Abstract
Objective
To evaluate whether temporal changes in prenatal detection and North Carolina abortion rates are associated with live-birth cleft palate (CP ± L) severity phenotypes.
Design
Retrospective cohort study linking patient-level clinical data to annual state abortion rates. Analyses included Pearson correlation, multivariable logistic, and interrupted time-series modeling.
Setting
Single-center craniofacial program, with linkage to annual statewide abortion rates as a population-level contextual variable.
Patients, Participants
554 patients undergoing primary palatoplasty (1999 to 2021).
Interventions
None.
Main Outcome Measure(S)
Primary outcomes: annual Veau classification distribution. Secondary outcomes: prenatal diagnosis rates and genetic syndrome or sequence rates.
Results
From 2004 to 2021, the proportion of Veau I cases increased from 18.2% to 25.0% (slope β= + 0.0074 proportion-units/year; P = .051), the proportion of Veau III cases declined significantly (slope β= −0.0085 proportion-units/year; P = .011), and the proportion of Veau IV cases trended downward non-significantly. Higher abortion rates correlated with greater Veau III proportions (r = 0.63, P = .005). Logistic regression demonstrated that each 1-SD increase in abortion rate raised severe-cleft odds by 26% (odds ratio (OR) 1.26; 95% CI 1.05-1.51; P = .014), prenatal diagnosis (OR 13.13; P < .001) and presence of genetic syndrome or sequence (OR 0.36; P < .001) were also significant. Cleft severity decreased prior to 2010 (β = –0.11 units/year; P = .036) and then plateaued. Prenatal diagnosis rates increased sevenfold during the study period.
Conclusions
Enhanced prenatal detection and abortion rate shifts had minimal impact on live-birth CL/P trends. The paradoxical association between higher abortion rates and severe clefts suggests that reproductive decision-making is influenced by multifactorial factors beyond detection.
Keywords
Introduction
Cleft lip and/or palate (CL/P) is one of the most prevalent congenital anomalies worldwide, with important implications for feeding, speech, and long-term craniofacial and dental development.1–5 Recent population-based data demonstrate that while prenatal detection rates for cleft lip with or without cleft palate (CL/P) are relatively high—approaching 85% in some cohorts—detection of isolated cleft palate remains markedly lower, with sensitivities as low as 4.4%, underscoring persistent limitations in current imaging modalities.6–9 Nevertheless, advances in prenatal imaging technology over the last 20 years have resulted in improved accuracy and earlier detection, especially in cases of embryonic clefting with syndromic features or co-existing anomalies.8,10,11
The increased accuracy of prenatal diagnosis has introduced complex ethical and clinical considerations, including decisions regarding pregnancy continuation. 12 In some regions, the option of pregnancy termination following the diagnosis of CL/P has influenced the reported prevalence and severity distribution of these anomalies at birth. 12 For instance, studies have observed a decline in the incidence of severe cleft cases over time, coinciding with enhanced prenatal screening and the availability of termination services. 13
Given the increased ability for prenatal detection coupled with new considerations of access to abortion, there are opportunities to consider how these factors might influence the epidemiology of congenital anomalies, including the phenotypic patterns of CL/P live births. While previous studies have investigated trends in prenatal diagnosis and the psychosocial outcomes of early diagnosis, there is a relative paucity of work investigating how improved prenatal detection and increased access to elective abortion may impact the distribution of CL/P phenotypes over time. In particular, it remains unclear if the increase in access to abortion or improved screening methods leading to detection of CL/P has created a measurable change in the severity of CL/P seen postnatally. We assessed temporal trends in the severity of CL/P (using the Veau classification), prenatal diagnosis, and associated clinical diagnoses in relation to annual abortion rates. 14
Methods
Data Collection
This study was approved by the Institutional Review Board. Patients who underwent primary palatoplasty between 1999 and 2021 were identified using the i2b2 platform via the Carolina Data Warehouse for Health (CDWH), an institutional enterprise clinical data repository that aggregates electronic medical record data across health system-affiliated hospitals and outpatient clinics. As such, CDWH captures patients treated within the study institution's health system and does not include cases managed at other cleft centers in the state. 15 Health electronic medical record. Demographic and clinical data were extracted from the electronic medical record and de-identified prior to analysis. Annual induced termination of pregnancy (abortion) rates for North Carolina was obtained from the Department of Health and Human Services. 16 Data were managed using REDCap on the institution's secure server.
