We performed a systematic review with meta-analysis following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for healthcare intervention evaluations [7].

We included all published studies that reported adult human subjects with aortic stenosis or aortic valve evaluation with data on aortic valve area by 2D TEE and 3D planimetry. Studies were eligible if they reported AVA measured by both 2D-TEE planimetry and 3D-TEE planimetry in the same subjects (paired measurements), or at least present group means with standard deviation (SD) and sample sizes for both modalities. We included original research (either prospective or retrospective). Primary outcomes: mean differences (2D − 3D) in AVA (cm²), secondary outcomes: pooled correlation coefficient (r) between 2D and 3D AVA; subgroup analyses (equipment or software vendor, year of study). We excluded case reports, editorials, reviews, studies using only transthoracic echocardiography (TTE), computed tomography (CT) or magnetic resonance imaging (MRI) without TEE planimetry comparison (unless they also included TEE 2D and 3D). A thorough search was performed across MEDLINE(PubMed), Cochrane Central Register, Web of Science and Scopus. We examined all the records retrieved from inception until 8th October 2025. Our search strategy is provided in supplementary files. The search was not restricted by language or date.

For study selection, a two-stage review process was employed: firstly, an independent screening process of titles and abstracts was performed by two reviewers. Subsequently, the full texts of potentially relevant studies were examined for inclusion or exclusion. Disagreements were resolved by consensus. Extracted variables included study ID (author, year), country, design, sample size, patient characteristics (age, sex), valve morphology (tricuspid/bicuspid), TEE equipment and software, AVA by the continuity equation (if reported), AVA means and SDs for 2D and 3D (ideally paired mean difference and SD of difference); correlation between modalities if reported.

Risk of bias was assessed using the QUADAS-2 tool [8]. The methodological quality of the included studies was evaluated using the QUADAS-2 tool across four domains: patient selection, index tests (2D and 3D TEE), reference standard (transthoracic echocardiography), and flow and timing. Each study was rated as having a low, high, or unclear (“some concerns”) risk of bias. Discrepancies were resolved by consensus. A “traffic-light” diagram was created using the Robvis tool [9].

The Meta-analyses were conducted using IBM SPSS Statistics 28 for mean differences and RStudio for meta-analysis of correlations using z Fischer transformation. The output included forest plots, showing both individual study outcomes and combined results. In the meta-analysis, we included continuous outcomes, reporting effect sizes as absolute mean differences between the two echocardiographic techniques, Pearson correlation coefficients (r) between the two techniques and the corresponding sample size (N). A random-effects model was employed to account for different technical aspects and the wide time range across published studies. Heterogeneity was quantified using the I2 percentage with values of > 50% considered with substantial heterogeneity. Sensitivity analyses were performed to assess the impact of potential confounders, including baseline characteristics and different equipment used.

The overall pooled effect size of correlations and its 95% confidence interval were calculated on the z-scale. For enhanced interpretability and ease of understanding, the pooled estimate and its confidence intervals were then back-transformed to the original correlation coefficient r scale for presentation in the forest plot.

Data Handling for paired data: when studies reported only means and standard deviations (SDs) for each modality, we estimated the standard deviation of the mean difference (SD_diff) using the following formula:

$$\text\_\text= \text+\text - 2 * \text*\text1*\text2).$$

Where SD₁ and SD₂ are the standard deviations for each modality and r is the correlation coefficient between the paired measurements. If the correlation coefficient was not provided, we used a conservative assumption of r = 0.5 based on typical values reported in similar studies.

For studies reporting unpaired data (i.e., separate groups for each modality), we directly used the reported means and SDs for each group. To pool data across studies, we applied random-effects models for meta-analysis, assuming that the variability between studies could not be entirely explained by sampling error.

Publication bias was assessed through visual inspection of funnel plots. In line with methodological recommendations, statistical tests for funnel plot asymmetry (Egger’s test) were interpreted with caution because fewer than 10 studies were included.