cycles for individual women to compare per-cycle and cumulative live-birth rates. Design: Historical cohort study. Setting: Clinic-based data. Patient(s): A total of 27,906 ART cycles with residency or treatment in Massachusetts during 2004–2006 and reported to the Society for Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS) on-line database. Intervention(s): None. Main Outcome Measure(s): Per-cycle and cumulative live-birth rates. Result(s): Linkage of cycles up to and including the first live-birth delivery revealed 14,265 women who averaged 1.9 1.2 SD cycles (range 1–11). These cycles yielded 9,452 pregnancies resulting in 7,675 live-birth deliveries. From cycle 1 to cycle 4, the cumulative live-birth rate for all patients increased from 30.4% to 43.3%, 49.1%, and 51.9%, respectively, and plateaued thereafter at about 53%. The cumulative live-birth rate after three cycles using donor oocytes was 60% for women aged <43 years and >50% for women R43 years; for autologous oocytes it was 60.1% for ages <35 years and declined steadily to 8.5% for ages R43 years. Conclusion(s): The results demonstrate the feasibility of linking ART cycles for individual women from SART CORS to characterize cumulative live-birth rates. (Fertil Steril 2010;94:1334–40. 2010 by American Society for Reproductive Medicine.) Key Words: Cumulative live-birth rate, live-birth rate per cycle, assisted reproductive technology, SART CORS
Success rates for assisted reproductive technology (ART) cycles have traditionally been reported as pregnancies per cycle according to maternal age. This cross-sectional measure, which can be easily calculated, is used by national ART registries in Europe, the Middle East, Australia and New Zealand, and North America, including the United States (1–3). This measure has limited usefulness, though, for the woman who undergoes multiple cycles. Longitudinal analysis linking all treatment cycles for an individual woman would have short- and long-term advantages. The immediate advantage is a more accurate estimate of success with continued
Received March 13, 2009; revised May 17, 2009; accepted May 28, 2009; published online July 10, 2009. J.S. has nothing to disclose. M.B. has nothing to disclose. B.L. has nothing to disclose. E.W. has nothing to disclose. A.L. has nothing to disclose. S.M. has nothing to disclose. M.H. has nothing to disclose. Supported by the Society for Assisted Reproductive Technology. Presented at the 64th annual meeting of the American Society for Reproductive Medicine, San Francisco, California, November 8–12, 2008. Reprint requests: Judy E. Stern, Ph.D., Department of Obstetrics and Gynecology, Dartmouth-Hitchcock Medical Center, Lebanon, NH 03756 (FAX: 603-650-0905; E-mail: [email protected]).
1334
treatment. Linking cycles would also facilitate assessing a woman’s future health risks from exposure to multiple treatment cycles over time. Concerns about the inadequacy of research on ART outcomes as a result of per-cycle reporting and other problems have been discussed previously (4). Studies that have linked treatment cycles have been limited by small sample size, inconsistent inclusion criteria and outcome measures, and statistical and methodologic shortcomings (5–12). The present project was undertaken to link cycles of ART from the national Society for Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS) on-line database. We evaluated a subset of cycles reported to this database from a single state to: 1) determine the feasibility of performing the linkage; and 2) compare the traditional measure of success of live-birth rates per cycle with cumulative live-birth rates per woman linking her cycles of treatment. We chose Massachusetts to demonstrate this linkage for several reasons. Massachusetts is one of only 14 states with mandated private insurance coverage for infertility therapy and one of only five states with full coverage, reducing an important barrier to care (13). As a result, in 2005 Massachusetts ranked fifth in highest number of ART cycles and ART births in the United States (14). In addition, all of the seven ART programs in
Fertility and Sterility Vol. 94, No. 4, September 2010 Copyright ª2010 American Society for Reproductive Medicine, Published by Elsevier Inc.
0015-0282/$36.00 doi:10.1016/j.fertnstert.2009.05.052
FIGURE 1 Study population and exclusions. SART CORS ¼ Society for Assisted Reproductive Technology Clinic Outcomes Reporting System.
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
Massachusetts report their data directly to the national SART CORS database.
MATERIALS AND METHODS The data source for this study was the SART CORS database, which contains comprehensive data from >90% of all reporting clinics performing ART in the U.S. Data were collected and verified by SART and reported to the Centers for Disease Control and Prevention in compliance with the Fertility Clinic Success Rate and Certification Act of 1992 (Public Law 102-493). The study was approved by the Committee for the Protection of Human Subjects at Dartmouth College, and all identifiable data were analyzed at the offices of the SART CORS programmers. Patient-identifiable data, such as names, dates, social security numbers, addresses, and zip codes, did not leave that location, and the datasets for subsequent analyses were stripped of identifiers, retaining only study numbers which were randomly assigned to link cycles. In this study we used data reported between January 1, 2004, and December 31, 2006, from patients who were either treated at Massachusetts ART centers or who had a Massachusetts zip code Fertility and Sterility
recorded at the time of treatment. This database includes in vitro fertilization with and without intracytoplasmic sperm injection (ICSI), frozen embryo transfer cycles, and cycles of other procedures, such as gamete intrafallopian transfer and preimplantation genetic diagnosis (PGD), that involve removal of oocytes from the woman, but it does not include intrauterine insemination or ovulation induction cycles. All cycle types (fresh, frozen, autologous egg, donor egg, donor sperm, PGD, banking, and research cycles) were included in the linkage analysis. The scheme to link cycles used identifying variables available in the SART CORS database, including date of birth, first and last names, social security number, and zip code. First and last names were adjusted to remove all nonletter characters and were coded using Soundex software (Soundex SQL Server 2000) to facilitate phonetic matches. Following Soundex coding, the first linkage used date of birth, last name, and first name to generate an initial random SART-Woman-ID. Because all cycles within a single clinic were generated from a single patient file, this linkage was highly accurate. In addition, an initial set of intraclinic linkages were revealed. A second linkage took these cycles at separate clinics and reordered them according to date of birth, first name, and last name to facilitate
1335
TABLE 1 Characteristics of the study population. Cycles
n
% of cycles
Total cyclesa Cycles at MA clinics Out-of-state cycles on MA residents Autologous oocyte cycles Donor oocyte cycles Fresh cycles Frozen embryo transfer cycles Use of ICSI in fresh cycles
27,906 26,778 1,128 25,695 2,211 23,959 3,947 9,914
100.0 96.0 4.0 92.1 7.9 85.9 14.1 41.4 (% fresh)
Total no. of women Pregnancies Ectopic pregnancies Fetal losses Stillbirths Live-birth deliveries
14,265 9,452 64 1,685 28 7,675
33.8 0.2 6.0 0.1 27.5
% of women
% of pregnancies
100.0 66.2 0.4 11.8 0.2 53.8
100.0 0.7 17.8 0.3 81.2
Note: ICSI ¼ intracytoplasmic sperm injection; MA ¼ Massachusetts. a Includes cycles performed at MA clinics plus all additional cycles to women with MA zip codes. Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
linkages between cycles in which the date of birth and first name were the same or similar but in which the last name was recorded differently or in modified format at subsequent clinics. Social security number and zip code (where present) as well as cycle order were used to confirm or refute the proposed matches from this second linkage. Social security number was found in only 47% of cycles and zip code in 67%. This, and the fact that some clinics had no entries for these identifiers on any cycles, made these of limited use for matching between clinics. Potential linkages were considered for hyphenated versus nonhyphenated names, first name and middle name combinations, and data entry errors, such as use of numbers or nonsensical characters in names. As a final measure for identifying accurate linkages among women with the same identifying variables, the outcome of a prior cycle and timing of a subsequent cycle were evaluated (i.e., a live birth in December 2005 followed by
TABLE 2 Assisted reproductive technology cycles per woman in Massachusetts cohort.