Cohort Definition
The study cohort included all patients who underwent primary palatoplasty during the defined study period, identified by current procedural terminology codes. Patients were excluded if they underwent primary palatoplasty at an outside hospital, if the surgical technique was incompletely or not documented, if they had submucous clefts treated for velopharyngeal insufficiency without primary palatoplasty, if the procedure was aborted intraoperatively, or if the palatal defect was iatrogenic. Only cases with complete documentation of cleft type and surgical approach were included in the final analysis.
Outcomes
The primary outcome was the annual proportion of Veau classifications (I-IV) relative to total cleft cases. This was examined in relation to annual state abortion rates to evaluate temporal trends and ecological associations within the institutional cohort. Statewide abortion rates were included as a population-level exposure and were not used to estimate statewide cleft incidence, prevalence, or treatment volume.
Secondary outcomes included the proportion of patients with a prenatal diagnosis of cleft palate and/or cleft lip (stratified by cleft type and gestational age at diagnosis) and the proportion of a confirmed genetic condition or sequence, such as Pierre Robin Sequence or 22q11.2 Deletion Syndrome.
Additional variables collected included type of palatoplasty performed (eg, Furlow, Bardach, Von Langenbeck), gestational age at birth (categorized as very preterm (<32 weeks), moderate-to-late preterm (32-37 weeks), term (37-42 weeks), or post-term (>42 weeks)), sex, race, ethnicity, requirement for tube feeding at birth, and age at the time of primary palatoplasty.
Statistical Methods
Analyses were conducted in Python (v3.10) using the statsmodels and scikit-learn package. Categorical variables were summarized as frequencies and percentages; continuous variables were reported as means ± SD or medians with interquartile ranges, as appropriate. Missing data due to adoption or unavailable clinical information were excluded from respective analyses; all other missing values were categorized as “unknown” and not imputed. All statistical analyses involving prenatal diagnosis were limited to cases in which cleft palate (as opposed to isolated cleft lip) was identified on prenatal imaging. Because cleft palate is more difficult to identify prenatally than cleft lip, temporal improvements in prenatal recognition of palate involvement may better reflect meaningful advances in ultrasound-based phenotypic characterization than detection of cleft lip alone.
Due to small sample sizes before 2004 and the absence of abortion data prior to 2003, patients born before 2004 were excluded from temporal trend analyses. Nonparametric locally weighted scatterplot smoothing (LOESS) with a span of 0.3 and 95% pointwise confidence intervals (from the residual SD) was applied to the annual proportions. Ordinary least-square regressions of proportion versus calendar year were then fitted for each Veau class to extract the slope, two-sided p-value, and R2 (α = 0.05).
Annual abortion rates were standardized as z-scores, and yearly proportions of Veau classifications, associated conditions, and prenatal diagnoses were computed and visualized using line plots. Pearson correlation coefficients with 95% confidence intervals (via Fisher's z-transformation) were calculated to assess the relationship between abortion rates and each of these variables.
A binary logistic regression was used to predict the probability of Veau III/IV cleft using three predictors: prenatal diagnosis (binary), presence of a genetic condition or sequence (binary), and state-level abortion rate (standardized, continuous). Odds ratios (ORs) and confidence intervals were visualized using a forest plot.
An interrupted time series (ITS) analysis was performed to evaluate changes in prenatal diagnosis rates and cleft severity, using 2010 as a prespecified breakpoint based on a visual inflection point in the data. For prenatal diagnosis, the outcome was whether each case was diagnosed prenatally (yes/no). A logistic regression model was fitted to individual-level data, incorporating terms for time (in years), a post-2010 indicator, and their interaction to estimate changes in slope and level. Predicted probabilities of prenatal diagnosis were generated for each case and averaged annually to visualize modeled trends over time. For cleft severity, Veau classification was treated as an ordered categorical variable (I-IV), and each case was assigned a single ordinal value. A cumulative logit model was used to perform an ordinal logistic ITS, with the same time-based terms. The model's predicted probabilities for each Veau category were used to calculate the expected mean Veau classification per year.
