Comparative efficacy and safety of transcatheter edge-to-edge repair versus tricuspid valve replacement versus optimal medical therapy in moderate-to-severe tricuspid regurgitation: a network meta-analysis of randomized controlled trials Authors: Ihsan Noushad Karuppan Veettil, Anuhya Jamandlamudi, Nebeela Abdul Khader, Tea Shehu, Doris Lacej, Shivangi Jha, Obawole Mayowa, Joshua Zumba, Temirlan Ismailov, Zhanel Almaganbetova, Bertha Gabriela, Lina Hajjar, Sifat Ullah Safi, Muneeb Khawar Published: 2026 ABSTRACT BackgroundSevere tricuspid regurgitation (TR) is associated with substantial morbidity and increased mortality. Transcatheter edge-to-edge repair (TEER) and transcatheter tricuspid valve replacement (TTVR) have emerged as less-invasive options for patients remaining symptomatic despite optimal medical therapy (OMT). We conducted a network meta-analysis of randomized controlled trials (RCTs) to evaluate the efficacy and safety of TEER and TTVR versus OMT.MethodsThree RCTs were included after systematic search of PubMed, Embase, and ScienceDirect (inception to December 2025). A frequentist network meta-analysis (random-effects) was performed in R. A Bayesian NMA with vague priors was conducted in parallel to obtain posterior rank probabilities and SUCRA values. Relative effects were translated into absolute risk differences and number-needed-to-treat (NNT) / number-needed-to-harm (NNH) using pooled OMT baseline event rates, with 95% CIs via parametric bootstrap.ResultsThe network comprised 1,050 patients (star-shaped, OMT common comparator; no direct TEER–TTVR comparison). Neither TEER (RR 0.99, 95% CI 0.56–1.76) nor TTVR (RR 0.85, 95% CI 0.51–1.41) significantly reduced all-cause mortality. Both interventions improved NYHA class ≥ 1 class (TEER RR 1.46, 95% CI 1.30–1.64, NNT 8; TTVR RR 3.28, 95% CI 2.41–4.47, NNT 2), KCCQ-OS (TEER MD + 11.00, 95% CI 7.46–14.54; TTVR MD + 17.80, 95% CI 12.78–22.82; both exceeding the 5-point MCID with ≥ 95% confidence), and 6MWD (TEER MD + 17.53 m; TTVR MD + 30.90 m; neither clearly exceeding the 30-m MCID). Both increased major bleeding (TEER NNH 29; TTVR NNH 21) and new pacemaker implantation (TEER NNH 91; TTVR NNH 10). Bayesian posterior probability that TTVR was best was 100% for NYHA improvement, 99% for KCCQ-OS, and 82% for 6MWD, but only 1% for avoidance of pacemaker implantation.ConclusionIn patients with symptomatic moderate-to-severe TR, both TEER and TTVR plus OMT provide consistent and clinically meaningful improvements in functional status and quality of life. Longer-term trials with direct head-to-head comparisons are warranted.Clinical trial registrationNot applicable. This study is a systematic review and meta-analysis of previously published randomized controlled trials.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13019-026-04265-9. Introduction Severe tricuspid regurgitation (TR) is a common valvular disease and an independent adverse prognostic marker associated with impaired quality of life, recurrent heart failure hospitalizations, and increased mortality [1]. Despite its well-established clinical impact, therapeutic options remain limited, as medical therapy is non-curative, surgical intervention carries high operative risk, and transcatheter therapies are relatively recent, supported by evolving evidence, and not universally available. According to current ESC/EACTS guidelines [2], patients with clinically relevant TR should initially be managed with medical therapy aimed at optimization of heart failure treatment, management of pulmonary hypertension, and rhythm control in atrial fibrillation. However, medical therapy has a limited effect on TR progression and should not delay further evaluation. Surgical intervention, although effective, is infrequently performed because of the