No. of cycles per woman 1 2 3 4 5 6 7 8 9 10 11 Total
No. of women with specified number of cycles
% of women
7,113 3,602 1,865 933 436 198 71 32 10 4 1 14,265
49.86 25.25 13.07 6.54 3.06 1.39 0.50 0.22 0.07 0.03 0.01 100.00
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
1336
Stern et al.
Linked ART cycles from SART CORS
a treatment cycle in January 2006 would indicate two different women, not linked cycles for the same woman). Fewer than 0.1% of cases suggested a match for which supporting information did not clearly permit a decision that the two cycles were indeed the same patient. Since we could not validate the data records against original clinical records, there may be errors due to data entry of names and/or dates of birth both when linking records within a clinic and when linking records between clinics. However, it is unlikely that the error rate due to these types of errors exceeds 2%–3%. After cycle matching, the data file was limited to exclude certain patients with specific types of cycles, as shown in Figure 1. The file was processed to delete banking cycles in which all embryos were cryopreserved and none were transfered (none of these patients had subsequent frozen embryo transfers). Also deleted were five cycles with unspecified treatment outcomes. Cycles were limited to those up to and including the first live birth delivery, i.e., all cycles were included for women in whom a live birth delivery did not occur, but cycles after the first delivery were excluded. Because women with or without frozen embryo transfer cycles may also have had ART cycles before 2004, we chose to include all women with ART cycles in the analysis. The data were analyzed either by total number of cycles, by cycle number, or by woman. Cumulative delivery rates represent deliveries per woman after repeated cycles of ART. These rates are not equivalent to the total deliveries from a single oocyte harvest (previously defined as total reproductive potential). Live-birth rates per cycle were computed for age groups <35, 35–37, 38–40, 41–42, and R43 years at the start of their first cycle with autologous versus donor oocytes. Because use of donor oocytes represents an entirely new phase of treatment, we renumbered the cycles with donor oocytes to start with 1. Both fresh and frozen embryo transfers were included. Differences in live-birth rates across the five age groups were analyzed using the c2 statistic. SAS 9.1 (SAS, Cary, NC) was used for all analyses.
RESULTS The cohort included 27,906 cycles of ART among 14,265 women resulting in 9,452 pregnancies with 7,675 live-birth deliveries. The Vol. 94, No. 4, September 2010
TABLE 3 Cumulative live-birth rates per woman.
Cycle of treatment 1 2 3 4 5 6 7 8 9–11
Live births (n)
Total cycles at each level (n)
Live birth/cycle (%)
4,331 1,848 825 396 169 68 25 10 3
14,265 7,152 3,550 1,685 752 316 118 47 21
30.4 25.8 23.2 23.5 22.5 21.5 21.2 21.3 14.3
95% CI
Cumulative live births (n)
Total women with ART (n)
Cumulative live birth/woman (%)
95% CI
29.6–31.1 24.8–26.9 21.9–24.6 21.5–25.5 19.5–25.5 17.0–26.1 13.8–28.6 9.6–33.0 0.0–29.3
4,331 6,179 7,004 7,400 7,569 7,637 7,662 7,672 7,675
14,265 14,265 14,265 14,265 14,265 14,265 14,265 14,265 14,265
30.4 43.3 49.1 51.9 53.1 53.5 53.7 53.8 53.8
29.7–31.1 42.5–44.1 48.3–49.9 51.1–52.7 52.3–53.9 52.6–54.4 52.9–54.5 53.0–54.6 53.0–54.6
Note: ART ¼ assisted reproductive technology; CI ¼ confidence interval. Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
characteristics of these cycles are presented in Table 1 and include 92.1% cycles with autologous oocytes and 14.1% cycles with frozen embryo transfers. Intracytoplasmic sperm injection was used to achieve fertilization in 41.4% of fresh cycles. The number of cycles for each woman is shown in Table 2. Women underwent between 1 and 11 cycles during this 3-year period, with a mean of 1.9 1.2 SD and a median of 2 cycles. Of the 14,265 women, 21% (3,000) had at least one frozen cycle and 10% (1,431) had at least one donor oocyte cycle. There were 7,113 women who underwent only one cycle, of whom 649 had frozen embryo transfers, indicating that this group included women with cycles before 2004. Table 3 shows the per-cycle and cumulative live-birth rate per woman as a function of cycle of treatment. The cumulative live-birth rate for all women was 53.8%. The live-birth rates by age group and autologous versus donor status are shown in Table 4. Only the first four cycles of treatment are shown, because the data for subsequent cycles (five or more) were too sparse to provide robust estimates. The live-birth rates per cycle with autologous oocytes declined significantly with age in all cycles, from 38.3% to 5.7% from the youngest to the oldest ages for the first cycle (P<.0001) and from 30.0% to 5.0% for the fourth cycle (P<.0001). In contrast, the live-birth rates with donor oocytes were not significantly different across age groups in all cycles, ranging from 46.8% to 40.3% from the youngest to the oldest ages for the first cycle (P¼.38) and from 29.4% to 21.9% for the third cycle (P¼.94). The cumulative live-birth rates per woman are shown in Figure 2. In line with the live-birth rates per cycle, the cumulative live-birth rates with autologous eggs differed substantially with advancing age. The cumulative live-birth rates among women with four cycles using autologous oocytes ranged from >60% for women aged <35 years to 8.7% among women aged R43 years. With donor oocytes, the cumulative live-birth rates among women with four cycles averaged about 60% for women aged %42 years and >50% for women aged R43 years.