Results
A total of 554 patients underwent primary palatoplasty and were included in the cohort. Most were born at term (71.7%) and were male (53.3%), with the predominant racial groups being White (58.1%) and ethnic group Hispanic/Latino (78.2%). Confirmed genetic condition or sequence was present in 26.0% of patients, most commonly Pierre Robin Sequence. Tube feeding at birth was required in 24.9%. Any orofacial cleft prenatal diagnosis was documented in 112 patients (20.2%), with a median gestational age at diagnosis of 20.1 weeks. Among prenatally diagnosed cases, 34 (30.4%) were identified as lip only, 68 (60.7%) had cleft palate with or without cleft lip identified prenatally, and 10 (8.9%) had unclear documentation regarding the specific defect detected. Most patients underwent either Bardach (56.9%) or Furlow (27.3%) palatoplasty, at a median age of 44.0 weeks (Table 1).
Demographic and Clinical Characteristics of Patients Undergoing Primary Palatoplasty (Mean ± Std Dev & Median (Q1-Q3)).
Across the 2004-2021 study period, Veau II remained the most prevalent subtype, representing a mean of 32.9% of cases each year and rising only from 27.3% to 30.0% overall (+2.7%; β = + 0.0047 year−1, P = .215; R2 = 0.09). The proportion of Veau I cases increased from 18.2% in 2004 to 25.0% in 2021 (mean 20.4%; absolute change +6.8%), peaking at 33.3% in 2019, with an annual slope of +0.0074 proportion-units/year (P = .051; R2 = 0.22). In contrast, Veau III exhibited a statistically significant decline, falling from 31.8% to 30.0% over the interval (β = –0.0085 year−1, P = .011). Veau IV, the least common classification (mean 16.5%), decreased from 22.7% to 15.0%, although this downward trend did not reach statistical significance (β = –0.0036 year−1, P = .37). More severe clefts (Veau III and IV) attained their highest annual proportions in 2005 (47.1% and 35.3%, respectively). Nonparametric LOESS-smoothed trajectories also illustrated a gradual shift toward milder cleft phenotypes over the 18-year span with a modest rise in proportions of Veau I, relative stability in Veau II, a gradual decline in Veau III, and minimal change in Veau IV (Figure 1).
Annual TRENDS in Veau classification proportions (2004-2021) with LOESS smoothing and linear fits. LOESS, locally weighted scatterplot smoothing.
Across the 18-year observation window, abortion rates averaged 12.63 ± 1.86 per 1000 women (range: 10.0-16.0). Cases with a genetic condition or sequence varied from 5.9% to 43.3% (mean 27.1% ± 12.9%), while prenatal diagnosis rates ranged from 4.2% to 36.8% (mean 16.4% ± 8.4%). Prenatal diagnosis increased over time, peaking in 2021 (36.8%), whereas syndromic diagnoses peaked in 2010 (Figure 2).
Temporal trends in standardized abortion rate and primary palatoplasty case characteristics (2004-2021).
Pearson correlation revealed a strong positive correlation between standardized abortion rate and the proportion of Veau III clefts (r = 0.63, P = .005, 95% CI [0.24, 0.85]). Weaker, non-significant trends were observed for Veau I (r = −0.43, P = .077, 95% CI [−0.75, 0.05]), Veau II (r = −0.37, P = .130, 95% CI [−0.71, 0.12]), and Veau IV (r = 0.20, P = .422, 95% CI [−0.29, 0.61]). Additionally, there were no significant associations between abortion rates and genetic condition or sequence cases (r = −0.21, P = .39, 95% CI [−0.62, 0.28]) or prenatal diagnosis (r = 0.10, P = .70, 95% CI [−0.39, 0.54]).
