This study was approved by the Institutional Review Board for Human Subjects Research at Baylor College of Medicine. If required by the state or institution, each surveillance program received specific approval through its respective institutional review board.

We obtained data on cases with Turner syndrome from seven, population-based birth defects surveillance programs (hereafter, programs) and on female live births from state birth records or state vital statistics offices. The programs were Arkansas, South Carolina, Texas, and Utah (statewide for birth years 2000–2017), California (10 counties for 2010–2012 and 2017, eight counties for 2013–2016), Massachusetts (statewide for 2011–2017), and the Metropolitan Atlanta Congenital Defects Program (MACDP; five counties for 2000–2011, three counties for 2012–2017).

Programs ascertained cases with Turner syndrome diagnosed within the first year of life. All cases were clinically reviewed and coded using the following codes: World Health Organization’s (WHO) International Classification of Disease (ICD) 9th revision: 758.6; ICD 10th revision: Q96.0-Q96.9; or the British Pediatric Association Classification of Diseases: 758.600–758.690 (Table S1). Case inclusion required confirmation by karyotype or chromosomal microarray analysis. Cases with a Turner syndrome phenotype that were diagnosed only by fluorescence in situ hybridization or noninvasive prenatal screening were excluded, as were those with <10 cell mosaic karyotype. Per recent diagnostic guidelines (Gravholt et al., 2017; Lin et al., 2019), we excluded the following karyotypes, which do not strictly result in Turner syndrome: 45,X/46,XY with male sex assigned at birth; X-autosome translocations and 45,X with trisomy 21; 46,XX,del(q24) and 46,X,del(X)(q25); and 46,X,idic(X) (q24). Case eligibility was confirmed by a medical geneticist (A.E.L.).

Programs abstracted the following demographic and clinical information on cases and female live births from medical or vital records: clinical and calculated estimates of infant gestational age at delivery (continuous), infant birth weight (continuous), maternal age at delivery (continuous), maternal race and ethnicity (Hispanic, non-Hispanic American Indian or Alaska Native, non-Hispanic Asian or Pacific Islander, non-Hispanic Black, non-Hispanic White, or other/unknown), and year of delivery (continuous). Programs submitted data on pregnancy outcomes as live birth, stillbirth (fetal death ≥20 weeks gestation), elective termination, miscarriage/unspecified non-live birth (fetal death <20 weeks gestation), or unknown. Massachusetts ascertained “other pregnancy losses,” defined as elective terminations and miscarriages <20 weeks. MACDP did not collect data on miscarriages occurring <20 weeks gestation.

Diagnoses of any of the “core” birth defects included in the NBDPN Guidelines for Conducting Birth Defects Surveillance Appendix 3.1 (National Birth Defects Prevention Network, 2021) were collected for each case. These defects were chosen because they are commonly ascertained by surveillance programs, although not all programs collected all birth defects across all years of the study. Importantly, these “core” birth defects do not represent all major defects that could occur or be characteristic of a Turner syndrome phenotype.

We collected survival data for live birth cases from states that follow cases up to at least 1 year of age: California and Massachusetts (follow-up data through 1 year of age) and Arkansas, Utah, Texas, and MACDP (follow-up data through 5 years of age). Survival data included infant death (yes/no) and length of follow-up time (if death occurred, the number of days from birth to death; if death did not occur, length of time in days for which the program ascertained deaths). All programs obtained death information through linkage to vital records or medical record abstraction.

We used data for gestational age based on the clinical (ultrasound) estimate. If the clinical estimate was missing, we used the calculated estimate (based on date of last menstrual period) or an estimate based on standard neonatal exam. Using the WHO definition, we classified birth weight for gestational age as small (birth weight < 10th centile), appropriate, and large (birth weight > 90th centile) using a published US birth weight reference (Talge et al., 2014). In analyses that assessed birth weight for gestational age, we excluded cases (N = 630) and live births (denominator data, N = 4261) with gestational age < 22 weeks to be consistent with the birth weight range used in the reference.

Due to small case counts, we grouped cases with maternal race and ethnicity designated as “non-Hispanic American Indian or Alaska Native” or “non-Hispanic Asian or Pacific Islander” into a single category, labeled as “non-Hispanic Alaska Native, American Indian, Asian, or Pacific Islander.” Across all programs, we also categorized pregnancy outcomes as live birth, other pregnancy loss (elective termination at any gestational age and miscarriage <20 weeks gestation), stillbirth (fetal death ≥20 weeks gestation), or unknown. Prevalence estimates and karyotype proportions were stratified by type of pregnancy outcome: (1) all pregnancy outcomes, and (2) live birth and stillbirth cases only.