DISCUSSION This is the first study to link successive cycles of treatment from the SART CORS database. We have used these linked cycles to determine the number of women, the number of cycles, and the outcomes for cycles performed at Massachusetts ART centers or for women with Massachusetts residency between January 2004 and December Fertility and Sterility
2006. Because we began to follow women at an arbitrary time point, the dataset includes women who may have had earlier cycles. Therefore, all estimates of live-birth rates are underestimated, because early cycles have the highest success rates. Censoring the data within this 36-month period also resulted in including women whose treatment was not finished by December 31, 2006. Most studies of ART have analyzed outcomes on a per-cycle basis; however, the more meaningful outcome measure is the outcome per woman. Per-cycle rates do not accurately measure the value of continued treatment for individual patients, nor do they allow determination of the effect of treatment on a woman’s long-term health. Because of the importance of determining these outcome measures, we performed the present analysis as the firstever attempt to assess the feasibility of using the SART CORS dataset to link ART cycles. We limited the analysis to patients in one geographic area, in a state with a large patient volume, with comprehensive reporting of all clinics to SART CORS, and with mandatory insurance coverage of ART services. Between-clinic linkages were possible but more problematic due to inconsistencies in entries of names and under-reporting of social security numbers and zip codes. The present study is the largest of its kind to date, based on 27,906 cycles among 14,265 women. We showed that over a 3-year period, patients underwent an average of 1.9 1.2 SD cycles of ART, which is in accordance with findings in earlier studies (12, 15). Witsenburg et al. (12), studying 1,826 cycles, found a mean number of 2.4 cycles per woman over a 4-year period. Linsten et al. (15), studying 4,928 couples, found that those treated over a 1-year period underwent an average of 1.8 cycles. We also found that the range of treatment cycles varied from 1 to 11 per woman, although only 21% continued beyond three cycles and <1% continued treatment past cycle 7 (Table 2). The live-birth rates per cycle confirm earlier reports, with declines paralleling advancing age with autologous but not donor oocytes (8, 16–19). Overall live-birth delivery rates per cycle are higher than rates per retrieval shown in a 1998 study of 4,043 SART clinic cycles (17). This is consistent with increases in success rates over time. Cumulative pregnancy and delivery rates for fertility treatments have been studied previously. In 1992, Hull et al. (7) reported on 1,280 cycles of fertility treatment in 804 couples, including IVF,
1337
TABLE 4 Live-birth rates per cycle by maternal age with autologous or donor oocytes. Maternal age (y) <35
35–37
38–40
41–42
R43
P value across age groups
Autologous cycle 1 Total cycles (n) Live births (n) Live-birth/cycle (%)
5,762 2,205 38.3
3,216 967 30.1
2,606 549 21.1
1,169 149 12.8
472 27 5.7
< .0001
Autologous cycle 2 Total cycles (n) Live births (n) Live-birth/cycle (%)
2,732 886 32.4
1,669 437 26.2
1,449 265 18.3
640 75 11.7
164 11 6.7
< .0001
Autologous cycle 3 Total cycles (n) Live births (n) Live-birth/cycle (%)
1,346 374 27.8
817 193 23.6
739 124 16.8
296 29 9.8
54 2 3.7
< .0001
Autologous cycle 4 Total cycles (n) Live births (n) Live-birth/cycle (%)
617 183 30.0
391 83 21.2
346 49 14.2
128 12 9.4
20 1 5.0
< .0001
Donor cycle 1 Total cycles (n) Live births (n) Live-birth/cycle (%)
158 74 46.8
172 75 43.6
297 137 46.1
268 122 45.5
536 216 40.3
.38
Donor cycle 2 Total cycles (n) Live births (n) Live-birth/cycle (%)
50 16 32.0
64 22 34.4
104 30 28.9
92 37 40.2
187 54 28.9
.36
Donor cycle 3 Total cycles (n) Live births (n) Live-birth/cycle (%)
17 5 29.4
24 7 29.2
42 10 23.8
26 6 23.1
73 16 21.9
.94
Donor cycle 4 Total cycles (n) Live births (n) Live-birth/cycle (%)
8 1 12.5
9 4 44.4
11 4 36.4
11 2 18.2
28 10 35.7
.51
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
gamete intrafallopian transfer, and intrauterine insemination. They found that cumulative rate of delivery for women aged <40 years was 70% after 6 cycles. This estimated rate was higher than that in some later studies (20–22). Tan et al. (20) reported cumulative live-birth rates between 29% and 59% after four cycles of treatment, with the variation dependent on the hormonal regime for ovarian hyperstimulation. Stolwijk et al. (10) estimated a probability of a cumulative live-birth rate of 54.4 10% after four cycles using autologous oocytes and sperm. Fukuda et al. (21) estimated cumulative delivery rates based on 2,339 cycles of conventional IVF-ET and found variation by age ranging from 50.8% in women aged %30 years to 7.7% in women aged R40 years. Most recently, Malizia et al. (22) reviewed 14,248 cycles in 6,164 patients and also showed cumulative live-birth rates that varied by age over six cycles They estimated these rates to be between 65% and 86% for women aged %35 years and demonstrated that the rate declined as women got older. Several studies have also looked at cumulative rates in women aged >40 years, finding these rates to be universally low (8, 11).
1338
Stern et al.
Linked ART cycles from SART CORS
Our estimates of cumulative live-birth rates are conservative; we have assumed that future treatment cycles would not be successful. In addition, because we did not know conclusively which women had treatment cycles before 2004, we assumed that the first cycle was the first one observed in our database that started in 2004. Because the first treatment cycles have higher success rates, this assumption also is conservative. Traditional analyses of cumulative pregnancy and delivery rates have used Kaplan-Meier analysis (23). Some investigators have questioned the utility of this statistical approach because it assumes that subjects dropping out of treatment would have responded similarly to those who continued to receive treatment and have advocated the use of alternate methods (9, 22, 22–26). Because there is currently no adequate method to estimate the success rate in women who declined further treatment, our approach of considering any woman who did not have a pregnancy with a live birth as a failure provides a conservative lower bound on the probability of treatment success. Vol. 94, No. 4, September 2010
FIGURE 2 Cumulative delivery rates per woman according to age of woman at first cycle and use of autologous versus donor oocytes. Cumulative delivery rates per woman from ART cycles in which women used their own (A) or donor (B) oocytes. Age of these patients at first cycle was: <35 years (red diamonds); 35–37 years (green squares); 38–40 years (green triangles); 41–42 years (blue circles); R43 years (purple circles).
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
Success rates for ART treatment options have evolved over time. It is therefore difficult to directly compare the present results with earlier studies that may have used cycles performed over many years and varied in cycle types. Some studies (15, 21) have limited their study population to women using fresh cycles, excluding those who went on to conceive using cryopreservation. Others have estimated the value of combining fresh and frozen embryo transfer cycles (27, 28). Analyses based on national data have demonstrated that live-birth rates per cycle using donor oocytes are consistently high and unchanged as a function of age (2). Witsenburg et al. (12) reported a 40% cumulative live-birth rate for donor cycles, but their study is limited by the small number of this type of cycle. In the present study, we chose to include all cycle types to achieve a more realistic measure of the outcomes seen in clinical practice.
Analysis of linked cycles has clear advantages for advancing our understanding of treatment course and outcome. Because this is a relatively new approach to evaluating ART data, and the data are complex and interrelated, the methodologies are still evolving. Factors in earlier cycles affecting outcomes in a future cycle should be weighed in any analysis. Studies of pregnancy and delivery rates could benefit from knowledge of predictive parameters from earlier treatment, including amount of medication used for ovarian stimulation, history of cycle cancellation, number of oocytes previously retrieved, and number of cryopreserved embryos. Studies of birth weight and gestational age could benefit from an understanding of the number of earlier cycles a patient has undergone, parity in earlier cycles, and previous birth outcomes. Linked cycle data could also be used to match ART patients and offspring to public records such as birth certificates and hospital discharge records. Future studies with this and other linked datasets should develop our ability to analyze these relationships and these parameters. This study is subject to several limitations. This is a retrospective study that depended on data entered from multiple sites into a clinical surveillance database system which was not designed for research purposes. Because we censored the data at two time periods, we could not determine which women had earlier and/or future cycles nor the characteristics and outcomes of those cycles. We were obligated to left-censor the data at 2004, because earlier SART datasets were no longer available for research. Restricting the data according to geographic region introduced an additional limitation, because some cycles, either of Massachusetts women who were treated out of state or out-of-state women treated at MA clinics, may have been missed. There is also the potential of having missed a small number of linkages between the few women whose date of birth and/or name were entered so incorrectly that our two-step linkage scheme was unsuccessful in correctly identifying them as a match. We also may have been missing cycles from patients who were Massachusetts residents but who were treated at out-of-state clinics that did not record zip codes. This limitation, inherent in linking a subset of the SART CORS database, will not be a problem for linkage of the entire database at some future date. The present results demonstrate that it is feasible to link cycles within the SART CORS dataset to determine the course of treatment of individual women and to generate cumulative treatment outcome information per woman. The use of the SART CORS database for analysis of cumulative live-birth rates has the potential of providing annual national estimates of ART success per woman, which would be very helpful to clinicians in counseling patients. Acknowledgments: The Society for Assisted Reproductive Technology thanks all of its members for providing clinical information to the SART CORS database for use by patients and researchers. Without the efforts of our members, this research would not have been possible.
