National Prevalence Estimates
The national prevalence estimates for 21 major birth defects were determined using data from 14 population-based surveillance systems, representing 32.3% of all live births in the United States from 2004 through 2006. The birth defects selected for this study included major structural malformations of several systems, such as the central nervous, cardiovascular, and gastrointestinal systems, and chromosomal anomalies. This study provides an update to previous national estimates of selected birth defects, which were based on data from 1999 through 2001 (Canfield et al., 2006).
The national estimates for the selected defects we presented have several strengths. The estimates were based on confirmed cases of birth defects from population-based surveillance systems that represented both demographic and geographic variations in the United States. Additionally, availability of data on maternal race and ethnicity allowed for adjustment, given reported variation in prevalence of defects based on race and ethnicity (Canfield et al., 2006; Canfield et al., 2009; Carmichael et al., 2004; Nembhard et al., 2010; Williams et al., 2005). Furthermore, adjustments for maternal age were made for chromosomal anomaly rates. The number of annual estimated cases increased slightly for Down syndrome when maternal age was taken into account. This increase was explained by the higher prevalence of Down syndrome among offspring of older mothers and the increasing contribution to the total birth population (Allen et al., 2009).
An additional strength of the national estimates was the inclusion of cases other than live births, which allowed for a more accurate estimation of the prevalence of the selected defects. Of the 14 programs used in determining the national estimates, nine ascertained cases among live births, stillbirths, and elective terminations, while the remaining five included both live births and stillbirths. Limiting the data to live birth cases only can greatly underestimate the prevalence of certain birth defects. The inclusion of stillbirths and elective terminations influences the prevalence of specific conditions more so than others, with fatal and severe conditions affected the most (Schechtman et al., 2002). For the majority of the selected defects, the updated national estimates remained similar to those reported previously (Canfield et al., 2006). Changes in the surveillance methodology of individual birth defects programs, coding modifications of congenital heart malformations between the two data collection periods, and slightly differing designs in the study methods could have accounted for some of the variability between the two reports.
Significant changes in the estimated national prevalence were observed for anencephaly, TGA, and gastroschisis. The decrease in prevalence of anencephaly from 2.51 to 2.06 (p < 0.01) was consistent with previous reports of the decrease of neural tube defect (NTD)-affected pregnancies since folic acid fortification. A study of the prevalence of spina bifida and anencephaly during the post-fortification period observed a significant decrease in the prevalence of anencephaly, but not spina bifida, from the period 1999 through 2000 to the period 2003 through 2004 (Boulet et al., 2008). A possible explanation for the post-fortification decrease in the birth prevalence of anencephaly (vs. spina bifida) over time might have been due to changes in prenatal diagnostic practices that could have made it more difficult for population-based surveillance systems to ascertain this fatal condition. Peller et al. (2004) found increases in prenatal screening and elective terminations for anencephaly from 1974 through 1999. If diagnostic tests were done earlier during pregnancy, or if elective terminations were performed in an outpatient clinic that surveillance systems did not routinely cover, then such cases would not have been captured by the surveillance system and the birth prevalence of selected defects could have been underestimated. Another reason for the decrease could have been the secular trend of decreasing NTD rates since the late 1960s (Lary and Edmonds, 1996; Yen et al., 1992).
The decline in the estimated prevalence of TGA from 4.73 to 3.00 (p < 0.01) could have been explained by changes made to the heart defect code inclusions in the 2009 NBDPN annual report data request. A study conducted using data from the Metropolitan Atlanta Congenital Defects Program (MACDP) demonstrated that relying strictly on administrative defect codes produced many false positives for TGA when compared to the use of a more clinically relevant nomenclature (Strickland et al., 2008). Subsequently, in an effort to improve the accuracy of reporting, the NBDPN revised its cardiovascular defect code list to exclude certain subtypes of TGA codes that might not actually have represented true cases of the TGA defect. Other cardiovascular defects that underwent changes in code inclusion criteria included tetralogy of Fallot and AVSD. This also explained the findings in our analysis of prevalence based on the method of surveillance, that passive systems showed a higher pooled prevalence for these heart defects. The elevated prevalence of specific heart defects among passive surveillance programs might be explained by the contribution of false positives. Such systems rely on reports from administrative data sources and do not use a follow-up system to confirm the cases. Reliance only on physician diagnoses might result in greater misclassification of cases than if a confirmatory diagnostic test is required for inclusion (Hobbs et al., 2001).