We estimated the prevalence of Turner syndrome per 10,000 female live births (number of Turner syndrome cases divided by the number of female live births within the individual state or for the total surveillance catchment area, multiplied by 10,000) for (1) each program and (2) all programs combined. Non-live births were not included in the denominator since including non-live births has not been shown to significantly impact prevalence estimates of other birth defects (Marshall et al., 2015). We also estimated the prevalence by maternal race and ethnicity. Sensitivity analyses of the crude prevalence included (1) case and denominator data excluding early losses (≤20 weeks gestational age) and (2) leave-one-out analysis to identify variability in program ascertainment. To calculate the proportion of cases with core birth defects and 95% confidence intervals (CI), we used the exact binomial method.

We estimated prevalence ratios (PR) for maternal age at delivery (categorical: <25, 25–29 [reference], 30–34, 35–39, or ≥ 40 years), maternal race and ethnicity (Hispanic, non-Hispanic Black, non-Hispanic White [reference], or non-Hispanic Alaska Native, American Indian, Asian, or Pacific Islander), infant birth weight for gestational age (small, appropriate [reference], or large), and year of delivery (categorical: 2000–2002 [reference], 2003–2005, 2006–2008, 2009–2011, 2012–2014, or 2015–2017) using Poisson regression. Cases with missing values for a specific variable were excluded from the respective analysis.

We estimated survival probability of infants with Turner syndrome at 1 day, 7 days, 28 days, 1 year, 2 years, and 5 years. Survival probability was defined as the time between birth and death due to any cause or length of follow-up. We estimated the survival probability among (1) all cases and (2) cases without hypoplastic left heart (HLH), a congenital anomaly that is associated with poor survival (Lara et al., 2017).

We constructed Kaplan–Meier curves and used the log-rank test to determine differences in survival by maternal race and ethnicity, year of delivery category, presence of at least one core birth defect, or karyotype category (45,X; mosaicism; structural abnormalities of the X; or other, defined as any karyotype not included in the previous groupings). Associations with p-values <0.05 were considered significant, and all log-rank tests met the proportional hazards assumption. All statistical analyses were conducted in R version 4.0.4.

For the study period 2000–2017, we identified 1872 cases with Turner syndrome diagnosed up to the first year of life among 5,911,109 female live births across the seven surveillance programs. The prevalence of Turner syndrome among all pregnancy outcomes was 3.2 per 10,000 female live births (95% CI = 3.0–3.3, Table 1). When we restricted the numerator to live birth and stillbirth cases, the prevalence of Turner syndrome was 1.9 per 10,000 live births (95% CI = 1.8–2.1). Turner syndrome prevalence among all outcomes was highest in Massachusetts (10.4, 95% CI = 9.2–11.8) and lowest in South Carolina (1.0, 95% CI = 0.8–1.3).

The crude prevalence also differed by maternal race and ethnicity. The prevalence was highest among cases born to women who identified as non-Hispanic White (3.6 per 10,000 female live births, 95% CI = 3.4–3.9) followed by cases born to women who identified as Hispanic (2.6 per 10,000 female live births, 95% CI = 2.4–2.8), non-Hispanic Alaska Native, American Indian, Asian, or Pacific Islander (2.3 per 10,000 female live births, 95% CI = 1.8–2.9), and non-Hispanic Black (1.7 per 10,000 female live births, 95% CI = 1.4–2.0).

Recognizing that Massachusetts ascertains very early losses and that MACDP does not collect data on miscarriages <20 weeks gestation, we conducted a sensitivity analysis to address skewness in the data. We estimated the prevalence among cases of any pregnancy outcome and female live births (denominator data) that were born >20 weeks gestation (Table S2). In this analysis, the prevalence across all programs was 2.1 per 10,000 female live births (95% CI = 2.0–2.2). For Massachusetts, the prevalence was 1.7 per 10,000 female live births.

To determine the effect of each individual program’s data on the combined overall prevalence estimate, we also conducted a leave-one-out sensitivity analysis, which estimated the overall prevalence by leaving out exactly one program in subsequent iterations. We found significant differences between the combined prevalence estimate and the resulting prevalence estimate after removing data from Utah, Massachusetts, Texas, and South Carolina (Figure S1).

