Estimates of the live births, natural losses, and elective terminations with Down syndrome in the United States

Definitions

“Live birth prevalence” is defined as the number of live-born babies with DS in 10,000 live births. “Non-selective live birth prevalence” is defined as the proportion of babies that would be live-born with DS in the absence of elective pregnancy terminations, following a prenatal diagnosis of DS. In this non-selective scenario, some of the fetuses in pregnancies that are not electively terminated would still not have survived to term, as there is natural loss. Therefore, in estimating non-selective live birth prevalence, a correction for this natural loss should be made. In the U.S., no reliable information on the number of DS-related elective pregnancy terminations is available. However, we describe an alternative way of estimating non-selective live birth prevalence below.

Estimating Non-Selective Live Birth Prevalence for DS

The link between DS and rising maternal age has been established for many years [Hecht and Hook, 1996; Bray et al., 1998; Hassold and Hunt, 2001; Morris et al., 2003]. In estimating non-selective live birth prevalence, we made use of the most recent model derived from data collected by the National Down Syndrome Cytogenetic Register (NDSCR) in the United Kingdom [Morris et al., 2002].

We obtained data on births by maternal age in the U.S. from the Centers for Disease Control and Prevention (CDC) (Table I). We estimated total live births in 1900 [Haines, 2010]. Total live births from 1901–1908 were interpolated. We obtained data on the total live births in the U.S. from 1909 onwards from the U.S. Census Bureau, adjusted for under-registration prior to 1960 [U.S. Census Bureau, 2003]. In Figure 1, the procedures for estimating non-selective live birth prevalence are explained.

Estimating Live Birth Prevalence of DS in the U.S.

An overview of the sources and procedures used for estimating live births with DS is presented in Table II and Figure 1. Additional details can be found in Supplementary Material S1.

Estimating DS-Related Elective Pregnancy Terminations

To estimate the number of DS-related elective pregnancy terminations, we calculated the difference between the estimated non-selective live births and the estimated live births with DS. We then adjusted this number to account for the natural loss rate associated with DS. See Figure 1 for a visual representation of the estimation procedures described below.

One way of estimating natural loss is comparing live birth prevalence with prevalence found in prenatal diagnoses. However, this is only valid in historical populations without risk screening procedures. These older studies, as reviewed by Morris et al. (1999) and Savva et al.(2006), are restricted to women ages 35 years and older. Furthermore, these studies might overestimate natural loss as the ascertainment in live birth studies is likely lower than the ascertainment in prenatal studies [Savva et al., 2006].

As an alternative approach, Savva et al. (2006) conducted a survival analysis, following up on 5,177 prenatally diagnosed cases of DS from the NDSCR database (around 27% were found after CVS and around 73% after amniocentesis). In this analysis, terminated pregnancies (around 90%) were treated as censored observations. On basis of the study of Savva et al. (2006), we estimated natural loss in women older than 35 years of age to be 34% during the period following diagnosis by CVS and 27% subsequent to diagnosis by amniocentesis. In women younger than 35 years of age, this was estimated at 25 and 20%, respectively. We used the NDSCR database, 1989–2011, to gather the number of prenatal diagnoses of DS, the type of prenatal testing, and the maternal age. Using these data, we calculated that for every 100 elective terminations of pregnancies diagnosed with DS, there were 73 live births of DS prevented, with a range of 71–75% for different years. In developing our model, we assumed that this 100–73%, or 27%, natural loss rate will be similar to that in the U.S., as the maternal age distribution in the U.S. is similar to that of the U.K., and both countries followed the same policy in offering prenatal screening for DS to all pregnant women from the early 1990s onwards. We assumed a 24% natural loss rate for the period before 1990, as most procedures would have been amniocenteses in this period. Morris et al. (1999) reported an average of 23% loss rate for pregnancies with DS following amniocenteses, and Savva et al. (2006) reported an average of 25%.)