REFERENCES 1. Andersen AN, Goossens V, Gianaroli L, Felberbaum R, de Mouzon J, Nygren KG, et al, Assisted reproductive technology in Europe, 2003. Results generated from European registers by ESHRE. Hum Reprod 2007;22:1513–25. 2. Department of Health and Human Services Centers for Disease Control and Prevention. ART success rates: national summary and fertility clinic report. 2006. Available at: http://www.cdc.gov/ART/ ART2006/index.htm. Last accessed March 3, 2009.
Fertility and Sterility
3. Adamson GD, de Mouzon J, Lancaster P, Nygren KG, Sullivan E, et al, International Committee for Monitoring Assisted Reproductive Technology. World collaborative report on in vitro fertilization, 2000. Fertil Steril 2006;85:1586–622. 4. Buck Louis GM, Schisterman EF, Dukic VM, Schieve LA. Research hurdles complicating the analysis of infertility treatment and child health. Hum Reprod 2005;20:12–8. 5. Elizur SE, Lerner-Geva L, Levron J, Shulman A, Bider D, Dor J, et al. Cumulative live birth rate fol-
lowing in vitro fertilization: study of 5,310 cycles. Gynecol Endocrinol 2006;22:25–30. 6. Engmann L, Maconochie N, Bekir JS, Jacobs HS, Tan SL. Cumulative probability of clinical pregnancy and live birth after a multiple cycle IVF package: a more realistic assessment of overall and agespecific success rates? Br J Obstet Gynaecol 1999;106:165–70. 7. Hull MG, Eddowes HA, Fahy U, Abuzeid MI, Mills MS, Cahill DJ, et al. Expectations of assisted conception for infertility. BMJ 1992;304:1465–9.
1339
8. Klipstein S, Regan M, Ryley DA, Goldman MB, Alper MM, Reindollar RH, et al. One last chance for pregnancy: a review of 2,705 in vitro fertilization cycles initiated in women age 40 years and above. Fertil Steril 2005;84:435–45. 9. Olivius K, Friden B, Lundin K, Bergh C, Olivius K, Friden B, et al. Cumulative probability of live birth after three in vitro fertilization/intracytoplasmic sperm injection cycles. Fertil Steril 2002;77:505–10. 10. Stolwijk AM, Wetzels AM, Braat DD. Cumulative probability of achieving an ongoing pregnancy after in-vitro fertilization and intracytoplasmic sperm injection according to a woman’s age, subfertility diagnosis and primary or secondary subfertility. Hum Reprod 2000;15:203–9. 11. van Disseldorp J, Eijkemans MJ, Klinkert ER, te Velde ER, Fauser BC, Broekmans FJ, et al. Cumulative live birth rates following IVF in 41- to 43-yearold women presenting with favourable ovarian reserve characteristics. Reprod Biomed Online 2007;14:455–63. 12. Witsenburg C, Dieben S, Van der Westerlaken L, Verburg H, Naaktgeboren N, Witsenburg C, et al. Cumulative live birth rates in cohorts of patients treated with in vitro fertilization or intracytoplasmic sperm injection. Fertil Steril 2005;84:99–107. 13. American Society for Reproductive Medicine: State Infertility Insurance Laws. Available at: http:// www.asrm.org/Patients/insur.html. Accessed February 26, 2009.
1340
Stern et al.
14. Wright VC, Chang J, Jeng G, Macaluso M. Assisted reproductive technology surveillance—United States 2005. Morb Mortal Wkly Rep Surveill Summ 2008;57:1–23. 15. Lintsen AM, Eijkemans MJ, Hunault CC, Bouwmans CA, Hakkaart L, Habbema JD, et al. Predicting ongoing pregnancy chances after IVF and ICSI: a national prospective study. Hum Reprod 2007;22:2455–62. 16. Lamarche C, Levy R, Felloni B, de Mouzon J, DenisBelicard E, Huss M, et al. Assisted reproductive techniques in women aged 38 years or more. Gynecol Obstet Fertil 2007;35:420–9. 17. Meldrum DR, Silverberg KM, Bustillo M, Stokes L. Success rate with repeated cycles of in vitro fertilization–embryo transfer. Fertil Steril 1998;69:1005–9. 18. Ron-El R, Raziel A, Strassburger D, Schachter M, Kasterstein E, Friedler S. Outcome of assisted reproductive technology in women over the age of 41. Fertil Steril 2000;74:471–5. 19. Spandorfer SD, Bendikson K, Dragisic K, Schattman G, Davis OK, Rosenwaks Z. Outcome of in vitro fertilization in women 45 years and older who use autologous oocytes. Fertil Steril 2007;87:74–6. 20. Tan SL, Maconochie N, Doyle P, Campbell S, Balen A, Bekir J, et al. Cumulative conception and live-birth rates after in vitro fertilization with and without the use of long, short, and ultrashort regimens of the gonadotropin-releasing hormone agonist buserelin. Am J Obstet Gynecol 1994;171:513–20. 21. Fukuda J, Kumagai J, Kodama H, Murata M, Kawamura K, Tanaka T. Upper limit of the number
Linked ART cycles from SART CORS
22.
23. 24.
25.
26.
27.
28.
of IVF-ET treatment cycles in different age groups, predicted by cumulative take-home baby rate. Acta Obstet Gynecol Scand 2001;80:71–3. Malizia B, Hacker M, Penzias A. Cumulative livebirth rates after in vitro fertilization. N Engl J Med 2009;360:236–43. Bland JM, Altman DG. Survival probabilities (the Kaplan-Meier method). BMJ 1998;317:1572. Daya S. Life table (survival) analysis to generate cumulative pregnancy rates in assisted reproduction: are we overestimating our success rates? Hum Reprod 2005;20:1135–43. Roest J, van Heusden AM, Zeilmaker GH, Verhoeff A. Cumulative pregnancy rates and selective drop-out of patients in in-vitro fertilization treatment. Hum Reprod 1998;13:339–41. Stolwijk AM, Hamilton CJ, Hollanders JM, Bastiaans LA, Zielhuis GA. A more realistic approach to the cumulative pregnancy rate after in-vitro fertilization. Hum Reprod 1996;11:660–3. de Jong D, Eijkemans MJ, Beckers NG, Pruijsten RV, Fauser BC, Macklon NS, et al. The added value of embryo cryopreservation to cumulative ongoing pregnancy rates per IVF treatment: is cryopreservation worth the effort? J Assist Reprod Genet 2002;19:561–8. Ubaldi F, Rienzi L, Baroni E, Ferrero S, Iacobelli M, Minasi MG, et al. Cumulative pregnancy rates after transfer of fresh and thawed embryos. Eur J Obstet Gynecol Reprod Biol 2004;115 (Suppl 1): S106–9.
