Data used in this study were obtained from National Assisted Reproductive Technology Surveillance System, which collects information about assisted reproductive technology cycles performed at United States fertility assisted reproductive technology clinics, as mandated by law (Fertility Clinic Success Rate and Certification Act of 1992, Public Law No. 102-493, October 24, 1992). Assisted reproductive technology procedures include those involving the laboratory handling of gametes, namely in vitro fertilization (IVF) transcervical embryo transfer, gamete intrafallopian transfer, and zygote intrafallopian transfer. National Assisted Reproductive Technology Surveillance System data include patient demographics, patient obstetrical and medical history, parental infertility diagnosis, clinical parameters of the assisted reproductive technology procedure, and information regarding resultant pregnancies and births. Approximately 5–12% of assisted reproductive technology clinics did not report data to Centers for Disease Control and Prevention during 2000–2008. Because most nonreporting clinics are small, we estimate that National Assisted Reproductive Technology Surveillance System contains information about more than 95% of all assisted reproductive technology cycles performed in the United States.

In vitro fertilization transcervical embryo transfer can be categorized further into the following groups: 1) fresh, donor; 2) fresh, nondonor; 3) frozen, donor; and 4) frozen, nondonor. Cycles involving fresh embryos are defined as those in which embryos were transferred after embryo culture. Cycles involving frozen embryos are defined as those in which thawed, previously frozen embryos were transferred. Donor embryo transfer refers to the transfer of donor embryos, whereas nondonor embryo transfer refers to the transfer of the patient’s own embryos.

We used data on all live births resulting from assisted reproductive technology cycles performed between 2000 and 2008 for the analysis of trends in good perinatal outcomes among all assisted reproductive technology live births (n=444,909) and among the subset of assisted reproductive technology liveborn singletons (n=222,500). Because we wanted to estimate the effects of factors other than multiple births on perinatal outcomes, we limited our analyses to singleton live births resulting from fresh, nondonor IVF transcervical embryo transfer cycles performed in 2008 (n=21,451), thereby ensuring homogeneity of the study population. We excluded 483 neonates for whom data were unavailable on birthweight or gestational age and 188 neonates born to gestational carriers. Our final study population for this analysis consisted of 20,780 neonates.

We defined a “good perinatal outcome” to be the live birth of a singleton neonate born at term (37 or more completed weeks of gestation) weighing 2,500 g or more. For fresh embryo transfers, we calculated gestational age by subtracting the date of oocyte retrieval from the date of birth and adding 14 days to adjust for theoretical last menstrual period. For assisted reproductive technology cycles involving transfer of previously frozen embryos, and in cases in which the date of oocyte retrieval was missing, we calculated gestational age by subtracting the date of embryo transfer from the date of birth and adding 17 days to adjust for theoretical last menstrual period and days in embryo culture.

All statistical analyses were performed using SAS statistical software 9.2. We estimated trends in proportion of good perinatal outcomes among all assisted reproductive technology live births and assisted reproductive technology singleton live births from 2000 to 2008 using χ² test for trend among the following cycle groups: fresh, nondonor; fresh, donor; frozen, nondonor; and frozen, donor. For each of these groups, we calculated the relative change in proportion of good perinatal outcomes from 2000 to 2008 and computed a χ² test for trend.

We examined the distribution of maternal and treatment characteristics for liveborn singletons and for all liveborn neonates after fresh, nondonor assisted reproductive technology cycles initiated in 2008. For singletons, univariable analyses were conducted to evaluate associations with good perinatal outcome with the χ² test for differences in proportions for the following characteristics: maternal age; gravidity; previous births; race or ethnicity; parental infertility diagnosis; use of intracytoplasmic sperm injection; use of assisted hatching; availability of supernumerary embryos for freezing; number of embryos transferred; number of fetal hearts on 6-week ultrasound examination; number of days of embryo culture; history of spontaneous abortion; and history of assisted reproductive technology cycles. Race or ethnicity data were missing for more than 30% of all assisted reproductive technology births. To allow use of data for these births, we created a category of “unknown” for those with missing data for the race or ethnicity variable. Because patients may have had more than one infertility diagnosis, the presence or absence of each diagnosis was handled as an individual variable.

Factors that were determined to be significant based on univariable analysis (P<.05) were included in a logistic regression model to describe predictors of good perinatal outcome. We used stepwise multiple logistic regression and included only those variables that had a two-tailed P<.05 in the final model. We calculated the adjusted odds ratios (ORs) and accompanying 95% confidence intervals (CIs) for each term. We also calculated a χ² statistic for goodness-of-fit on the final model, which indicated an adequate fit.

We conducted sensitivity analyses using multiple logistic regression in the manner described, limiting the original sample of 2008 liveborn singletons after fresh, nondonor assisted reproductive technology to those with: 1) one fetal heart on 6-week ultrasound examination (n=17,289); 2) one embryo transferred (n=2,128); and 3) nonmissing data (n=13,613) for the race or ethnicity variable. Additionally, we repeated univariable analysis for tubal factor infertility, limiting this group to those solely with tubal factor infertility diagnosed (n=3,376) compared with all others (n=17,404).