Among all cases, 40.5% (N = 759) were live births, 20.6% (N = 386) were stillbirths, 35.0% (N = 655) were other pregnancy losses, and 3.9% (N = 72) were of unknown outcome. The most frequent karyotype among all cases was 45,X (75.4%, Table 2). Mosaicism and structural X abnormalities comprised 9.2% and 6.5% of the study cohort, respectively. The proportion of all cases with at least one core birth defect was 25.2% (N = 472). The most common defects were coarctation of the aorta (11.6%), atrial septal defect (6.8%), and ventricular septal defect (3.8%, Table 3).

We excluded cases and female live births (denominator data) with missing data for maternal age (48 cases; 123,210 live births), maternal race and ethnicity (223 cases; 169,598 live births), and birth weight for gestational age (876 cases; 455,368 live births). Resulting counts are shown in Table 4.

The prevalence of Turner syndrome did not significantly differ for any maternal age group compared to the 25–29 year group nor by any year of birth category compared to 2000–2002 (Table 4). We did find that the prevalence of Turner syndrome was 50% lower for cases born to women who identified as non-Hispanic Black (PR = 0.5, 95% CI = 0.4–0.6) compared to women who identified as non-Hispanic White. We also observed a significantly lower prevalence of Turner syndrome among cases born to women who identified as Hispanic (PR = 0.7, 95% CI = 0.6–0.8) or non-Hispanic Alaska Native, American Indian, Asian, or Pacific Islander (PR = 0.6, 95% CI = 0.5–0.8) compared to women who identified as non-Hispanic White.

Among all cases, overall survival for individuals with Turner syndrome was >90% at all-time points from 1 to 5 years (Table 5). Among cases without HLH, survival was also >90% at all-time points. One-year survival did not significantly differ by karyotype, maternal race and ethnicity, year of delivery, or the presence of at least one core birth defect among all cases nor the subset of cases without HLH (Figures S2–S4).

Using data from seven NBDPN birth defects surveillance programs, we report the population-based estimate of the prevalence of Turner syndrome in the United States, which, in previous reports, has largely been unaccompanied by karyotype examination, assessments of the association of demographic characteristics, or survival analyses. The lack of such data in the United States compared to European countries (Berglund et al., 2019, 2020; Gravholt et al., 2022; Ji et al., 2016; Schoemaker et al., 2008; Zelinska et al., 2018) can be largely attributed to differences in health care structure. Unlike European countries, which often have access to nationalized health care registry data (Laugesen et al., 2021; Schmidt et al., 2019), the United States does not have a nationwide birth defects or Turner syndrome registry and relies on medical care centers and state birth defects surveillance programs for case ascertainment.

Within the present study, there were substantial differences by program. For example, the prevalence of Turner syndrome was lowest in South Carolina. Turner syndrome cases in South Carolina were primarily ascertained through the Greenwood Genetic Center, which contributed data for approximately one-third of the state (National Birth Defects Prevention Network, 2022); data from cytogenetic centers in other regions of the state were not available. Although the South Carolina Department of Health and Environmental Control began an active case-finding birth defects program in 2006, these ascertainment issues likely resulted in a significant underestimate of the true prevalence of Turner syndrome in this state. These data can serve as a reference to inform strategies that aim to improve case ascertainment.

Turner syndrome prevalence across all pregnancy outcomes was highest in Massachusetts and was significantly different compared to other programs. In 2011, Massachusetts began collecting data on other pregnancy losses, including early miscarriages and terminations, in addition to live births and stillbirths (Massachusetts Department of Public Health, 2013). For this reason, our analyses did not include data from Massachusetts prior to 2011. Although five other programs conducted surveillance for these outcomes, Massachusetts ascertains karyotype information from very early losses (<18 weeks gestational age) at several large, tertiary care hospitals in the state. Our sensitivity analysis showed that the prevalence of Turner syndrome in Massachusetts was similar to other states when excluding early losses.

We also observed a significantly higher prevalence of Turner syndrome in Utah compared to other programs. After a study from the Utah Birth Defect Network found that 12.6% of infant girls with coarctation of the aorta had a confirmed Turner syndrome karyotype, Utah’s Primary Children’s Hospital began offering cytogenetic testing to all girls who presented with this congenital heart defect (Eckhauser et al., 2015), which might be responsible for the higher prevalence. Additional factors that could have contributed include the unique position of the Primary Children’s Hospital as the sole referral and speciality care center in the state, which may simplify the ascertainment of individuals with Turner syndrome.