Vol. 94, No. 4, September 2010
Objective: To determine the feasibility of linking assisted reproductive technology (ART) cycles for individual women to compare per-cycle and cumulative live-birth rates. Design: Historical cohort study. Setting: Clinic-based data. Patient(s): A total of 27,906 ART cycles with residency or treatment in Massachusetts during 2004–2006 and reported to the Society for Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS) on-line database. Intervention(s): None. Main Outcome Measure(s): Per-cycle and cumulative live-birth rates. Result(s): Linkage of cycles up to and including the first live-birth delivery revealed 14,265 women who averaged 1.9 1.2 SD cycles (range 1–11). These cycles yielded 9,452 pregnancies resulting in 7,675 live-birth deliveries. From cycle 1 to cycle 4, the cumulative live-birth rate for all patients increased from 30.4% to 43.3%, 49.1%, and 51.9%, respectively, and plateaued thereafter at about 53%. The cumulative live-birth rate after three cycles using donor oocytes was 60% for women aged <43 years and >50% for women R43 years; for autologous oocytes it was 60.1% for ages <35 years and declined steadily to 8.5% for ages R43 years. Conclusion(s): The results demonstrate the feasibility of linking ART cycles for individual women from SART CORS to characterize cumulative live-birth rates. (Fertil Steril 2010;94:1334–40. 2010 by American Society for Reproductive Medicine.) Key Words: Cumulative live-birth rate, live-birth rate per cycle, assisted reproductive technology, SART CORS
Success rates for assisted reproductive technology (ART) cycles have traditionally been reported as pregnancies per cycle according to maternal age. This cross-sectional measure, which can be easily calculated, is used by national ART registries in Europe, the Middle East, Australia and New Zealand, and North America, including the United States (1–3). This measure has limited usefulness, though, for the woman who undergoes multiple cycles. Longitudinal analysis linking all treatment cycles for an individual woman would have short- and long-term advantages. The immediate advantage is a more accurate estimate of success with continued
Received March 13, 2009; revised May 17, 2009; accepted May 28, 2009; published online July 10, 2009. J.S. has nothing to disclose. M.B. has nothing to disclose. B.L. has nothing to disclose. E.W. has nothing to disclose. A.L. has nothing to disclose. S.M. has nothing to disclose. M.H. has nothing to disclose. Supported by the Society for Assisted Reproductive Technology. Presented at the 64th annual meeting of the American Society for Reproductive Medicine, San Francisco, California, November 8–12, 2008. Reprint requests: Judy E. Stern, Ph.D., Department of Obstetrics and Gynecology, Dartmouth-Hitchcock Medical Center, Lebanon, NH 03756 (FAX: 603-650-0905; E-mail: [email protected]).
1334
treatment. Linking cycles would also facilitate assessing a woman’s future health risks from exposure to multiple treatment cycles over time. Concerns about the inadequacy of research on ART outcomes as a result of per-cycle reporting and other problems have been discussed previously (4). Studies that have linked treatment cycles have been limited by small sample size, inconsistent inclusion criteria and outcome measures, and statistical and methodologic shortcomings (5–12). The present project was undertaken to link cycles of ART from the national Society for Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS) on-line database. We evaluated a subset of cycles reported to this database from a single state to: 1) determine the feasibility of performing the linkage; and 2) compare the traditional measure of success of live-birth rates per cycle with cumulative live-birth rates per woman linking her cycles of treatment. We chose Massachusetts to demonstrate this linkage for several reasons. Massachusetts is one of only 14 states with mandated private insurance coverage for infertility therapy and one of only five states with full coverage, reducing an important barrier to care (13). As a result, in 2005 Massachusetts ranked fifth in highest number of ART cycles and ART births in the United States (14). In addition, all of the seven ART programs in
Fertility and Sterility Vol. 94, No. 4, September 2010 Copyright ª2010 American Society for Reproductive Medicine, Published by Elsevier Inc.
0015-0282/$36.00 doi:10.1016/j.fertnstert.2009.05.052
FIGURE 1 Study population and exclusions. SART CORS ¼ Society for Assisted Reproductive Technology Clinic Outcomes Reporting System.
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
Massachusetts report their data directly to the national SART CORS database.
MATERIALS AND METHODS The data source for this study was the SART CORS database, which contains comprehensive data from >90% of all reporting clinics performing ART in the U.S. Data were collected and verified by SART and reported to the Centers for Disease Control and Prevention in compliance with the Fertility Clinic Success Rate and Certification Act of 1992 (Public Law 102-493). The study was approved by the Committee for the Protection of Human Subjects at Dartmouth College, and all identifiable data were analyzed at the offices of the SART CORS programmers. Patient-identifiable data, such as names, dates, social security numbers, addresses, and zip codes, did not leave that location, and the datasets for subsequent analyses were stripped of identifiers, retaining only study numbers which were randomly assigned to link cycles. In this study we used data reported between January 1, 2004, and December 31, 2006, from patients who were either treated at Massachusetts ART centers or who had a Massachusetts zip code Fertility and Sterility
recorded at the time of treatment. This database includes in vitro fertilization with and without intracytoplasmic sperm injection (ICSI), frozen embryo transfer cycles, and cycles of other procedures, such as gamete intrafallopian transfer and preimplantation genetic diagnosis (PGD), that involve removal of oocytes from the woman, but it does not include intrauterine insemination or ovulation induction cycles. All cycle types (fresh, frozen, autologous egg, donor egg, donor sperm, PGD, banking, and research cycles) were included in the linkage analysis. The scheme to link cycles used identifying variables available in the SART CORS database, including date of birth, first and last names, social security number, and zip code. First and last names were adjusted to remove all nonletter characters and were coded using Soundex software (Soundex SQL Server 2000) to facilitate phonetic matches. Following Soundex coding, the first linkage used date of birth, last name, and first name to generate an initial random SART-Woman-ID. Because all cycles within a single clinic were generated from a single patient file, this linkage was highly accurate. In addition, an initial set of intraclinic linkages were revealed. A second linkage took these cycles at separate clinics and reordered them according to date of birth, first name, and last name to facilitate
1335
TABLE 1 Characteristics of the study population. Cycles
n
% of cycles
Total cyclesa Cycles at MA clinics Out-of-state cycles on MA residents Autologous oocyte cycles Donor oocyte cycles Fresh cycles Frozen embryo transfer cycles Use of ICSI in fresh cycles
27,906 26,778 1,128 25,695 2,211 23,959 3,947 9,914
100.0 96.0 4.0 92.1 7.9 85.9 14.1 41.4 (% fresh)
Total no. of women Pregnancies Ectopic pregnancies Fetal losses Stillbirths Live-birth deliveries
14,265 9,452 64 1,685 28 7,675
33.8 0.2 6.0 0.1 27.5
% of women
% of pregnancies
100.0 66.2 0.4 11.8 0.2 53.8
100.0 0.7 17.8 0.3 81.2
Note: ICSI ¼ intracytoplasmic sperm injection; MA ¼ Massachusetts. a Includes cycles performed at MA clinics plus all additional cycles to women with MA zip codes. Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
linkages between cycles in which the date of birth and first name were the same or similar but in which the last name was recorded differently or in modified format at subsequent clinics. Social security number and zip code (where present) as well as cycle order were used to confirm or refute the proposed matches from this second linkage. Social security number was found in only 47% of cycles and zip code in 67%. This, and the fact that some clinics had no entries for these identifiers on any cycles, made these of limited use for matching between clinics. Potential linkages were considered for hyphenated versus nonhyphenated names, first name and middle name combinations, and data entry errors, such as use of numbers or nonsensical characters in names. As a final measure for identifying accurate linkages among women with the same identifying variables, the outcome of a prior cycle and timing of a subsequent cycle were evaluated (i.e., a live birth in December 2005 followed by
TABLE 2 Assisted reproductive technology cycles per woman in Massachusetts cohort.
