Abstract

Background: Prenatal ultrasound screening for fetal structural anomalies is usually summarised by a single pooled sensitivity. Published syntheses report figures around 46% to 56% for first-trimester screening, alongside heterogeneity statistics indicating that the contributing studies are measuring markedly different things. Methods: Systematic reviews with meta-analysis of prenatal ultrasound accuracy against a postnatal or post-mortem reference standard were eligible. Detection rates were extracted overall and by organ system. The pooled estimate was translated into a two-by-two table at the reported prevalence. Heterogeneity was summarised as reported. No new pooling across organ systems was performed, since the appropriateness of such pooling is the question under examination. Analyses were performed in Python 3. Results: Pooled first-trimester sensitivity for major anomalies was 46.10% (95% CI 36.88-55.46) with I² of 90.1% (86.5-92.4), meaning that approximately 10% of the observed variation between studies is attributable to chance. Detection rates by organ system spanned 89 percentage points: 89.0% for lethal malformations, 88.3% for central nervous system and 84.8% for urinary tract anomalies, against 38.8% for heart and great vessels and 0.0% for genital and for ear, face and neck anomalies in the majority of contributing studies. The pooled figure of 46.1% falls in the gap between these clusters and corresponds to no organ system. For major cardiac anomalies in a non-high-risk population, pooled first-trimester sensitivity was 55.80% (45.87-65.50) with specificity 99.98% and positive predictive value 94.85%. At the reported prevalence of 0.41%, this corresponds to 229 anomalies detected and 181 missed per 100,000 fetuses screened, with 20 false-positive results. Conclusions: A single pooled sensitivity for prenatal anomaly ultrasound averages across organ systems whose true detection rates differ by up to 89% points, and describes no clinical situation that arises in practice. Reporting should be organ-system specific. The favourable specificity and predictive values that accompany the pooled figure are properties of low prevalence rather than of detection, and should not be read as reassurance about the quantity a screening programme exists to maximise.

Keywords: Prenatal ultrasound, Fetal anomaly screening, Diagnostic accuracy, Heterogeneity, Organ-system specific reporting, Congenital heart disease