We found that 75.4% of Turner syndrome cases had karyotype 45,X, a proportion that is consistent with a retrospective French study of prenatally diagnosed Turner syndrome (79.6% with 45,X; Gruchy et al., 2014). However, this proportion is higher than clinical estimates (Gravholt et al., 2017). This discrepancy is likely due to a delay in diagnosis, which is more common among individuals with mosaic karyotypes, who often present with milder phenotypes compared to individuals with 45,X (Stochholm et al., 2006).

Like previous assessments (Gravholt et al., 2017), coarctation of the aorta, atrial septal defect, and ventricular septal defect were among the most common anomalies. The proportion of cases with bicuspid aortic valve (3.5%) or horseshoe kidney (3.4%), two other common anomalies associated with Turner syndrome, were lower than previous estimates of 14%–34% and 10%, respectively. The challenges of diagnosing these defects prenatally and at birth (Silberbach et al., 2018; Strauss et al., 2000) may explain this discrepancy. Notably, we restricted this study’s collection of major birth defects to the NBDPN “core” birth defects (National Birth Defects Prevention Network, 2021) for consistent data collection across the seven surveillance programs. However, further research should address this limitation by including population-based assessments of the proportion of cases with other birth defects that are observed within the Turner syndrome phenotype.

We found that the prevalence of Turner syndrome did not significantly differ by maternal age. Findings that conclude that advanced maternal age (≥40 years) is associated with Turner syndrome (Hagman et al., 2010) are likely a reflection of incidental diagnosis due to increased screening for other birth defects in older women (Papp et al., 2006; Saenger et al., 2001). Others have found that the majority of Turner syndrome cases retain the maternal X chromosome and that loss or abnormality of the second X chromosome is due to paternal meiotic errors (Uematsu et al., 2002), which could be part of the explanation for why Turner syndrome is not thought to be associated with advanced maternal age (Gunther et al., 2004). Although this study did not collect paternal age data, future studies should consider evaluating the effect of paternal age on Turner syndrome prevalence. Such investigation could support hypotheses related to the paternal mechanisms underlying Turner syndrome development (Huang et al., 2021).

Few studies have described Turner syndrome among racial and ethnic groups in the United States. In a study of 81 individuals with Turner syndrome in North Carolina, 86.4% identified as White and 13.6% identified as African American, American Indian, or Asian. Of the individuals who were diagnosed prenatally, all were White (Savendahl & Davenport, 2000). We found that the prevalence of Turner syndrome was significantly lower for individuals born to women who were Hispanic, non-Hispanic American Indian, Alaska Native, Asian, Pacific Islander, or non-Hispanic Black compared to non-Hispanic White. Importantly, our findings could be explained by differences in case ascertainment by race and ethnicity. Therefore, surveillance programs should consider evaluating and improving current ascertainment strategies by maternal race and ethnicity to enhance ascertainment of birth defects in these populations.

We also hypothesize that these findings could be a result of structural racism in healthcare, including lack of access to and utilization of prenatal testing and care, lower likelihood of affordable care, and distrust in the healthcare system, all of which are known to disproportionately affect Black women (Gadson et al., 2017) and some of which have also been shown to impact attitudes toward prenatal testing for Down syndrome in Black women (Kuppermann et al., 2006; Learman et al., 2003). An assessment of the association of Turner syndrome prevalence with structural racism and the effects of structural racism might identify factors related to inequitable access to maternal and infant care, which could also be influencing case ascertainment. Findings from such studies could inform diagnostic protocols, medical management, and engagement strategies, especially for the populations that we identified to be at highest risk in this study: Hispanic, non-Hispanic American Indian, Alaska Native, Asian, or Pacific Islander, and non-Hispanic Black populations.

We found that survival for Turner syndrome was greater than 90% from birth to 5 years. Additionally, previous studies have shown that individuals with HLH-associated Turner syndrome have a significantly greater risk of death compared to those without HLH (Lara et al., 2017; Reis et al., 1999). Our analysis showed that Turner syndrome without HLH had survival probabilities greater than 90% at all-time points that we analyzed. This information is important for counseling parents about survival when HLH is not a part of the Turner syndrome phenotype.