Live Birth Modeling With and Without Elective Pregnancy Terminations

For the U.S., we estimated the number of (i) live births with DS, (ii) elective pregnancy terminations of fetuses with DS, and (iii) natural losses between a gestational age of 10 weeks and expected date of birth. We did this for two different scenarios as explained in our Introduction. The estimation procedures for constructing these scenarios are visually depicted in Figure 1.

Differences by Region and Race/Ethnic Group

In addition, we analyzed the available data by region (Northeast, Midwest, South, West) and race/ethnic group (non-Hispanic whites, non-Hispanic blacks/Africans, Hispanics, Asians/Pacific Islanders, American Indians).

Confidence Intervals

For all of the outcome variables, we estimated confidence intervals. The procedures are described in Supplementary Material S2.

Impact of Elective Terminations of Pregnancies

For the U.S., the live birth prevalence for DS in the most recent years (2006–2010) was estimated at approximately 12.6 per 10,000 with a total of around 5,300 births annually. During this period, an estimated 3,100 DS-related elective pregnancy terminations were performed annually. If these DS-related elective pregnancy terminations had not occurred, we estimate that there would have been around 7,600 live births with DS annually, taking natural losses into account. For the U.S., as a whole, an estimated 30% of fetuses with DS were selectively terminated in recent years, more or less stable since 1996.

According to our modeled estimates, the annual number of DS-related elective pregnancy terminations rose sharply during the 1970s, 1980s, and 1990s, starting with around 46 in 1974 and reaching around 3,200 in 2000. In the 2000s these numbers stayed fairly constant around 3,000–3,300.

Regional Differences

In our analysis for the period 2000–2009, live birth prevalence turns out to be more or less similar in almost all surveillance programs—i.e., between 11–14 live births with DS per 10,000 live births. However, this conceals large differences in reduction percentages (Table III). In the U.S., according to our modeled estimates, there appears to be a relatively low overall reduction percentage of 30% (95% CI: 27.3–31.9% as of 2007). However, according to our modeled estimates, some states turn out to have much higher reduction percentage estimates, in particular Hawaii (62% as of 2005) and the Northeast states included in our analysis (on average 46% as of 2007). However, according to our modeled estimates, despite these large regional differences in reduction percentage estimates, in none of the four geographical regions of the U.S. has the average overall reduction percentage estimate increased in recent years. In three of these regions (Midwest, South, and West), there was even a slight decrease after 2005 (Table III).

Analysis of our modeled estimates reveals that states with high reduction percentage estimates have relatively older mothers. Older mothers tend to do more screening. However, there also appear to be cultural differences between states. The states from the Northeast (and Hawaii) seem to follow a different pattern than states from other parts of the U.S. In the Northeast, reduction percentage estimates were much higher for both older and younger women—around an estimate of 56% for mothers 35 years of age and older and 30% for younger mothers; whereas in the other parts of the U.S. (with the exception of Hawaii that also has high reduction percentage estimates of 64 and 58%, respectively), the corresponding values were at a lower level of around 38 and 15%, respectively (Table III).

Differences by Race/Ethnic Group

For the most recent 5-year period (2005–2009), we have race/ethnic group-specific data on live births of children with DS for twelve states (Arizona, Connecticut, Florida, Illinois, Indiana, Kentucky, Louisiana, Massachusetts, Michigan, Minnesota, New Jersey, New York). In addition, under the assumption that the model of Morris et al. (2002) for maternal age specific chances can be applied to different races/ethnic groups, we can estimate the non-selective live birth prevalence by race/ethnic group for these twelve states as well (Table IV). Reduction percentage estimates turn out to differ greatly by race/ethnic group—highest for Asians/Pacific Islanders (61%), followed by non-Hispanic whites (39%), non-Hispanic blacks/Africans (27%), Hispanics (18%), and American Indians (16%, though with a very wide 95% CI for this group) (Table IV).

