We estimated the prevalence of trisomy 21 among all pregnancy outcomes for the three race–ethnicity groups during the years 1996–2005 (Table I). Compared with non-Hispanic whites, a significantly lower prevalence of trisomy 21 was observed among non-Hispanic black and Hispanic pregnancies in the combined maternal age category (prevalence ratio [PR] = 0.62, 95% confidence interval [CI] 0.53–0.73 and PR = 0.64, 95% CI 0.52–0.8, respectively). Non-Hispanic black mothers aged <35 years had a significantly lower prevalence of trisomy 21 pregnancies compared with non-Hispanic white mothers aged <35 years (PR = 0.74, 95% CI 0.58–0.95). However, the prevalence of trisomy 21 among Hispanic mothers aged <35 years was similar to that among non-Hispanic white mothers aged <35 years; the prevalence of trisomy 21 among mothers 35 years and older did not vary significantly by maternal race–ethnicity.

The live birth prevalence of trisomy 21, and the live birth prevalence adjusted for electively terminated trisomy 21 pregnancies, were calculated for each of the three race–ethnicity subgroups (Table I). The live birth prevalence of trisomy 21 was significantly lower among non-Hispanic black mothers of all ages compared with non-Hispanic white mothers of all ages (PR = 0.77, 95% CI 0.64–0.93; Table I). When adjusted for elective termination, the live birth prevalence of trisomy 21 among both non-Hispanic black and among Hispanic mothers of all ages were statistically significantly lower compared with non-Hispanic white mothers (PR = 0.64, 95% CI 0.54–0.76, and PR = 0.69, 95% CI 0.55–0.86, respectively). Non-Hispanic black mothers aged <35 years also had a significantly lower live birth prevalence when adjusted for elective termination compared with non-Hispanic white mothers aged <35 (PR = 0.76, 95% CI 0.59–0.98). However, the adjusted live birth prevalence of trisomy 21 among Hispanic mothers aged <35 years was similar to that among non-Hispanic white mothers aged <35; the adjusted live birth prevalence of trisomy 21 among mothers 35 years and older did not vary significantly by maternal race–ethnicity.

Prenatal cytogenetic testing was reported significantly less frequently among Hispanic women (Table II) compared with non-Hispanic white women (odds ratio [OR] 0.66, 95% CI 0.56–0.78) using a multivariate logistic regression model to adjust for birth year and maternal age. In contrast, non-Hispanic black women underwent a similar frequency of prenatal cytogenetic testing (OR 0.95, 95% CI 0.87–1.00). However, compared with non-Hispanic white women, both non-Hispanic black and Hispanic women were significantly less likely to undergo elective termination after an abnormal prenatal cytogenetic test result (OR 0.50, 95% CI 0.36–0.70 and OR 0.49, 95% CI 0.27–0.88, respectively).

The outcomes of all pregnancies with defects and of those pregnancies affected by trisomy 21 were analyzed for the years 1996–2005 by maternal race–ethnicity (Table III). Among all pregnancies with defects ascertained by MACDP, live births occurred most frequently among Hispanics and fetal deaths occurred most frequently among non-Hispanic blacks; elective termination occurred most frequently among non-Hispanic white pregnancies. Among pregnancies affected by trisomy 21, live births occurred most frequently among Hispanic and non-Hispanic black pregnancies, while elective terminations occurred most frequently among non- Hispanic white pregnancies. Fetal deaths were similar between race-ethnicities but occurred most frequently among non-Hispanic black pregnancies.

We found that maternal race–ethnicity was associated with both the utilization of prenatal cytogenetic testing and termination of a pregnancy after diagnosis of a birth defect. Compared with non-Hispanic white mothers, Hispanic mothers underwent prenatal cytogenetic testing significantly less often. This finding is in keeping with the literature which also has shown lower use of amniocentesis among Hispanic women [Baker et al., 2004]. It has been hypothesized that race–ethnicity differences in the utilization of prenatal diagnosis are due in part to differences in attitudes towards elective terminations [Press and Browner, 1998; Li et al., 2008]. This is supported by our finding that non-Hispanic black and Hispanic women were significantly less likely to undergo elective termination after an abnormal result from a prenatal cytogenetic test.

In comparing the overall prevalence of trisomy 21 among race–ethnicity groups, both the non-Hispanic black and Hispanic populations had a statistically significantly lower prevalence than the non-Hispanic white population; however, when only the live birth prevalence was examined, a statistically significant difference remained only for the non-Hispanic black population. The proportion of pregnancies with trisomy 21 that resulted in fetal death was similar among race–ethnicity groups (3.9% non-Hispanic white, 4.5% non-Hispanic black, and 3.8% Hispanic), suggesting that the differences in the overall prevalence of trisomy 21 largely reflect variations in the proportion of affected pregnancies that resulted in elective termination (32.3% non-Hispanic white, 15.2% non-Hispanic black, and 5.7% Hispanic). Indeed, when the live birth prevalence was adjusted for the estimated 74% of elective terminations that would have otherwise resulted in a live birth, both the non-Hispanic black and Hispanic populations again had a significantly lower prevalence of trisomy 21 compared with the non-Hispanic white population in the same age groups.