No. of cycles per woman 1 2 3 4 5 6 7 8 9 10 11 Total
No. of women with specified number of cycles
% of women
7,113 3,602 1,865 933 436 198 71 32 10 4 1 14,265
49.86 25.25 13.07 6.54 3.06 1.39 0.50 0.22 0.07 0.03 0.01 100.00
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
1336
Stern et al.
Linked ART cycles from SART CORS
a treatment cycle in January 2006 would indicate two different women, not linked cycles for the same woman). Fewer than 0.1% of cases suggested a match for which supporting information did not clearly permit a decision that the two cycles were indeed the same patient. Since we could not validate the data records against original clinical records, there may be errors due to data entry of names and/or dates of birth both when linking records within a clinic and when linking records between clinics. However, it is unlikely that the error rate due to these types of errors exceeds 2%–3%. After cycle matching, the data file was limited to exclude certain patients with specific types of cycles, as shown in Figure 1. The file was processed to delete banking cycles in which all embryos were cryopreserved and none were transfered (none of these patients had subsequent frozen embryo transfers). Also deleted were five cycles with unspecified treatment outcomes. Cycles were limited to those up to and including the first live birth delivery, i.e., all cycles were included for women in whom a live birth delivery did not occur, but cycles after the first delivery were excluded. Because women with or without frozen embryo transfer cycles may also have had ART cycles before 2004, we chose to include all women with ART cycles in the analysis. The data were analyzed either by total number of cycles, by cycle number, or by woman. Cumulative delivery rates represent deliveries per woman after repeated cycles of ART. These rates are not equivalent to the total deliveries from a single oocyte harvest (previously defined as total reproductive potential). Live-birth rates per cycle were computed for age groups <35, 35–37, 38–40, 41–42, and R43 years at the start of their first cycle with autologous versus donor oocytes. Because use of donor oocytes represents an entirely new phase of treatment, we renumbered the cycles with donor oocytes to start with 1. Both fresh and frozen embryo transfers were included. Differences in live-birth rates across the five age groups were analyzed using the c2 statistic. SAS 9.1 (SAS, Cary, NC) was used for all analyses.
RESULTS The cohort included 27,906 cycles of ART among 14,265 women resulting in 9,452 pregnancies with 7,675 live-birth deliveries. The Vol. 94, No. 4, September 2010
TABLE 3 Cumulative live-birth rates per woman.
Cycle of treatment 1 2 3 4 5 6 7 8 9–11
Live births (n)
Total cycles at each level (n)
Live birth/cycle (%)
4,331 1,848 825 396 169 68 25 10 3
14,265 7,152 3,550 1,685 752 316 118 47 21
30.4 25.8 23.2 23.5 22.5 21.5 21.2 21.3 14.3
95% CI
Cumulative live births (n)
Total women with ART (n)
Cumulative live birth/woman (%)
95% CI
29.6–31.1 24.8–26.9 21.9–24.6 21.5–25.5 19.5–25.5 17.0–26.1 13.8–28.6 9.6–33.0 0.0–29.3
4,331 6,179 7,004 7,400 7,569 7,637 7,662 7,672 7,675
14,265 14,265 14,265 14,265 14,265 14,265 14,265 14,265 14,265
30.4 43.3 49.1 51.9 53.1 53.5 53.7 53.8 53.8
29.7–31.1 42.5–44.1 48.3–49.9 51.1–52.7 52.3–53.9 52.6–54.4 52.9–54.5 53.0–54.6 53.0–54.6
Note: ART ¼ assisted reproductive technology; CI ¼ confidence interval. Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
characteristics of these cycles are presented in Table 1 and include 92.1% cycles with autologous oocytes and 14.1% cycles with frozen embryo transfers. Intracytoplasmic sperm injection was used to achieve fertilization in 41.4% of fresh cycles. The number of cycles for each woman is shown in Table 2. Women underwent between 1 and 11 cycles during this 3-year period, with a mean of 1.9 1.2 SD and a median of 2 cycles. Of the 14,265 women, 21% (3,000) had at least one frozen cycle and 10% (1,431) had at least one donor oocyte cycle. There were 7,113 women who underwent only one cycle, of whom 649 had frozen embryo transfers, indicating that this group included women with cycles before 2004. Table 3 shows the per-cycle and cumulative live-birth rate per woman as a function of cycle of treatment. The cumulative live-birth rate for all women was 53.8%. The live-birth rates by age group and autologous versus donor status are shown in Table 4. Only the first four cycles of treatment are shown, because the data for subsequent cycles (five or more) were too sparse to provide robust estimates. The live-birth rates per cycle with autologous oocytes declined significantly with age in all cycles, from 38.3% to 5.7% from the youngest to the oldest ages for the first cycle (P<.0001) and from 30.0% to 5.0% for the fourth cycle (P<.0001). In contrast, the live-birth rates with donor oocytes were not significantly different across age groups in all cycles, ranging from 46.8% to 40.3% from the youngest to the oldest ages for the first cycle (P¼.38) and from 29.4% to 21.9% for the third cycle (P¼.94). The cumulative live-birth rates per woman are shown in Figure 2. In line with the live-birth rates per cycle, the cumulative live-birth rates with autologous eggs differed substantially with advancing age. The cumulative live-birth rates among women with four cycles using autologous oocytes ranged from >60% for women aged <35 years to 8.7% among women aged R43 years. With donor oocytes, the cumulative live-birth rates among women with four cycles averaged about 60% for women aged %42 years and >50% for women aged R43 years.