The logistic regression analysis revealed that abortion rate was a significant variable of severe cleft (Veau III/IV), with each standard deviation increase associated with 26% higher odds of a severe cleft diagnosis (OR = 1.26, 95% CI: 1.05-1.51, P = .014). Prenatal diagnosis was strongly associated with severe cleft (OR = 13.13, 95% CI: 6.06-28.46, P < .0001). In contrast, the presence of a genetic condition or clinical sequence was associated with lower odds of severe cleft (OR = 0.36, 95% CI: 0.23-0.57, P < .0001). The model showed moderate explanatory power (McFadden's R2 = 0.127) and discriminative ability (AUC = 0.714) (Figure 3).
Forest plot of adjusted odds ratios for abortion rate, prenatal diagnosis, and genetic condition/sequence.
ITS modeling demonstrated a significant pre-2010 decline in cleft severity, with a 17.7% annual reduction in the odds of presenting with a more severe phenotype (OR = 0.823; 95% CI: 0.700-0.968; P = .019), and a corresponding drop in the expected mean Veau class from 2.69 to 2.17. Although there was a non-significant immediate decrease at the 2010 breakpoint (OR = 0.465; 95% CI: 0.199-1.087; P = .077), severity subsequently increased at a significant rate (OR = 1.202; 95% CI: 1.013-1.426; P = .035), with expected mean Veau class rising from 2.23 to 2.37 between 2010 and 2021 (Figure 4).
Interrupted time series (2004-2021) of probability-weighted expected Veau class means from ordinal logistic regression.
In parallel, logistic regression for prenatal diagnosis showed no significant trend prior to 2010 (OR = 0.886; 95% CI: 0.647-1.214; P = .452), with predicted probabilities declining from 11.7% to 7.5%. A non-significant immediate drop was observed in 2010 (OR = 0.320; 95% CI: 0.073-1.407; P = .132; predicted probability: 6.2%), followed by a gradual but non-significant increase through 2021 (OR = 1.275; 95% CI: 0.920-1.768; P = .144), reaching a predicted rate of 13.6% (Figure 5).
Interrupted time series (2004-2021) of predicted probability of prenatal diagnosis from logistic regression.
Discussion
This study contributes to the limited United States literature analyzing population-level trends in cleft severity within the context of changing abortion rates and prenatal diagnostic capabilities over a two-decade period. Among 554 patients who underwent primary palatoplasty, we found a paradoxical association: higher abortion rates were linked to increased odds of severe cleft phenotypes (Veau III/IV). Prenatal diagnosis was strongly associated with cleft severity, while the presence of a genetic condition or sequence was inversely associated with a Veau III or IV cleft. Cleft severity declined markedly prior to 2010 and then stabilized, suggesting a temporal inflection likely shaped by changes in reproductive decision-making, diagnostic technologies, and policy environments.
The nearly sevenfold rise in prenatal diagnosis rates—from 4.2% in 2012 to 36.8% in 2021—reflects substantial improvements in detection capabilities. The transition from 2D to high-resolution 3D and 4D ultrasonography has enhanced fetal facial visualization, improving identification of clefts. 17 While 2D ultrasound remains the standard initial modality, it has notable limitations in detecting cleft palate, especially when isolated. 18 The introduction of 3D ultrasound has markedly improved visualization of fetal facial structures, offering more accurate surface rendering and better detection rates of cleft palate compared to 2D alone.19,20 Moreover, when clefts are suspected, supplemental imaging—such as 3D or fetal MRI—can enhance diagnostic precision. 8 However, diagnostic accuracy remains influenced by gestational age, fetal position, and maternal characteristics, underscoring the continued importance of skilled sonographers and access to advanced imaging modalities. 21
Paradoxically, we found higher rates of severe clefts during years with higher overall abortion rates. This may reflect the multifactorial nature of termination decisions, which are shaped by cultural, socioeconomic, and legal factors beyond anomaly detection alone. 22 State-level abortion data are inherently limited by underreporting and the inability to capture out-of-state procedures, particularly in restrictive environments such as North Carolina.16,23 In California, where abortion access is broad, termination rates for cleft lip and palate remain low and are likely even lower in more restrictive settings like North Carolina. 24 Thus, aggregate abortion rates are a poor proxy for anomaly-specific decisions, complicating the interpretation of these trends. A more likely explanation involves syndromic associations: up to 50% of isolated cleft palates are syndromic, compared to fewer than 15% of cleft palates with cleft lip. 25 Syndromic cases linked to Veau I and II may be more readily identified and thus result in termination, especially when accompanied by other anomalies, whereas non-syndromic Veau III & IV clefts may be more likely to remain in the live-birth cohort.
