Publications were included for review if they met the following inclusion criteria: English language, full-length articles, human subjects aged 0 to 3 years, comparative study of at least one transcatheter option (DS or RVOT intervention) against SPS as a control group, and described at least one outcome of interest.

Publications which were not a comparative study were excluded (such as case reports, case series, topical reviews, and opinion pieces), as were abstracts, oral presentations, and conference proceedings without a full-length article. While systematic reviews and meta-analyses were excluded from the analysis itself, their references were manually searched for possible source of relevant publications.

A comprehensive literature search of MEDLINE and EMBASE databases was conducted. All publications from January 2003 to December 2023 inclusive were included. Additionally, the Cochrane Database of Systematic Reviews was searched for relevant systematic reviews.

The search strategy was developed by two reviewers (CQZ and SB) in consensus. The search syntax for MEDLINE and EMBASE were constructed to replicate the search logic as similarly as possible. The complete search syntax may be found in Table 1.

The following search terms were used as free text, or where available, equivalent standard MeSH and Emtree terms: human; paediatric or children; ductus arteriosus and stent; right ventricular outflow tract, and stent or balloon; cyanotic congenital heart disease or Tetralogy of Fallot; Blalock-Taussig shunt or systemic-to-pulmonary shunt.

Two reviewers (CQZ and SB) screened titles and abstracts of the search results for publications relevant to study question. Where the eligibility of the publication was unclear from the title or abstract, the full text was retrieved for screening. Duplicate publications were removed, and data from included publications were extracted for further analysis.

The following outcomes were sought from each publication: Overall and Early Mortality, Overall and Early Re-intervention rate, Procedural Complications, Intra-procedural Mortality, time to second-stage surgery, length of overall hospitalization, and length of stay in intensive care unit (ICU).

As the exact criteria for defining the “Early” period varied across publications, for the purpose of our review, mortality and re-intervention outcome was considered “Early” if it occurred: within 30 days of the initial intervention, within the same admission, or considered by study authors to be directly related to the initial procedure. “Overall” referred to any instance of the outcome reported by the publication, regardless of duration of follow-up. Procedural Complications referred to complications relating to the initial procedure, as reported by each publication. Time to second-stage surgery or definitive surgery was recorded as the same outcome.

Where mean and standard deviation of study groups were not directly reported, these statistics were estimated from the statistics reported by each publication, using the Data Estimation and Conversion for Meta-Analysis (DECoMA) tool (Cochrane Taiwan) (9, 10). Meta-analysis and network meta-analysis were conducted with Stata 18.0 software package (StataCorp). Two-tailed alpha was set at 0.05. Adjusted point estimates were pooled using the generic inverse variance method under the fixed-effect model.

Network meta-analysis allowed simultaneous comparison of all three interventions by direct and indirect evidence. The mean rank and surface under the cumulative ranking (SUCRA) of each intervention was calculated for each outcome of interest. Local incoherence was tested by node-splitting, while global incoherence was tested using Design-by-Treatment model to check for the entire network. The overall heterogeneity was assessed using the Chi-square test, with a threshold of p < 0.1 for significant heterogeneity. Papers that might cause significant heterogeneity were identified and removed from the network meta-analysis.

Quality appraisal of each included publication was performed by two reviewers (CQZ and SB) with the relevant risk-of-bias tool (Newcastle-Ottawa Scale was used for retrospective or prospective cohorts) (11).

Non-consensus between the two reviewers was resolved by a third senior author (CCK).

Total of 337 publications were identified from initial search of each database, and 30 duplicate publications removed. From the remaining, 270 publications were excluded during the initial screening of titles and abstracts for relevance. Of 37 left, the following exclusions were made: 2 (non-English language), 6 (not full-length articles), 1 (non-comparative study), 8 (study groups and outcomes were not relevant). Furthermore, 2 studies were identified as a subgroup analysis of a larger study already being included, and hence were removed from analysis (12, 13).

Additionally, 40 publications were identified for screening from the references of relevant systematic reviews (2–7),—from which 2 publications were eventually identified as eligible for inclusion. In effect, a total of 20 publications were selected for analysis (14–33). The PRISMA flow diagram is reported in Figure 1.

The basic characteristics of each included publication are detailed in Table 2 (14–33). Among the 20 selected publications, 15 publications compared DS against SPS and 4 studies compared RVOT intervention against SPS. A single study compared all three modalities of treatment.

A total of 4,441 patients were analysed across all publications. Patients underwent intervention from 0 to 11 months of age. Surgical SPS included not only the most common modified Blalock-Taussig-Thomas shunt, but also central aorto-pulmonary shunts. All RVOT interventions were performed via RVOT stenting. Among patients with a reported cardiac diagnosis, 352 of 3,694 patients were classified as Tetralogy of Fallot. The baseline characteristics of the included patients are presented in Table 3.