Some of these findings are in contrast to previous reports in the literature. Similar to our observations, some studies have reported a lower prevalence of trisomy 21 among non-Hispanic blacks compared with non-Hispanic whites [Bishop et al., 1997; Canfield et al., 2006; Shin et al., 2009]. Unlike our data, however, others have reported as much as a 20% higher prevalence of trisomy 21 in Hispanic populations compared with non-Hispanic whites, while others have found no statistically significant difference between these race–ethnicity groups [Centers for Disease Control and Prevention, 1994; Bishop et al., 1997; Hook et al., 1999; Canfield et al., 2006; Shin et al., 2009]. Bishop et al. [1997] did report a slightly lower prevalence of trisomy 21 among Hispanics compared with non-Hispanic whites, but only after adjusting for the probability of survival to birth if elective termination had not been chosen; the unadjusted live birth prevalence reported by these authors is similar to other literature reporting a higher prevalence of trisomy 21 among Hispanics [Bishop et al., 1997]. We hypothesize that the cause of the lower prevalence of trisomy 21 among Hispanics in the MACDP surveillance area may be at least twofold, as described below.

The Hispanic population in the MACDP area might be younger than those populations reported in other studies, since they did not provide mean maternal age data. During the 1996–2005 time period, the mean age of all women with pregnancies ≥20 weeks gestation (regardless of birth defect status) in the five-county MACDP area was 30.1 years (standard deviation 5.7) for non-Hispanic white mothers and 25.7 (standard deviation 5.7) for Hispanic mothers. There also was a greater proportion of mothers aged ≥35 years in the non-Hispanic white population than in the Hispanic population (22.2% of non-Hispanic white mothers compared with 7.9% of Hispanic mothers). While not reaching the level of significance, the live birth prevalence of trisomy 21 for Hispanic mothers aged ≥35 years was higher than that for white mothers aged ≥35 (48.7 vs. 33.6 per 10,000 live births, respectively, Table I). The role of maternal age is supported by other data presented in Table I. The prevalence ratios for pregnancies affected by trisomy 21 in Hispanic women stratified by maternal age are 1.0 and 0.93 (<35 and ≥35, respectively), which are not significantly different from those for non-Hispanic white women. This comparison is most likely confounded by maternal age, as the overall PR is 0.64 for Hispanic women. If the maternal age distributions for these two race–ethnicity populations were more similar, the overall trisomy 21 live birth prevalence might shift towards a higher prevalence among Hispanics as has been reported by others.

As seen in our data and noted by others [Baker et al., 2004; Jones et al., 2010; Pazol et al., 2011], there is a significant difference in the prevalence of prenatal cytogenetic testing and of elective termination among race–ethnicity groups, which would affect the live birth prevalence [Mikkelsen, 1992; Cornel et al., 1993; Krivchenia et al., 1993; Bishop et al., 1997; Forrester and Merz, 1999]. Although speculative, if the same frequencies of prenatal cytogenetic testing and elective termination observed in the non-Hispanic white trisomy 21 population were applied to the Hispanic trisomy 21 population, the live birth prevalence of trisomy 21 would be nearly identical for non-Hispanic whites and Hispanics (maternal age <35 years, 7.3 and 7.4 per 10,000 live births, respectively; maternal age ≥35 years, 33.9 and 31.4 per 10,000 live births respectively). To our knowledge, no other study has analyzed the prevalence of trisomy 21 for these race–ethnicity groups in light of differences in the maternal age distribution, the frequency of prenatal testing, and the frequency of elective termination after a prenatal diagnosis, which may explain some of the disparate results seen in the literature and why our results differ from previous reports.

There are several limitations to these analyses. These results represent the use of prenatal cytogenetic testing and elective termination within a population with confirmed birth defects, rather than within the general population of all pregnant women. Also, MACDP does not receive results from all cytogenetic laboratories that service Atlanta, which probably results in some under-ascertainment of the actual use of cytogenetic testing and of chromosomal diagnoses. Further, there may be differential ascertainment between race–ethnicity groups as a result of the distribution of these groups within the catchment area, as well as differences in access to care. The limited number of cases particularly among the Hispanic population could influence the statistical stability of our estimates. Finally, the probability of survival to birth if elective termination had not been chosen might differ by race–ethnicity; the population used by Hook et al. [1989] to establish the 74% chance of survival was based on survey responses from prenatal cytogenetics laboratories in North America, however no demographic data or information about the responding laboratories were provided.

This study has several strengths. This is a population-based study rather than hospital-based or a passive registry. Also, MACDP uses multiple ascertainment sources (such as birth hospitals, pediatric hospitals, specialty clinics, perinatal offices, cytogenetics laboratories, and vital records) to identify all infants, fetuses, or children with birth defects within the five-county catchment area. Highly trained MACDP abstractors actively search multiple sources at each ascertainment site and all medical records of identified cases are thoroughly reviewed to assure completeness of information. Finally, case abstractions are reviewed through a multi-tiered system involving pediatric and clinical genetics staff to ensure accuracy of diagnosis.