The lack of correlation between cleft severity and abortion rates also likely reflects the effectiveness of prenatal counseling provided by cleft care teams. As shown in Matthews’ study, the vast majority of plastic surgeons engage in prenatal consultations primarily to educate families about cleft conditions, discussing treatment options, expected outcomes, and long-term care. Termination was rarely addressed, and most surgeons reported that families sought information to prepare for the birth and care of their child, rather than to consider pregnancy termination. 26 Society for Maternal-Fetal Medicine guidance further recommends a detailed evaluation for associated anomalies, offering diagnostic testing with chromosomal microarray, consideration of fetal echocardiography or magnetic resonance imaging when indicated, and referral for prenatal counseling with pediatric plastic surgery, oral and maxillofacial surgery, or a craniofacial clinic. Counseling should also address prognosis, newborn feeding, surgical repair, and postnatal genetic evaluation. 27 These findings suggest that consistent, informative counseling helps families better understand the treatable nature of cleft lip and palate, supporting informed and confident decision-making during a vulnerable time.
The inflection in cleft severity and prenatal diagnosis rates around 2010 likely reflects intersecting policy, technological, and societal shifts that influenced detection and reproductive decision-making. The Affordable Care Act, enacted in 2010, expanded insurance coverage and designated prenatal care as an essential health benefit, improving access to ultrasounds. 28 Around the same time, widespread adoption of high-resolution 3D/4D ultrasound—including GE's HDlive system in 2011—enhanced fetal facial imaging.29,30 The early 2010s also marked the introduction of non-invasive prenatal testing, reflecting a broader shift toward comprehensive prenatal screening. 31 However, these improvements occurred alongside rising state-level abortion restrictions, including North Carolina's 2011 “Women's Right to Know Act,” which may have limited termination access despite improved detection. 32 Together, these trends suggest that while diagnostic capability expanded post-2010, the cohort of live births with severe clefts may have reached a lower bound, constrained by legal and cultural barriers to termination.
Interpretation of these findings requires attention to the timing of North Carolina abortion law. Importantly, the current North Carolina law limiting abortion after 12 weeks of pregnancy did not take effect until July 1, 2023 and therefore falls outside the study period. During the years analyzed here, North Carolina law generally permitted abortion through the first 20 weeks of pregnancy for any reason and after 20 weeks in the case of medical emergency. 33 Because a detailed fetal anatomic ultrasonography is typically performed at 18-22 weeks of gestation, prenatal cleft diagnosis during most of the study period would often have occurred near this legal threshold rather than clearly before it. 34 For isolated cleft palate, which is less reliably detected prenatally, diagnosis may occur even later and further constrain anomaly-specific decision-making.
Our findings differ from international trends that have demonstrated sharper declines in severe cleft phenotypes in settings with varying cultural, legal, and moral attitudes toward pregnancy termination following prenatal diagnosis. In Israel, for example, Emodi et al reported a 60% reduction in severe cleft palate cases over a decade, coinciding with high access to advanced prenatal screening and a healthcare system that facilitates termination for fetal anomalies under medical committee approval—reflecting cultural norms more accepting of selective termination. 13 Similarly, in Taiwan, where abortion for congenital anomalies is legally permitted and culturally accepted, Chang et al observed a marked decline in cleft lip with or without palate following the implementation of universal health coverage and expanded prenatal ultrasound. 35 In contrast, the U.S. context is characterized by state-level heterogeneity in abortion policy, complicating the downstream effects of improved prenatal detection on birth prevalence.
This study's strengths include a large, well-characterized clinical cohort evaluated over two decades at a high-volume craniofacial center. The integration of clinical data with population-level abortion statistics provides a novel lens to explore how shifts in reproductive policy and diagnostic access may influence cleft phenotypes in the live-birth population.