high operative risk in an elderly and comorbid population. Consequently, less invasive transcatheter therapies, including transcatheter edge-to-edge repair, direct annuloplasty, and orthotopic or heterotopic tricuspid valve replacement, have emerged as important alternatives, with patient selection guided by clinical status and anatomical suitability. Recent randomized controlled trials evaluating these transcatheter approaches have demonstrated improvements in tricuspid regurgitation severity, functional status, and patient-reported quality of life, with acceptable safety profiles [3−6]. However, results across trials have been heterogeneous, and individual studies have been underpowered to reliably assess hard clinical endpoints such as all-cause mortality and heart failure hospitalization. Although only three trials are available, pooling data enhances statistical power for secondary endpoints, allows assessment of consistency across different device platforms (e.g., edge-to-edge repair vs. replacement), and provides a more robust evaluation of safety signals. As a result, the magnitude, consistency, and overall clinical relevance of the benefit associated with transcatheter tricuspid interventions remain uncertain. We therefore conducted a network meta-analysis of randomized controlled trials to compare transcatheter tricuspid valve interventions plus guideline-directed medical therapy versus medical therapy alone in adult patients with severe symptomatic tricuspid regurgitation, evaluating clinical, functional, quality-of-life, and safety outcomes. Beyond the conventional pairwise and network estimates, we further integrated a Bayesian framework to report posterior rank probabilities, translated relative effects into absolute risk differences and number-needed-to-treat/harm, conducted threshold (robustness) analysis against null and clinically important effect boundaries, and applied the CINeMA framework to formally rate confidence in network estimates. Methods This systematic review and meta-analysis was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [7]. The protocol was predefined to ensure transparency, reproducibility, and methodological rigor. Information sources and search strategy A comprehensive literature search was conducted in PubMed, Embase, and ScienceDirect (Elsevier) from database inception through December 2025. The strategy combined Medical Subject Headings (MeSH) and free-text terms for high sensitivity. The primary search string was: (“tricuspid regurgitation”[MeSH]) AND (“transcatheter edge-to-edge repair” OR “TEER” OR “TriClip” OR “transcatheter tricuspid valve replacement” OR “TTVR” OR “EVOQUE”). Broader keyword searches included “tricuspid regurgitation,” “transcatheter tricuspid repair,” and “transcatheter tricuspid valve replacement” to identify studies not captured by device-specific terms. The complete search string for all databases is given in the Supplementary File. Searches were limited to English-language publications involving human subjects and original research articles. No date restrictions were applied initially. Reference lists of included trials and relevant reviews were manually screened for additional eligible studies. Eligibility criteria We included RCTs comparing TTVI (transcatheter edge-to-edge repair or transcatheter tricuspid valve replacement) plus optimal medical therapy versus optimal medical therapy alone. Participants were adults with symptomatic moderate-to-severe or severe TR on stable guideline-directed medical therapy for ≥ 30 days before randomization. Studies were required to report ≥ 1 prespecified outcome: all-cause mortality, heart failure hospitalization, change in KCCQ-OS score, NYHA functional class improvement ≥ 1 class, change in six-minute walk distance (6MWD), major bleeding, or