Non-Selective Live Birth Prevalence

According to our estimates, between 1920–1940 in the U.S., both the number of live births with DS and live birth prevalence of DS decreased. During this period, an increasing number of women obtained access to birth control information and family planning services, resulting in a decrease in the average family size with fewer births and fewer older mothers [Centers for Disease Control and Prevention, 1999].

From 1940–1957, total fertility rates increased [Wetzel, 1990; Centers for Disease Control and Prevention, 1999]. Consequently, the total number of live births increased, and, so, too, did the total number of live births with DS. In regards to the live birth prevalence for DS, a small peak occurred between 1944–1945. During these years, millions of young men became involved in World War II, resulting in a temporary dip in the number of children born to their female partners between 15 and 30 years of age. As such, there was a higher percentage of older mothers during these years, and an increase in the live birth prevalence for DS. With the exception of this temporary peak, the live birth prevalence for DS stayed fairly constant during the period 1940–1957. Family size increased, resulting in a higher number of older mothers. However, the median maternal age at the first birth simultaneously dropped [Wetzel, 1990], resulting in a higher number of younger mothers, too. These two trends counterbalanced to result in a near zero net effect on live birth prevalence for DS.

In 1960, the era of modern contraception began when both the birth control pill and intrauterine device (IUD) became available [Centers for Disease Control and Prevention, 1999]. During the next two decades, there was a sharp decrease in the total number of births and the total number of live births with DS. Live birth prevalence for DS declined as well, caused by a decreasing number of older mothers and smaller average family size.

From the early 1980s onwards, as a consequence of postponed motherhood, the percentage of mothers older than 35 years of age grew again. This demographic trend produced a clear increase in non-selective live birth prevalence for DS. This development appears to plateau in most recent years.

Live Birth Prevalence

From the early 1970s onwards, prenatal services were developed and expanded. In the early years, prenatal diagnoses had little to no influence on the live birth prevalence of DS. However, from the late 1970s onwards, while non-selective live birth prevalence for DS greatly increased, there was a small decrease in live birth prevalence for DS in the late 1970s and only a small increase in the period 1980–2010.

With regards to reduction percentages, there appear to be cultural differences between states. In comparison with other parts of the U.S., in the Northeast (and in Hawaii), reduction percentage estimates were much higher for both older and younger women (Table III).

In addition, there appear to be racial/ethnic differences in reduction percentages. In our analysis, reduction percentage estimates were highest among Asians/Pacific Islanders, followed by non-Hispanic whites, non-Hispanic blacks/Africans, Hispanics, and American Indians. Using data from the Metropolitan Atlanta Congenital Defects Program for the period 1996–2005, other researchers concluded that DS-related elective pregnancy terminations were largest for non-Hispanic whites, followed by non-Hispanic blacks, and Hispanics, suggesting a similar racial/ethnic trend as in our analyzes [Jackson et al., 2014].

We made an estimation of the reduction percentage in the U.S. for women of 35 years of age and older versus their younger counterparts. If we were to make a similar estimation for only the 12 states with racial information, we could apply these estimates to the non-selective number of births of children with DS by maternal age group for each of the races/ethnic groups. This yields a prediction of live birth prevalence that would occur in pregnant women from the same age group, absent any behavioral differences by race/ethnic group in regard to prenatal screening, diagnostics, and DS-related elective pregnancy terminations. For non-Hispanic whites, the prediction is 5% higher than the live birth counts, and for non-Hispanic blacks, the prediction is 10% lower than the live birth counts. For Asians/Pacific Islanders, there is a 62% higher prediction, whereas for American Indians, there is a 20% lower prediction, and for Hispanics, a 19% lower prediction. This implies that the racial differences in reduction percentages can only be partly ascribed to maternal age differences between races/ethnic groups. Women in the same age group, but of different race/ethnic group, on average appear to act differently. In previous research, maternal race–ethnicity was associated with both the utilization of prenatal cytogenetic testing and with termination of a pregnancy after diagnosis of a birth defect [Jackson et al., 2014]. Compared to non-Hispanic whites, non-Hispanic black and Hispanic women were significantly less likely to undergo elective pregnancy termination after an abnormal result from a prenatal cytogenetic test, suggesting cultural differences in attitudes towards DS-related elective pregnancy terminations.