Out of the 20 publications, all reported data for Overall Mortality, while 17 reported data for Early Mortality. Overall Re-intervention data was reported in 16 publications, while 6 reported data for Early Re-intervention. Procedural Complication rates was recorded from 10 publications, and Intra-procedural Mortality from 11 publications.

Network meta-analysis was performed for indirect comparisons and ranking of treatment modalities. There was sufficient data for network meta-analysis for the outcomes of Overall Mortality, Early Mortality, Intra-procedural Mortality, Procedural Complications, Early Re-interventions, and Overall Re-interventions. All network meta-analyses were closed loops, with at least one direct comparison between each treatment modality.

The results of risk of bias assessment done using Newcastle-Ottawa Scale (11), are presented in Table 4.

By network meta-analysis, RVOT intervention was ranked as best modality for the outcome of Early Mortality (Figure 2), and DS was best for Overall Mortality (Figure 3). Surgical SPS was ranked the poorest modality for both Early and Overall Mortality. However, surgical SPS was the best-ranked modality for Intra-procedural Mortality (mean rank 1.5; cf. DS 2.2 and RVOT 2.3). There was no statistically significant differences between pairwise comparisons of each treatment modality for both Early Mortality and Overall Mortality (Figuress 2,3).

DS is the best-ranked modality for Procedural Complications (Mean SUCRA = 92.5, Mean Rank = 1.2), by network meta-analysis (Figure 4). RVOT intervention was worst-ranked for Procedural Complications (Mean SUCRA = 13.2, Men Rank = 2.7). Surgical SPS was worst-ranked by network meta-analysis for Early Re-intervention rate (Mean SUCRA = 11.3, Mean Rank = 2.8) (Figure 5), yet was best-ranked for Overall Re-intervention rate (Mean SUCRA = 95.8, Mean Rank = 1.1) (Figure 6).

There was no difference in time to second-stage surgery (or definitive repair) between DS or surgical SPS.

Patients who underwent DS compared to surgical SPS as initial treatment had a shorter ICU stay [mean difference −0.60 (−1.06, −0.15)] (Figure 7). Both DS and RVOT intervention conferred a shorter overall hospitalization duration compared to surgical SPS, with results slightly favouring RVOT intervention over DS [RVOT vs. SPS mean difference −0.54 [−0.88 to −0.20]; DS vs. SPS mean difference −0.49 [−0.62 to −0.36]] (Figure 7).

There was insufficient data to perform meta-analysis of RVOT intervention against SPS for time to second-stage surgery and length of ICU stay. Due to small sample sizes, network meta-analysis of secondary outcomes was not performed.

Funnel plots for pairwise outcomes of Early Mortality, Procedural Complications, and Early Re-intervention did not demonstrate significant asymmetry (Figure 8). These were chosen as they were deemed the most directly consequent adverse outcomes of each intervention. Funnel plots of RVOT vs. SPS was not shown as there was only a single publication with this respective direct comparison. Egger regression-based testing confirmed no statistically significant asymmetry, suggesting the absence of publication bias.

Due to the rarity of cyanotic congenital heart disease, almost all included studies were retrospective cohorts. There were no randomised studies comparing one intervention against another, and therefore the outcomes are subject to selection bias. Among studies, there remained ambiguity in the definition of “complications”, making this outcome subject to each study's interpretation. There was also significant variability in follow-up durations, and lost-to-follow-up rates were generally unreported, introducing further biases that are difficult to address through statistical adjustments alone.

The greatest limitation is the scarcity of comparative data between RVOT intervention compared to other interventions. In published literature, only 4 studies directly compared RVOT intervention with SPS, and only 1 study performed a three-way comparison. We have attempted to mitigate this limitation in our study by use of network meta-analysis and indirect comparisons.

The patient population selected for RVOT intervention was generally older than the patients selected for DS (Table 3), which limits the comparability of outcomes between the two transcatheter intervention groups.

Innovative advances in interventional options for cyanotic congenital heart disease are promising, and continue to offer exciting alternatives to treatment. However, the choice of initial intervention remains highly dependent on the training and experience of each centre, as well as patient-specific factors, such as the native anatomy. Tailoring the intervention to both the institution's expertise and the unique anatomical needs of each patient is key to optimizing outcomes.

This study lends further credibility to transcatheter options as viable, safe, and attractive alternatives to traditional surgical shunts in the treatment of duct-dependent cyanotic heart disease. Our findings suggest that, with continued experience and refinement, especially in techniques like RVOT stenting, transcatheter approaches could see wider adoption and potentially lead to further improvements in patient outcomes. Expanding the use of these less-invasive methods may also help build collective expertise, paving the way for optimized care in this complex patient population.