new permanent pacemaker implantation. Exclusion criteria included trials enrolling patients with LVEF ≤ 20%, prohibitive pulmonary hypertension, or unsuitable anatomy for device implantation; non-randomized designs; observational studies; single-arm trials; abstracts; reviews; editorials; non-English publications; or studies lacking an OMT-only comparator arm. Study selection and data extraction Two independent reviewers screened titles and abstracts using Zotero software. Potentially eligible studies underwent full-text review, with discrepancies resolved by consensus or a third reviewer. Data were extracted independently using a standardized, pilot-tested Google spreadsheet. Extracted items included study design, sample size, randomization ratio, baseline demographics and clinical characteristics, procedural details, and all relevant outcomes. Means ± SD were recorded for continuous variables; medians (IQR) were noted when means were unavailable. Corresponding authors were contacted for missing data. Discrepancies were resolved through discussion. Risk of bias assessment Risk of bias was independently assessed by two reviewers using the Cochrane Risk of Bias 2 (RoB 2) tool [8], evaluating domains including randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of reported results. Each domain was rated as low risk, some concerns, or high risk, with overall judgments per Cochrane guidance. Disagreements were resolved by consensus or third-reviewer adjudication. Statistical analysis primary (frequentist) network meta-analysis Pooled risk ratios (RR) for dichotomous outcomes and mean differences (MD) for continuous outcomes were calculated with 95% CI using a frequentist network meta-analysis framework (random-effects model, DerSimonian–Laird estimator [9]) implemented in the netmeta package [10] (R version 4.5.1). Pairwise data were generated with the pairwise() function; netmeta() was run with OMT as the reference group. Heterogeneity was quantified with the I² statistic. Treatment rankings were determined using P-scores [11]. Inconsistency was assessed via the design-by-treatment decomposition model (decomp.design); because the network was star-shaped with no closed loops, incoherence was interpreted descriptively. League tables and rankograms were generated for visual summary. Bayesian network meta-analysis A parallel Bayesian NMA was performed to quantify posterior probabilities of treatment ranking and to provide inferential outputs directly interpretable as probability statements. Under the star-shaped network with vague priors, posterior marginals for each relative effect are well-approximated by the normal sampling distribution of the corresponding frequentist estimate; we therefore sampled 200,000 Monte-Carlo draws from the joint normal distribution of the log-RR (or MD) estimates for TEER versus OMT and TTVR versus OMT. For each draw we computed within-sample ranks (1 = best, 3 = worst) for OMT, TEER, and TTVR according to the direction of benefit for each outcome. We report posterior probabilities of being ranked first, second, or third; mean ranks; and SUCRA values [12]. A sensitivity Bayesian analysis using half-normal prior with scale 0.5 on the between-study standard deviation yielded qualitatively identical rankings. Absolute effect translation and NNT / NNH Relative effects were translated into absolute risk differences (RD) and number-needed-to-treat (NNT) or number-needed-to-harm (NNH) using pooled OMT baseline event rates derived from the control arms of the three included trials (mortality 8.5%, NYHA improvement 27.0%, major bleeding 2.5%, new pacemaker 1.5% at ~ 1-year follow-up). For each outcome we simulated 200,000 draws of log-RR from its normal sampling distribution, converted to a posterior distribution of RD and NNT, and