In addition to anencephaly and TGA, gastroschisis showed a significant change between the two time periods. The increase observed in the prevalence of gastroschisis, 3.73 to 4.49 (p < 0.01), compared to previous national estimates, was in concordance with studies from California, North Carolina, Texas, and Utah, all of which reported significant increases in gastroschisis the past decade (Benjamin et al., 2010; Hougland et al., 2005; Laughon et al., 2003; Vu et al., 2008).
Several defects showed slight, nonsignificant decreases in the national estimated prevalence between the two time periods. These changes in prevalence might have been explained by the inclusion of programs with passive case ascertainment in the updated estimates. With the exception of the chromosomal anomalies, all estimates remained similar or increased when the analysis was limited to just the active programs. None of the observed changes were significant (unpublished data).
Although we limited the programs included in the national estimates to those with either an active ascertainment methodology or an active case confirmation component, some variation in state-specific prevalence still remained across programs. Birth defect studies using data from multiple surveillance systems have observed variations in the reported prevalence of defects such as spina bifida and Down syndrome (Lary and Edmonds, 1996; Shin et al., 2009). A study of the effect of grain fortification on the prevalence of 16 birth defects based on data from 23 surveillance programs showed considerable state-to-state variation in the calculated prevalence ratios comparing prefortification and postfortification estimates. Heterogeneity in state-specific prevalence rates is to be expected and does not invalidate aggregating state-level data (Canfield et al., 2005). Sources of variation might include varying case inclusion criteria, such as age of case ascertainment or disease coding system. Variation also could be explained by differing surveillance methodologies, such as whether a state includes all pregnancy outcomes. The higher prevalence of anencephaly, anophthalmia/microphthalmia, and trisomy 18 among active programs might be explained by the inclusion of elective terminations (Ethen and Canfield, 2002). Of the 11 active surveillance programs, nine included elective terminations in their case definition; only one of the passive surveillance programs did. Furthermore, several of these active programs conducted specialized prenatal ascertainment involving abstraction at prenatal diagnostic facilities and clinical genetics facilities, ultimately increasing their ability to ascertain additional prenatal cases (National Birth Defects Prevention Network, 2009b).
The availability of pregnancy outcome information for a subset of the selected defects provided the opportunity to describe the contribution of each individual pregnancy outcome toward the overall prevalence. Pregnancy outcome data for NTDs, anencephaly and spina bifida, and chromosomal anomalies were collected based on previous studies indicating the contribution of elective terminations on the prevalence of these defects (Forrester et al., 1998; Schechtman et al., 2002). The prevalence of anencephaly was impacted the most with the inclusion of elective terminations, increasing 149% from the prevalence among live births and stillbirths. Studies consistently have shown the strong impact of including elective terminations on the prevalence of anencephaly (Forrester et al., 1998; Peller et al., 2004; Velie and Shaw, 1996). A study from MACDP showed a 128% increase in the prevalence of anencephaly when elective terminations were included (Cragan and Gilboa, 2009).
Other defects influenced by the inclusion of elective terminations were trisomy 13 and trisomy 18. When elective terminations were included, they made up 41.5% and 34.0% of all cases ascertained, respectively. These findings were consistent with an analysis of MACDP data that indicated elective terminations accounted for 45.8% of trisomy 13 cases and 48.4% of trisomy 18 cases (Crider et al., 2008). The three birth defects surveillance programs selected for the pregnancy outcome analysis had enhanced prenatal data ascertainment from multiple sources, including prenatal diagnostic facilities and cytogenetic laboratories; however, underascertainment of prenatal cases was still likely. These findings highlight the importance of including additional pregnancy outcomes to enhance the surveillance of birth defects.
Although our study provided prevalence estimates based on surveillance methodology, updated national estimates, and detailed information on the impact of ascertaining additional pregnancy outcomes on the prevalence of selected defects, several limitations should be noted. First, the birth defect categories were not composed of a homogenous group of cases. Isolated and nonisolated patterns were combined within categories and might have had very different etiologies. Second, with the exception of race and ethnicity and maternal age, information on other possible covariates was unavailable, and was not accounted for in the national estimates. Maternal age adjustments were made for the chromosomal anomalies, but such adjustments could not be made for other defect categories. Third, cases with multiple birth defects were included in each relevant category, thereby overestimating the total number of affected births.