Because this cohort was derived from the study institution's health system, our study reflects a single-center tertiary referral experience rather than a statewide cleft birth cohort. Other craniofacial centers in the state were not captured in CDWH. Although a statewide birth defects registry could provide broader population-level case ascertainment, such sources rely on administrative diagnostic codes and lack the patient-level phenotypic resolution required for this analysis, including surgically confirmed Veau classification, prenatal diagnosis characteristics, associated genetic conditions or sequences, and operative details. Accordingly, findings may be influenced by institutional referral patterns, case mix, and access to care, and should be interpreted as an ecological analysis linking a single-center surgical cohort with statewide abortion statistics rather than as a population-based analysis of all cleft cases in the state. Other limitations include underreporting of abortion data, particularly from 2011 to 2014, as acknowledged by the state, and lack of anomaly-specific granularity. 16 In addition, the strong association between prenatal diagnosis and severe cleft phenotype should be interpreted cautiously. Because prenatal ultrasound detection is substantially higher for clefts involving the lip than for isolated cleft palate, the prenatal diagnosis variable is inherently linked to cleft phenotype and likely reflects differential detectability rather than an independent determinant of severity. This introduces ascertainment bias and limits causal interpretation of the corresponding OR in the logistic model. The retrospective design introduces potential for documentation bias, and the use of ecological abortion data prevents patient-level linkage. Finally, reliance on the Veau classification may not fully capture anatomic or functional heterogeneity in cleft severity.
Conclusions and Future Directions
This study reveals a complex interplay between prenatal diagnostic advances, reproductive policy, and cleft phenotype severity over nearly two decades. While cleft severity declined prior to 2010, it has since plateaued or risen modestly, with higher abortion rates and prenatal diagnosis associated with increased odds of severe clefts, and genetic conditions linked to milder presentations. In the post-Roe v. Wade era, access to pregnancy termination is increasingly restricted across many U.S. states, even in the setting of fetal anomalies. 36 At the same time, recent advances in artificial intelligence offer promising tools for improving prenatal detection of cleft lip and palate, particularly in challenging cases like isolated cleft palate, by reducing inter-operator variability and enhancing diagnostic precision. 37 Future research should examine how these technological and policy shifts affect the phenotypic distribution of craniofacial anomalies, ensuring that innovations in prenatal care are evaluated not only for their clinical utility but also for their broader individual, familial, and societal implications. These tools should also be used to deliver a clear, detailed information to support informed decision-making within today's dynamic political, legal, and cultural environment. Additional work is needed to better characterize how prenatal cleft diagnoses are managed in contemporary U.S. Maternal-Fetal Medicine practice, as published objective data describing MFM-specific counseling and management pathways remain limited.
Footnotes
Acknowledgments
i2b2 software was used in conducting this study. i2b2 is the flagship tool developed by the i2b2 (Informatics for Integrating Biology and the Bedside) Center, an NIH-funded National Center for Biomedical Computing based at Partners HealthCare System. The i2b2 instance at the University of North Carolina is supported by the National Center for Advancing Translational Sciences (NCATS), National Institutes of Health, through Grant Award Number UL1TR002489. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Ethical Considerations and Informed Consent Statements
This study received ethical approval from the University of North Carolina at Chapel Hill IRB (approval # 22-2668) on October 17, 2022. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008 (5). The IRB waived informed consent for research [45 CFR 46.116 (f)] and waiver of HIPAA authorization [45 CFR 164.512 (i) (2) (ii)] due to minimal risk.
Author Contributions
All authors read and approved the final manuscript. GAL contributed to data curation, formal analysis, writing—original draft, and writing—review and editing. IEN contributed to conceptualization, data curation, writing—original draft, writing—review and editing, supervision, and project administration. CNK contributed to conceptualization, data curation, writing—original draft, writing—review and editing, supervision, and project administration. PNS contributed to conceptualization, data curation, formal analysis, writing—original draft, writing—review and editing, supervision, project administration. NRM contributed to data curation and writing—review and editing. MP contributed to data curation and writing—review and editing. JW contributed to conceptualization, writing—review and editing, supervision, and project administration. KC contributed to funding acquisition, supervision, and project administration.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
Data are available from the corresponding author on reasonable request.