report median and 95% percentile CIs. NNT is reported where the intervention was beneficial; NNH where it was harmful. Threshold (robustness) analysis For each relative effect we computed the distance (in standard errors) of the point estimate from the null boundary (RR = 1 for dichotomous outcomes; MD = 0 for continuous outcomes) and, for continuous outcomes, from the minimal clinically important difference (MCID = 5 points for KCCQ-OS; 30 m for 6MWD). A robustness buffer was defined as |Z| − 1.96; larger buffers indicate greater resilience of the statistical inference to re-analysis with additional data or alternative methods. Pre-specified additional analyses A pairwise random-effects meta-analysis pooling TEER and TTVR together as “TTVI” versus OMT was performed as a secondary synthesis to maximise power for mortality and heart failure endpoints. We also examined, qualitatively, whether treatment effects appeared to differ by baseline TR severity (massive vs. torrential) using trial-level reported data. Given the three-trial network, formal meta-regression, leave-one-out analysis, comparison-adjusted funnel plots, and Egger tests were not considered informative and were not performed; this is explicitly acknowledged as a limitation. Confidence in the evidence – CINeMA Confidence in each network estimate was rated using the CINeMA (Confidence In Network Meta-Analysis) framework [13], evaluating six domains: (i) within-study bias, (ii) reporting bias, (iii) indirectness, (iv) imprecision, (v) heterogeneity, and (vi) incoherence. Each domain was graded as ‘no concerns’, ‘some concerns’, or ‘major concerns’. Domain judgments were aggregated into a single confidence rating (high, moderate, low, very low). Within-study bias was informed by the Cochrane RoB 2 results. Reporting bias was qualitatively assessed. Indirectness was judged against PICO of the review. Imprecision was evaluated against a range of equivalence of 0.80–1.25 for RRs. Heterogeneity was assessed using I² and the prediction interval. Incoherence was not estimable directly because of the star network geometry and was carried forward with ‘some concerns’ by default. Two reviewers performed assessments independently; disagreements were resolved by discussion. Results The PRISMA flow diagram (Fig. 1) summarizes the study selection process. A total of 924 records were identified through database searches (PubMed: 202; ScienceDirect: 430; Embase: 292). After removal of duplicates, 859 records underwent title and abstract screening. Of these, 214 articles were reviewed in full text, and three randomized controlled trials [3–5] met eligibility criteria and were included in the final quantitative synthesis. Fig. 1PRISMA flow diagram of study selection process PRISMA flow diagram of study selection process Study characteristics The three included multicenter, prospective, open-label randomized controlled trials were the TRILUMINATE Pivotal trial, the TRISCEND II pivotal trial, and the Tri.Fr randomized clinical trial. Collectively, these studies enrolled 1050 adult patients with symptomatic moderate-to-severe or severe tricuspid regurgitation receiving stable optimal medical therapy. Two trials evaluated transcatheter edge-to-edge repair using the TriClip system, whereas one trial evaluated transcatheter tricuspid valve replacement using the EVOQUE system. All studies compared TTVI plus OMT versus OMT alone. Across trials, participants were elderly, with mean ages ranging from 78 to 79 years, and were predominantly female (56% to 75%). Atrial fibrillation was highly prevalent (87% to 96%), and the majority of patients had advanced heart failure symptoms, with substantial proportions classified as NYHA class III or IV. Comorbidities including hypertension, chronic kidney disease, diabetes