DISCUSSION This is the first study to link successive cycles of treatment from the SART CORS database. We have used these linked cycles to determine the number of women, the number of cycles, and the outcomes for cycles performed at Massachusetts ART centers or for women with Massachusetts residency between January 2004 and December Fertility and Sterility
2006. Because we began to follow women at an arbitrary time point, the dataset includes women who may have had earlier cycles. Therefore, all estimates of live-birth rates are underestimated, because early cycles have the highest success rates. Censoring the data within this 36-month period also resulted in including women whose treatment was not finished by December 31, 2006. Most studies of ART have analyzed outcomes on a per-cycle basis; however, the more meaningful outcome measure is the outcome per woman. Per-cycle rates do not accurately measure the value of continued treatment for individual patients, nor do they allow determination of the effect of treatment on a woman’s long-term health. Because of the importance of determining these outcome measures, we performed the present analysis as the firstever attempt to assess the feasibility of using the SART CORS dataset to link ART cycles. We limited the analysis to patients in one geographic area, in a state with a large patient volume, with comprehensive reporting of all clinics to SART CORS, and with mandatory insurance coverage of ART services. Between-clinic linkages were possible but more problematic due to inconsistencies in entries of names and under-reporting of social security numbers and zip codes. The present study is the largest of its kind to date, based on 27,906 cycles among 14,265 women. We showed that over a 3-year period, patients underwent an average of 1.9 1.2 SD cycles of ART, which is in accordance with findings in earlier studies (12, 15). Witsenburg et al. (12), studying 1,826 cycles, found a mean number of 2.4 cycles per woman over a 4-year period. Linsten et al. (15), studying 4,928 couples, found that those treated over a 1-year period underwent an average of 1.8 cycles. We also found that the range of treatment cycles varied from 1 to 11 per woman, although only 21% continued beyond three cycles and <1% continued treatment past cycle 7 (Table 2). The live-birth rates per cycle confirm earlier reports, with declines paralleling advancing age with autologous but not donor oocytes (8, 16–19). Overall live-birth delivery rates per cycle are higher than rates per retrieval shown in a 1998 study of 4,043 SART clinic cycles (17). This is consistent with increases in success rates over time. Cumulative pregnancy and delivery rates for fertility treatments have been studied previously. In 1992, Hull et al. (7) reported on 1,280 cycles of fertility treatment in 804 couples, including IVF,
1337
TABLE 4 Live-birth rates per cycle by maternal age with autologous or donor oocytes. Maternal age (y) <35
35–37
38–40
41–42
R43
P value across age groups
Autologous cycle 1 Total cycles (n) Live births (n) Live-birth/cycle (%)
5,762 2,205 38.3
3,216 967 30.1
2,606 549 21.1
1,169 149 12.8
472 27 5.7
< .0001
Autologous cycle 2 Total cycles (n) Live births (n) Live-birth/cycle (%)
2,732 886 32.4
1,669 437 26.2
1,449 265 18.3
640 75 11.7
164 11 6.7
< .0001
Autologous cycle 3 Total cycles (n) Live births (n) Live-birth/cycle (%)
1,346 374 27.8
817 193 23.6
739 124 16.8
296 29 9.8
54 2 3.7
< .0001
Autologous cycle 4 Total cycles (n) Live births (n) Live-birth/cycle (%)
617 183 30.0
391 83 21.2
346 49 14.2
128 12 9.4
20 1 5.0
< .0001
Donor cycle 1 Total cycles (n) Live births (n) Live-birth/cycle (%)
158 74 46.8
172 75 43.6
297 137 46.1
268 122 45.5
536 216 40.3
.38
Donor cycle 2 Total cycles (n) Live births (n) Live-birth/cycle (%)
50 16 32.0
64 22 34.4
104 30 28.9
92 37 40.2
187 54 28.9
.36
Donor cycle 3 Total cycles (n) Live births (n) Live-birth/cycle (%)
17 5 29.4
24 7 29.2
42 10 23.8
26 6 23.1
73 16 21.9
.94
Donor cycle 4 Total cycles (n) Live births (n) Live-birth/cycle (%)
8 1 12.5
9 4 44.4
11 4 36.4
11 2 18.2
28 10 35.7
.51
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
gamete intrafallopian transfer, and intrauterine insemination. They found that cumulative rate of delivery for women aged <40 years was 70% after 6 cycles. This estimated rate was higher than that in some later studies (20–22). Tan et al. (20) reported cumulative live-birth rates between 29% and 59% after four cycles of treatment, with the variation dependent on the hormonal regime for ovarian hyperstimulation. Stolwijk et al. (10) estimated a probability of a cumulative live-birth rate of 54.4 10% after four cycles using autologous oocytes and sperm. Fukuda et al. (21) estimated cumulative delivery rates based on 2,339 cycles of conventional IVF-ET and found variation by age ranging from 50.8% in women aged %30 years to 7.7% in women aged R40 years. Most recently, Malizia et al. (22) reviewed 14,248 cycles in 6,164 patients and also showed cumulative live-birth rates that varied by age over six cycles They estimated these rates to be between 65% and 86% for women aged %35 years and demonstrated that the rate declined as women got older. Several studies have also looked at cumulative rates in women aged >40 years, finding these rates to be universally low (8, 11).
1338
Stern et al.
Linked ART cycles from SART CORS
Our estimates of cumulative live-birth rates are conservative; we have assumed that future treatment cycles would not be successful. In addition, because we did not know conclusively which women had treatment cycles before 2004, we assumed that the first cycle was the first one observed in our database that started in 2004. Because the first treatment cycles have higher success rates, this assumption also is conservative. Traditional analyses of cumulative pregnancy and delivery rates have used Kaplan-Meier analysis (23). Some investigators have questioned the utility of this statistical approach because it assumes that subjects dropping out of treatment would have responded similarly to those who continued to receive treatment and have advocated the use of alternate methods (9, 22, 22–26). Because there is currently no adequate method to estimate the success rate in women who declined further treatment, our approach of considering any woman who did not have a pregnancy with a live birth as a failure provides a conservative lower bound on the probability of treatment success. Vol. 94, No. 4, September 2010
FIGURE 2 Cumulative delivery rates per woman according to age of woman at first cycle and use of autologous versus donor oocytes. Cumulative delivery rates per woman from ART cycles in which women used their own (A) or donor (B) oocytes. Age of these patients at first cycle was: <35 years (red diamonds); 35–37 years (green squares); 38–40 years (green triangles); 41–42 years (blue circles); R43 years (purple circles).
Stern. Linked ART cycles from SART CORS. Fertil Steril 2010.
Success rates for ART treatment options have evolved over time. It is therefore difficult to directly compare the present results with earlier studies that may have used cycles performed over many years and varied in cycle types. Some studies (15, 21) have limited their study population to women using fresh cycles, excluding those who went on to conceive using cryopreservation. Others have estimated the value of combining fresh and frozen embryo transfer cycles (27, 28). Analyses based on national data have demonstrated that live-birth rates per cycle using donor oocytes are consistently high and unchanged as a function of age (2). Witsenburg et al. (12) reported a 40% cumulative live-birth rate for donor cycles, but their study is limited by the small number of this type of cycle. In the present study, we chose to include all cycle types to achieve a more realistic measure of the outcomes seen in clinical practice.
Analysis of linked cycles has clear advantages for advancing our understanding of treatment course and outcome. Because this is a relatively new approach to evaluating ART data, and the data are complex and interrelated, the methodologies are still evolving. Factors in earlier cycles affecting outcomes in a future cycle should be weighed in any analysis. Studies of pregnancy and delivery rates could benefit from knowledge of predictive parameters from earlier treatment, including amount of medication used for ovarian stimulation, history of cycle cancellation, number of oocytes previously retrieved, and number of cryopreserved embryos. Studies of birth weight and gestational age could benefit from an understanding of the number of earlier cycles a patient has undergone, parity in earlier cycles, and previous birth outcomes. Linked cycle data could also be used to match ART patients and offspring to public records such as birth certificates and hospital discharge records. Future studies with this and other linked datasets should develop our ability to analyze these relationships and these parameters. This study is subject to several limitations. This is a retrospective study that depended on data entered from multiple sites into a clinical surveillance database system which was not designed for research purposes. Because we censored the data at two time periods, we could not determine which women had earlier and/or future cycles nor the characteristics and outcomes of those cycles. We were obligated to left-censor the data at 2004, because earlier SART datasets were no longer available for research. Restricting the data according to geographic region introduced an additional limitation, because some cycles, either of Massachusetts women who were treated out of state or out-of-state women treated at MA clinics, may have been missed. There is also the potential of having missed a small number of linkages between the few women whose date of birth and/or name were entered so incorrectly that our two-step linkage scheme was unsuccessful in correctly identifying them as a match. We also may have been missing cycles from patients who were Massachusetts residents but who were treated at out-of-state clinics that did not record zip codes. This limitation, inherent in linking a subset of the SART CORS database, will not be a problem for linkage of the entire database at some future date. The present results demonstrate that it is feasible to link cycles within the SART CORS dataset to determine the course of treatment of individual women and to generate cumulative treatment outcome information per woman. The use of the SART CORS database for analysis of cumulative live-birth rates has the potential of providing annual national estimates of ART success per woman, which would be very helpful to clinicians in counseling patients. Acknowledgments: The Society for Assisted Reproductive Technology thanks all of its members for providing clinical information to the SART CORS database for use by patients and researchers. Without the efforts of our members, this research would not have been possible.