Last, to assess the validity of our results among the population of assisted reproductive technology live births, we conducted separate analyses to assess associations with good outcomes among the population of all live births after 2008 fresh, nondonor assisted reproductive technology cycles (n=40,288). We used univariable analyses for the factors described and defined good perinatal outcomes as live births at 37 completed weeks or more, with birth-weights of at least 2,500 g. In addition to these factors, we tested singleton compared with nonsingleton live birth for association with the outcome. Factors determined to be significant were included in a stepwise logistic regression model controlling for clustering effects of mothers with multiple birth outcomes by incorporating a random effects model. This study was approved by the Institutional Review Board of the Centers for Disease Control and Prevention.

The total number of liveborn neonates conceived after assisted reproductive technology increased 76%, from 34,861 in 2000 to 61,430 in 2008 (Table 1). Similarly, the number of liveborn singleton neonates conceived after assisted reproductive technology increased 95% during the same time period. The percentage of good perinatal outcomes among all assisted reproductive technology liveborn neonates increased from 38.6% in 2000 to 42.5% in 2008 (P<.05). An increase in good perinatal outcomes among all assisted reproductive technology live births was observed in all assisted reproductive technology cycle groups, with a significant trend (P<.05) for all with the exception of frozen, donor cycles. The percentage of good perinatal outcomes among all singleton live births, however, declined slightly from 83.6% in 2000 to 83.4% in 2008. In addition, there was a small decline in good perinatal outcomes among liveborn singletons in the fresh, nondonor cycle group from 84.1% to 83.9%. A decline also was observed in the frozen, nondonor cycle group, although a significant trend was not detected (P=0.13). Some year-to-year variation was seen for the other cycle types, but overall the percentages of births with good perinatal outcomes changed little, with no significant trend detected.

The study population chosen to investigate factors associated with good perinatal outcomes—singleton neonates born after fresh, nondonor IVF transcervical embryo transfer in 2008 (n=20,780)—was similar in most characteristics to all liveborn neonates after fresh, nondonor IVF transcervical embryo transfer from the same year (Table 2), but was substantially different with respect to the frequency of good perinatal outcomes. The greatest differences between the two groups were observed in the proportion of neonates born at or after 37 weeks of gestation (87% of singletons and 62% of all liveborn neonates) and with a birthweight of at least 2,500 g (90% of singletons and 66% of all liveborn neonates). For both groups, most neonates were born to women 30–39 years of age and to women with no previous births. Approximately half of the neonates in both groups were born to women with one or more previous pregnancies. Although more than 30% of reported assisted reproductive technology cycles did not contain data for the race or ethnicity variable, most neonates for whom maternal race was reported were born to women of non-Hispanic white race. The most common infertility diagnosis in both groups was male factor, and uterine factor was the least common diagnosis. For both groups, in most procedures, intracytoplasmic sperm injection was used and use of assisted hatching was less common. Availability of supernumerary embryos for cryopreservation after assisted reproductive technology cycles was somewhat less common among singleton neonates (44%) than for all liveborn neonates (49%). Ten percent of singleton live births, and 6% of all live births, resulted from the transfer of a single embryo. Two or more fetal hearts were observed much less commonly in the assisted reproductive technology singleton group, and the assisted reproductive technology singleton group had a smaller proportion of embryos transferred on day 5 (at blastocyst stage). The groups were similar in terms of number of previous spontaneous abortions and number of previous assisted reproductive technology cycles.

Results of univariable analyses indicated that one previous birth, the transfer of one or two embryos compared with more than two, and the presence of one or two fetal hearts on 6-week ultrasound examination compared with more than two hearts were associated with good perinatal outcomes (Table 3). All of these associations remained significant in the final multiple logistic regression model. Non-Hispanic black race, tubal factor infertility, uterine factor infertility, ovulatory disorder, and 5 days of embryo culture were inversely associated with good perinatal outcomes in the final model. The strongest positive association with good perinatal outcome was estimated for the presence of one fetal heart on 6-week ultrasound examination as compared with more than two fetal hearts (adjusted OR 2.4, 95% CI 1.7–3.4), and the strongest negative association was estimated for non-Hispanic black as compared with non-Hispanic white race (adjusted OR 0.5, 95% CI 0.4–0.6). Limiting the sample to those with one heart on 6-week ultrasound examination (n=17,289) did not change the other associations retained in the final model to any appreciable extent (results not shown). When we limited the sample to those with one embryo transferred (n=2,128), most of the associations were in the same direction as observed in the original analysis, although not significant (results not shown). However, the transfer of blastocyst stage embryos was not associated with the outcome (adjusted OR 1.0, CI 0.8–1.4) in this additional analysis.