mellitus, and prior heart failure hospitalization were common. Baseline LVEF was generally preserved or mildly reduced (Table 1). Table 1Baseline characteristics of the included patientsCharacteristicTRILUMINATE (n = 350) [5]TRISCEND II (n = 400) [4]Tri.Fr (n = 300) [3]Study DesignMulticenter RCT (1:1)Multicenter RCT (2:1)Multicenter RCT (1:1)Age (years)TTVI 78.0 ± 7.4 / OMT 77.8 ± 7.2TTVI 79.3 ± 7.4 / OMT 79.1 ± 7.7TTVI 78.3 ± 6.4 / OMT 78.7 ± 6.4Female, n (%)TTVI 98 (56.0) / OMT 94 (53.7)TTVI 194 (74.9) / OMT 102 (76.7)TTVI 98 (65.3) / OMT 93 (62.8)Atrial fibrillation, n (%)TTVI 153 (87.4) / OMT 162 (92.6)TTVI 249 (96.1) / OMT 123 (92.5)TTVI 143 (94.1) / OMT 142 (95.9)NYHA III/IV, n (%)TTVI 104 (59.4) / OMT 97 (55.4)TTVI 189 (73.0) / OMT 92 (69.2)TTVI 59 (38.9) / OMT 68 (45.9)Baseline KCCQ-OSTTVI 56.0 ± 23.4 / OMT 54.1 ± 24.2TTVI 52.8 / OMT 50.6TTVI 54 ± 23.4 / OMT 54 ± 25Baseline 6MWD (m)TTVI 240.5 ± 117.1 / OMT 253.6 ± 129.1TTVI 236.4 / OMT 240.8TTVI 302 ± 104 / OMT 309 ± 112LVEF (%)59.0 ± 9.9TTVI 54.4 / OMT 54.3–Hypertension, n (%)TTVI 142 (81.1) / OMT 141 (80.6)TTVI 235 (90.7) / OMT 122 (91.7)TTVI 106 (69.7) / OMT 102 (68.9)CKD, n (%)TTVI 62 (35.4)TTVI 140 (54.1) / OMT 79 (59.4)TTVI 13 (8.55) / OMT 6 (4.05)Diabetes, n (%)TTVI 28 (16.0) / OMT 27 (15.4)TTVI 88 (34.0) / OMT 48 (36.1)TTVI 22 (14.5) / OMT 26 (17.6)Prior HF hospitalizationTTVI 44 (25.1)–TTVI 55 (36.2) / OMT 66 (44.6)TTVI = transcatheter tricuspid valve intervention arm; OMT = optimal medical therapy arm; NYHA = New York Heart Association; KCCQ-OS = Kansas City Cardiomyopathy Questionnaire overall summary score; 6MWD = six-minute walk distance; LVEF = left ventricular ejection fraction; CKD = chronic kidney disease. Dashes (–) indicate data not reported in the original trial publication Table 1Baseline characteristics of the included patientsCharacteristicTRILUMINATE (n = 350) [5]TRISCEND II (n = 400) [4]Tri.Fr (n = 300) [3]Study DesignMulticenter RCT (1:1)Multicenter RCT (2:1)Multicenter RCT (1:1)Age (years)TTVI 78.0 ± 7.4 / OMT 77.8 ± 7.2TTVI 79.3 ± 7.4 / OMT 79.1 ± 7.7TTVI 78.3 ± 6.4 / OMT 78.7 ± 6.4Female, n (%)TTVI 98 (56.0) / OMT 94 (53.7)TTVI 194 (74.9) / OMT 102 (76.7)TTVI 98 (65.3) / OMT 93 (62.8)Atrial fibrillation, n (%)TTVI 153 (87.4) / OMT 162 (92.6)TTVI 249 (96.1) / OMT 123 (92.5)TTVI 143 (94.1) / OMT 142 (95.9)NYHA III/IV, n (%)TTVI 104 (59.4) / OMT 97 (55.4)TTVI 189 (73.0) / OMT 92 (69.2)TTVI 59 (38.9) / OMT 68 (45.9)Baseline KCCQ-OSTTVI 56.0 ± 23.4 / OMT 54.1 ± 24.2TTVI 52.8 / OMT 50.6TTVI 54 ± 23.4 / OMT 54 ± 25Baseline 6MWD (m)TTVI 240.5 ± 117.1 / OMT 253.6 ± 129.1TTVI 236.4 / OMT 240.8TTVI 302 ± 104 / OMT 309 ± 112LVEF (%)59.0 ± 9.9TTVI 54.4 / OMT 54.3–Hypertension, n (%)TTVI 142 (81.1) / OMT 141 (80.6)TTVI 235 (90.7) / OMT 122 (91.7)TTVI 106 (69.7) / OMT 102 (68.9)CKD, n (%)TTVI 62 (35.4)TTVI 140 (54.1) / OMT 79 (59.4)TTVI 13 (8.55) / OMT 6 (4.05)Diabetes, n (%)TTVI 28 (16.0) / OMT 27 (15.4)TTVI 88 (34.0) / OMT 48 (36.1)TTVI 22 (14.5) / OMT 26 (17.6)Prior HF hospitalizationTTVI 44 (25.1)–TTVI 55 (36.2) / OMT 66 (44.6)TTVI = transcatheter tricuspid valve intervention arm; OMT = optimal medical therapy arm; NYHA = New York Heart Association; KCCQ-OS = Kansas City Cardiomyopathy Questionnaire overall summary score; 6MWD = six-minute walk distance; LVEF = left ventricular ejection fraction; CKD = chronic kidney disease. Dashes (–) indicate data not reported in the original trial publication Baseline characteristics of the included patients TTVI = transcatheter tricuspid valve intervention arm; OMT = optimal medical therapy arm; NYHA = New York Heart Association; KCCQ-OS = Kansas City Cardiomyopathy Questionnaire overall summary score; 6MWD = six-minute walk distance; LVEF = left ventricular ejection fraction; CKD = chronic kidney disease. Dashes (–) indicate data not reported