REFERENCES 1. Andersen AN, Goossens V, Gianaroli L, Felberbaum R, de Mouzon J, Nygren KG, et al, Assisted reproductive technology in Europe, 2003. Results generated from European registers by ESHRE. Hum Reprod 2007;22:1513–25. 2. Department of Health and Human Services Centers for Disease Control and Prevention. ART success rates: national summary and fertility clinic report. 2006. Available at: http://www.cdc.gov/ART/ ART2006/index.htm. Last accessed March 3, 2009.
Fertility and Sterility
3. Adamson GD, de Mouzon J, Lancaster P, Nygren KG, Sullivan E, et al, International Committee for Monitoring Assisted Reproductive Technology. World collaborative report on in vitro fertilization, 2000. Fertil Steril 2006;85:1586–622. 4. Buck Louis GM, Schisterman EF, Dukic VM, Schieve LA. Research hurdles complicating the analysis of infertility treatment and child health. Hum Reprod 2005;20:12–8. 5. Elizur SE, Lerner-Geva L, Levron J, Shulman A, Bider D, Dor J, et al. Cumulative live birth rate fol-
lowing in vitro fertilization: study of 5,310 cycles. Gynecol Endocrinol 2006;22:25–30. 6. Engmann L, Maconochie N, Bekir JS, Jacobs HS, Tan SL. Cumulative probability of clinical pregnancy and live birth after a multiple cycle IVF package: a more realistic assessment of overall and agespecific success rates? Br J Obstet Gynaecol 1999;106:165–70. 7. Hull MG, Eddowes HA, Fahy U, Abuzeid MI, Mills MS, Cahill DJ, et al. Expectations of assisted conception for infertility. BMJ 1992;304:1465–9.
1339
8. Klipstein S, Regan M, Ryley DA, Goldman MB, Alper MM, Reindollar RH, et al. One last chance for pregnancy: a review of 2,705 in vitro fertilization cycles initiated in women age 40 years and above. Fertil Steril 2005;84:435–45. 9. Olivius K, Friden B, Lundin K, Bergh C, Olivius K, Friden B, et al. Cumulative probability of live birth after three in vitro fertilization/intracytoplasmic sperm injection cycles. Fertil Steril 2002;77:505–10. 10. Stolwijk AM, Wetzels AM, Braat DD. Cumulative probability of achieving an ongoing pregnancy after in-vitro fertilization and intracytoplasmic sperm injection according to a woman’s age, subfertility diagnosis and primary or secondary subfertility. Hum Reprod 2000;15:203–9. 11. van Disseldorp J, Eijkemans MJ, Klinkert ER, te Velde ER, Fauser BC, Broekmans FJ, et al. Cumulative live birth rates following IVF in 41- to 43-yearold women presenting with favourable ovarian reserve characteristics. Reprod Biomed Online 2007;14:455–63. 12. Witsenburg C, Dieben S, Van der Westerlaken L, Verburg H, Naaktgeboren N, Witsenburg C, et al. Cumulative live birth rates in cohorts of patients treated with in vitro fertilization or intracytoplasmic sperm injection. Fertil Steril 2005;84:99–107. 13. American Society for Reproductive Medicine: State Infertility Insurance Laws. Available at: http:// www.asrm.org/Patients/insur.html. Accessed February 26, 2009.
1340
Stern et al.
14. Wright VC, Chang J, Jeng G, Macaluso M. Assisted reproductive technology surveillance—United States 2005. Morb Mortal Wkly Rep Surveill Summ 2008;57:1–23. 15. Lintsen AM, Eijkemans MJ, Hunault CC, Bouwmans CA, Hakkaart L, Habbema JD, et al. Predicting ongoing pregnancy chances after IVF and ICSI: a national prospective study. Hum Reprod 2007;22:2455–62. 16. Lamarche C, Levy R, Felloni B, de Mouzon J, DenisBelicard E, Huss M, et al. Assisted reproductive techniques in women aged 38 years or more. Gynecol Obstet Fertil 2007;35:420–9. 17. Meldrum DR, Silverberg KM, Bustillo M, Stokes L. Success rate with repeated cycles of in vitro fertilization–embryo transfer. Fertil Steril 1998;69:1005–9. 18. Ron-El R, Raziel A, Strassburger D, Schachter M, Kasterstein E, Friedler S. Outcome of assisted reproductive technology in women over the age of 41. Fertil Steril 2000;74:471–5. 19. Spandorfer SD, Bendikson K, Dragisic K, Schattman G, Davis OK, Rosenwaks Z. Outcome of in vitro fertilization in women 45 years and older who use autologous oocytes. Fertil Steril 2007;87:74–6. 20. Tan SL, Maconochie N, Doyle P, Campbell S, Balen A, Bekir J, et al. Cumulative conception and live-birth rates after in vitro fertilization with and without the use of long, short, and ultrashort regimens of the gonadotropin-releasing hormone agonist buserelin. Am J Obstet Gynecol 1994;171:513–20. 21. Fukuda J, Kumagai J, Kodama H, Murata M, Kawamura K, Tanaka T. Upper limit of the number
Linked ART cycles from SART CORS
22.
23. 24.
25.
26.
27.
28.
of IVF-ET treatment cycles in different age groups, predicted by cumulative take-home baby rate. Acta Obstet Gynecol Scand 2001;80:71–3. Malizia B, Hacker M, Penzias A. Cumulative livebirth rates after in vitro fertilization. N Engl J Med 2009;360:236–43. Bland JM, Altman DG. Survival probabilities (the Kaplan-Meier method). BMJ 1998;317:1572. Daya S. Life table (survival) analysis to generate cumulative pregnancy rates in assisted reproduction: are we overestimating our success rates? Hum Reprod 2005;20:1135–43. Roest J, van Heusden AM, Zeilmaker GH, Verhoeff A. Cumulative pregnancy rates and selective drop-out of patients in in-vitro fertilization treatment. Hum Reprod 1998;13:339–41. Stolwijk AM, Hamilton CJ, Hollanders JM, Bastiaans LA, Zielhuis GA. A more realistic approach to the cumulative pregnancy rate after in-vitro fertilization. Hum Reprod 1996;11:660–3. de Jong D, Eijkemans MJ, Beckers NG, Pruijsten RV, Fauser BC, Macklon NS, et al. The added value of embryo cryopreservation to cumulative ongoing pregnancy rates per IVF treatment: is cryopreservation worth the effort? J Assist Reprod Genet 2002;19:561–8. Ubaldi F, Rienzi L, Baroni E, Ferrero S, Iacobelli M, Minasi MG, et al. Cumulative pregnancy rates after transfer of fresh and thawed embryos. Eur J Obstet Gynecol Reprod Biol 2004;115 (Suppl 1): S106–9.
Vol. 94, No. 4, September 2010