Finally, our estimate of live birth prevalence of DS for 2010 is around 1 in 792 live births. Our estimate of total case prevalence, including both live births with DS, DS-related elective pregnancy terminations and modeled natural loss (fetal loss and stillbirths) after the gestational age of around 10 weeks, is around 1 in 365 live births. Parker et al. (2010) estimated total case prevalence of DS for 2004–2006 at around 1 in 691 live births. The discrepancy between our estimates can partly be ascribed to Parker et al. (2010) including only natural loss occurring after 20 weeks gestational age. In addition, there is an under-ascertainment of both natural loss and DS-related elective pregnancy terminations even in surveillance programs that count both. For instance, for the period 2006–2010, Mai et al. (2013) calculated a pooled live birth prevalence of DS of 14.2 per 10,000 live births in surveillance programs counting all pregnancy outcomes (live births, still births, and DS-related terminations). The pooled live birth prevalence of DS was only 12.6 per 10,000 live births in surveillance programs counting only live births and stillbirths and 12.5 per 10,000 live births in surveillance programs counting only live births. As of 2007, our model would predict 26.4 per 10,000 live births for all pregnancy outcomes, 19.1 per 10,000 for live births and natural loss after 10 weeks gestational age (GA), and 12.6 per 10,000 for live births only. Our model would predict around 3,100 DS-related elective terminations as of 2007. Mai et al.'s model for 2006 would suggest around 700 DS-related elective terminations (1.6 per 10,000 live births), suggesting significant under-ascertainment. According to Appendix 6.1 of the Guidelines for Conducting Birth Defects Surveillance, in most states early fetal deaths (GA < 20 weeks) are not reported on fetal death certificates and are under-reported in states that register early fetal deaths [National Birth Defects Prevention Network, 2014]. Elective terminations (which actually will often occur before GA of 20 weeks) are under-reported on both fetal death certificates and in termination certificates, as compliance to guidelines of the selected states that ask for termination certificates is notably poor [National Birth Defects Prevention Network, 2014]. Regarding DS, our study corroborates this poor ascertainment of both fetal loss and elective terminations in the surveillance programs. Our estimates avoid this issue by constructing these variables on basis of the difference between our maternal-age-based estimate of non-selective live births with DS and our surveillance-program-based estimate of live births with DS, in combination with the fetal survival analysis of Savva et al. (2006).

Comparison With International Studies

According to long-term studies from England and Wales [Morris and Alberman, 2009], Slovenia [Tul et al., 2007], Australia [Bittles et al., 2007; Collins et al., 2008], and most EUROCAT regions [Dolk et al., 2005], the effect of increasing maternal age was counterbalanced by the more widespread use of prenatal screening, resulting in a stable, or slightly decreasing, live birth prevalence for DS, beginning in the 1990s. In the Netherlands, by contrast, live birth prevalence for DS has been increasing since the early 1990s [de Graaf et al., 2011a,2011b]. Our present study shows that the U.S. is more like the Netherlands in this respect, with the live birth prevalence of DS slightly increasing since the 1990s. In addition, both in the Netherlands and in the U.S., the reduction percentage associated with DS-related elective pregnancy terminations was relatively low, around 35% for the Netherlands, as of 2007 [de Graaf et al., 2011b] and 30% for the U.S by our calculations (2006–2010). By comparison, the reduction percentages were 55% for Australia, as of 2004 [Bittles et al., 2007], 48% for the UK, as of 2008 [Morris and Alberman, 2009], and 47% for Slovenia, as of 2005 [Tul et al., 2007].