in the original trial publication Risk of bias assessment Using the Cochrane RoB 2 tool, the TRILUMINATE Pivotal and TRISCEND II trials were judged to have low overall risk of bias. The Tri.Fr trial was assessed as having some concerns, primarily related to the measurement of patient-reported outcomes in the context of an open-label design. Domain-specific assessments are presented in Fig. 2. Fig. 2Risk of bias assessment using the Cochrane Risk of Bias 2 tool through Summary and Traffic Light plot Risk of bias assessment using the Cochrane Risk of Bias 2 tool through Summary and Traffic Light plot Confidence in the evidence (CINeMA and GRADE) Under the CINeMA framework, confidence in the network estimates was rated as high for NYHA functional class improvement, KCCQ-OS, 6MWD, and major bleeding; and moderate for all-cause mortality and new permanent pacemaker implantation (down-rated for imprecision related to low event counts). Incoherence was carried with ‘some concerns’ for all comparisons because of the star geometry but was not estimable directly. Heterogeneity was low (I² <30%) for all estimable comparisons. The full domain-by-domain CINeMA assessment is provided in Supplementary Table S4; the traditional GRADE summary is given in Table 2. Table 2GRADE summary of the network clinical outcomesOutcomePatients (studies)Effect estimate (95% CI)ImprecisionCertaintyAll-cause mortality1050 (3)TEER RR 0.99 (0.56–1.76); TTVR RR 0.85 (0.51–1.41)SeriousModerateNYHA improvement ≥ 1 class1050 (3)TEER RR 1.46 (1.30–1.64); TTVR RR 3.28 (2.41–4.47)Not seriousHighNew permanent pacemaker1050 (3)TEER RR 1.71 (0.39–7.46); TTVR RR 7.64 (1.31–44.51)SeriousModerateChange in 6MWD (m)1050 (3)TEER MD + 17.53 (0.53–34.52); TTVR MD + 30.90 (7.90–53.90)Not seriousHighMajor bleeding1050 (3)TEER RR 2.40 (1.01–5.70); TTVR RR 2.92 (1.35–6.33)Not seriousHighKCCQ-OS change1050 (3)TEER MD + 11.00 (7.46–14.54); TTVR MD + 17.80 (12.78–22.82)Not seriousHigh Table 2GRADE summary of the network clinical outcomesOutcomePatients (studies)Effect estimate (95% CI)ImprecisionCertaintyAll-cause mortality1050 (3)TEER RR 0.99 (0.56–1.76); TTVR RR 0.85 (0.51–1.41)SeriousModerateNYHA improvement ≥ 1 class1050 (3)TEER RR 1.46 (1.30–1.64); TTVR RR 3.28 (2.41–4.47)Not seriousHighNew permanent pacemaker1050 (3)TEER RR 1.71 (0.39–7.46); TTVR RR 7.64 (1.31–44.51)SeriousModerateChange in 6MWD (m)1050 (3)TEER MD + 17.53 (0.53–34.52); TTVR MD + 30.90 (7.90–53.90)Not seriousHighMajor bleeding1050 (3)TEER RR 2.40 (1.01–5.70); TTVR RR 2.92 (1.35–6.33)Not seriousHighKCCQ-OS change1050 (3)TEER MD + 11.00 (7.46–14.54); TTVR MD + 17.80 (12.78–22.82)Not seriousHigh GRADE summary of the network clinical outcomes Heart failure hospitalization — narrative synthesis Heart failure hospitalization (HFH) data across the three included trials are summarized in Table X. Definitions, follow-up durations, and analytic approaches differed sufficiently that formal network pooling was not appropriate; results are therefore presented descriptively. TRILUMINATE Pivotal. At 1 year, HFH rates were similar between T-TEER and OMT arms, with no statistically significant difference (the primary analysis population showed no HFH benefit; log-rank P = 0.40). At 2 years, the annualized recurrent HFH rate was 0.19 events/patient-year in the T-TEER arm versus 0.26 events/patient-year in the OMT arm, with a joint frailty model hazard ratio of 0.72 (one-sided upper 95% CI limit 0.93; P = 0.02), corresponding to a 28% relative reduction in recurrent HFH (Kar et al., Circulation 2025). Tri.Fr. At 1 year, HFH was reported as a component of both the Packer clinical composite primary endpoint and a hierarchical composite secondary endpoint (win ratio 2.06, 95% CI 1.38–3.08; P < 0.001, driven by KCCQ improvement and HFH). Individual-component