To assess the effect of the missing race or ethnicity data on our results, we limited the sample to those with nonmissing data (n=13,613). Most of the observed associations were retained in the final model (results not shown), with effect estimates more than 1.0 observed for one previous birth, the presence of fewer fetal hearts on 6-week ultrasound examination, and the transfer of one embryo; the effect estimate for tubal factor infertility remained less than 1.0. Most notably, one fetal heart on 6-week ultrasound examination remained the strongest predictor of good outcomes (adjusted OR 1.7, 95% CI 1.1–2.7) and non-Hispanic black race remained the strongest negative predictor (adjusted OR 0.5, 95% CI 0.4–0.6).

Because we did not expect tubal factor infertility to be inversely associated with good outcomes, we investigated this in greater detail. Because the groups of fertility diagnosis were not mutually exclusive, we repeated univariable analysis, comparing those solely diagnosed with tubal factor infertility (n=3,376) with all others (n=17,404), and association with the outcome was retained (results not shown).

Last, we tested for associations with good outcomes among all assisted reproductive technology live births in 2008 after fresh, nondonor assisted reproductive technology and found that most observed associations were retained in the final model, with effect estimates more than 1.0 for one previous birth, the presence of fewer fetal hearts on 6-week ultrasound examination, and the transfer of one embryo; the effect estimates for non-Hispanic black race, tubal factor infertility, ovarian disorder, and transfer of blastocyst stage embryo were less than 1.0. The presence of one fetal heart had the strongest association with good outcomes (adjusted OR 48.0, 95% CI 40.9–56.3) and non-Hispanic black race remained the strongest negative predictor (adjusted OR 0.6, 95% CI 0.5–0.7). Additionally, maternal age older than 30 was associated with good outcomes in this analysis. We also found that there were fewer multiple births in older women despite increased numbers of embryos being transferred.

Traditionally, success in assisted reproductive technology has been viewed as clinical live birth. More recently, it has been suggested that success should be defined as the birth of a healthy singleton.²⁹ We report an increase in good perinatal outcomes among all assisted reproductive technology live births from 2000 to 2008, a finding that likely reflects changes in the practice of assisted reproductive technology in the United States. After the first publication of American Society for Reproductive Medicine guidelines regarding embryo transfer in 1998 and 1999, the numbers of embryos transferred decreased sharply in the subsequent years.³⁰ Between 2000 and 2008, the percentage of fresh, nondonor IVF cycles that involved the transfer of a single embryo increased from less than 1% to 10%.¹ Although techniques and practices may have changed, singletons born after assisted reproductive technology remain at risk for adverse outcomes.

In our analysis, the strongest predictor of good outcomes was the presence of a single fetal heart. This finding is consistent with a recent analysis that demonstrated that spontaneous losses increase the odds for preterm birth and low birthweight in singletons.³¹ Furthermore, we found that transferring fewer embryos resulted in better outcomes.

Race was also an important predictor of perinatal outcomes. A recent large analysis regarding racial disparities in assisted reproductive technology with respect to birth outcomes reported increased risks for low birthweight and preterm delivery among singletons born to non-Hispanic black women.³² Although the reasons for racial disparity in assisted reproductive technology remain unclear, we found non-Hispanic black race to be strongly associated with a lower likelihood of good perinatal outcomes.

Uterine factor infertility, ovulatory disorders, tubal factor infertility, and embryo transfer after 5 days of embryo culture were inversely associated with good perinatal outcome. Tubal disease may alter the intrauterine environment during pregnancy. In a study of patients using donor oocytes, hydrosalpinges in the recipient were associated with higher miscarriage and ectopic pregnancy rates,³³ and a larger meta-analysis demonstrated that early pregnancy loss was more common in assisted reproductive technology patients with hydrosalpinx among those with tubal factor infertility.³⁴

Our analysis of national surveillance data should be considered in light of its strengths and limitations. The use of nationwide surveillance data provides increased statistical power and the restriction of our analysis to data from 1 year decreases the chances that our results are affected by changes in assisted reproductive technology.

Our findings regarding race should be interpreted with caution given that more than 30% of data regarding maternal race were missing. However, the similarity in the effect estimates for the unknown and the non-Hispanic white race or ethnicity groups, the lack of a change in our finding for non-Hispanic black race on restriction of the analyses to nonmissing data, and the fact that most women who undergo assisted reproductive technology are of non-Hispanic white race suggest that those for whom race or ethnicity data were missing were mostly non-Hispanic white women. Sufficient data were unavailable to examine the effects of body mass index, chronic disease, socioeconomic status, and behavioral or lifestyle characteristics on outcomes. Last, our study describes factors associated with outcomes among singleton live births and should be interpreted with this in mind.

These results demonstrate that good perinatal outcomes among singleton births after fresh, nondonor IVF are predicted by multiple factors. Single embryo transfer was associated with superior outcomes, whereas non-Hispanic black race was associated with less favorable outcomes. Future work is needed to investigate the mechanisms behind these effects and to identify reasons for racial disparities to improve outcomes.