HFH rates at 1 year favored T-TEER numerically without reaching independent statistical significance, attributed by investigators to lower-than-expected event rates. Two-year data recently presented at ACC.26 (Donal et al., 2026) reported HFH in 13.8% of T-TEER patients versus 23.0% of OMT patients (40% relative risk reduction) with no crossover permitted by trial design. TRISCEND II. At 1 year, TTVR was associated with numerically fewer HFHs than OMT; HFH was embedded as a component of the hierarchical win-ratio primary endpoint (win ratio 2.02, 95% CI 1.56–2.62), which significantly favored TTVR overall, although the independent HFH component did not reach conventional statistical significance. Two-year intention-to-treat data (Thourani, ACC.26, 2026) reported HFH of 26.8% (TTVR) versus 32.3% (OMT), P = 0.312; crossover-adjusted Bayesian analyses suggested a larger benefit in non-crossover patients but were post-hoc. Network geometry and primary (frequentist) estimates The network included three multicenter RCTs and 1050 patients. Treatments were TEER (n = 327), TTVR (n = 267), and OMT (n = 456). The geometry was star-shaped with OMT as the common comparator and no direct TEER–TTVR comparison (Supplementary Figure S1). No closed loops existed; therefore, inconsistency could not be tested statistically but was assumed to be low. All-cause mortality. Neither TEER (RR 0.99, 95% CI 0.56–1.76) nor TTVR (RR 0.85, 95% CI 0.51–1.41) significantly reduced all-cause mortality compared with OMT (Fig. 3A). NYHA functional class improvement (≥ 1 class). Both TEER (RR 1.46, 95% CI 1.30–1.64) and TTVR (RR 3.28, 95% CI 2.41–4.47) were superior to OMT (Fig. 3B). Major bleeding. Increased risk with both devices versus OMT (TEER RR 2.40, 95% CI 1.01–5.70; TTVR RR 2.92, 95% CI 1.35–6.33) (Fig. 3C). New permanent pacemaker implantation. Both interventions increased risk versus OMT, with TTVR showing a numerically higher signal (TEER RR 1.71, 95% CI 0.39–7.46; TTVR RR 7.64, 95% CI 1.31–44.51) (Fig. 3D). Change in KCCQ-OS score. Both devices produced clinically meaningful improvements versus OMT (TEER MD + 11.00, 95% CI 7.46–14.54; TTVR MD + 17.80, 95% CI 12.78–22.82) (Fig. 3E). Change in 6MWD. Significant improvement with both TEER (MD + 17.53 m, 95% CI 0.53–34.52) and TTVR (MD + 30.90 m, 95% CI 7.90–53.90) versus OMT (Fig. 3F). Fig. 3Network meta-analysis forest plots showing treatment effects of TEER and TTVR versus OMT for (A) all-cause mortality, (B) NYHA functional class improvement ≥ 1 class, (C) major bleeding, (D) new permanent pacemaker implantation, (E) KCCQ-OS score change, and (F) six-minute walk distance change Network meta-analysis forest plots showing treatment effects of TEER and TTVR versus OMT for (A) all-cause mortality, (B) NYHA functional class improvement ≥ 1 class, (C) major bleeding, (D) new permanent pacemaker implantation, (E) KCCQ-OS score change, and (F) six-minute walk distance change Bayesian rank probabilities and SUCRA Posterior rank probabilities quantified the strength of device ranking in a way that is directly interpretable as a probability statement (Table 3; Supplementary Figure S2). For patient-important efficacy outcomes, the probability that TTVR was the best treatment was 100% for NYHA improvement, 99% (SUCRA 0.993) for KCCQ-OS, and 82% (SUCRA 0.907) for 6MWD. TEER had the highest posterior probability of being the safest intervention with respect to new pacemaker implantation among active treatments (P(better than TTVR) for pacemaker avoidance ≈ 90%), but OMT remained the safest option with respect to both bleeding (SUCRA 0.987) and pacemaker (SUCRA 0.875). For all-cause mortality, TTVR had the highest posterior probability of being best (57.3%, SUCRA 0.695), but posterior overlap was substantial, reflecting the imprecise point estimates. Table 3Bayesian posterior rank probabilities and SUCRA (higher SUCRA = better)OutcomeTreatmentP(best), %P(middle), %P(worst), %Mean rankSUCRAAll-cause mortalityTTVR57.324.418.41.610.695TEER29.826.443.82.140.430OMT12.949.237.82.250.376NYHA improvement ≥ 1TTVR100001.001.000TEER010002.000.500OMT001003.000.000KCCQ-OS changeTTVR98.51.501.010.993TEER1.598.501.990.507OMT001003.000.0006MWD changeTTVR81.817.70.41.190.907TEER18.179.72.21.840.580OMT02.697.42.970.013Major bleeding (safety)OMT97.42.601.030.987TEER2.360.737.12.350.326TTVR0.336.762.92.630.187New pacemaker (safety)OMT75.324.50.31.250.875TEER23.766.310.01.860.569TTVR1.09.289.72.890.056 Table 3Bayesian posterior rank probabilities and SUCRA (higher SUCRA = better)OutcomeTreatmentP(best), %P(middle), %P(worst), %Mean rankSUCRAAll-cause mortalityTTVR57.324.418.41.610.695TEER29.826.443.82.140.430OMT12.949.237.82.250.376NYHA improvement ≥ 1TTVR100001.001.000TEER010002.000.500OMT001003.000.000KCCQ-OS changeTTVR98.51.501.010.993TEER1.598.501.990.507OMT001003.000.0006MWD changeTTVR81.817.70.41.190.907TEER18.179.72.21.840.580OMT02.697.42.970.013Major bleeding (safety)OMT97.42.601.030.987TEER2.360.737.12.350.326TTVR0.336.762.92.630.187New pacemaker (safety)OMT75.324.50.31.250.875TEER23.766.310.01.860.569TTVR1.09.289.72.890.056 Bayesian posterior rank probabilities and SUCRA (higher SUCRA = better) Absolute risk differences and number needed to treat / harm Relative effects translated into clinically interpretable absolute effects (Table 4). For every 100 patients with symptomatic severe TR treated for approximately one year, TTVR produced approximately 61 additional patients with ≥ 1-class NYHA improvement (NNT 2; 95% CI 1–3) and TEER produced approximately 12 additional patients (NNT 8; 95% CI 6–12). Absolute mortality differences were small and non-significant for both devices (NNT values of 59–60 with very wide CIs). Safety trade-offs were meaningful: TTVR increased pacemaker implantation by approximately 15 per 100 (NNH 10; 95% CI 2–182), whereas TEER added approximately 2 per 100 (NNH 91; 95% CI 10–2152). Major bleeding was increased by both devices, with NNH of 29 for TEER and 21 for TTVR. Table 4Absolute risk differences and number-needed-to-treat/harm per 100 patients (baseline OMT rate = pooled 1-year control-arm estimate)OutcomeTreatmentOMT rate, %Treatment rate, %RD per 100 (95% CI)NNT / NNH (95% CI)All-cause mortalityTEER8.58.4+ 0.28 (− 3.76 to + 6.46)NNT 60 (15–1286)TTVR8.57.2−1.03 (− 4.15 to + 3.51)NNT 59 (22–1201)NYHA improvement ≥ 1TEER27.039.4+ 12.48 (+ 8.09 to + 17.27)NNT 8 (6–12)TTVR27.088.6†+ 60.53 (+ 38.00 to + 73.00)NNT 2 (1–3)Major bleedingTEER2.56.0+ 4.12 (+ 0.02 to + 11.75)NNH 29 (9–408)TTVR2.57.3+ 5.39 (+ 0.89 to + 13.33)NNH 21 (8–112)New pacemakerTEER1.52.6+ 1.91 (− 0.91 to + 9.67)NNH 91 (10–2152)TTVR1.511.5+ 15.35 (+ 0.46 to + 65.50)NNH 10 (2–182)† The point estimate for NYHA improvement under TTVR is driven by the large RR observed in TRISCEND II and is truncated at the natural ceiling (≤ 100%); the CI range is wide because the underlying baseline NYHA-improvement rate in OMT patients is itself uncertain Table 4Absolute risk differences and number-needed-to-treat/harm per 100 patients (baseline OMT rate = pooled 1-year control-arm estimate)OutcomeTreatmentOMT rate, %Treatment rate, %RD per 100 (95% CI)NNT / NNH (95% CI)All-cause mortalityTEER8.58.4+ 0.28 (− 3.76 to + 6.46)NNT 60 (15–1286)TTVR8.57.2−1.03 (− 4.15 to + 3.51)NNT 59 (22–1201)NYHA improvement ≥ 1TEER27.039.4+ 12.48 (+ 8.09 to + 17.27)NNT 8 (6–12)TTVR27.088.6†+ 60.53 (+ 38.00 to + 73.00)NNT 2 (1–3)Major bleedingTEER2.56.0+ 4.12 (+ 0.02 to + 11.75)NNH 29 (9–408)TTVR2.57.3+ 5.39 (+ 0.89 to + 13.33)NNH 21 (8–112)New pacemakerTEER1.52.6+ 1.91 (− 0.91 to + 9.67)NNH 91 (10–2152)TTVR1.511.5+ 15.35 (+ 0.46 to + 65.50)NNH 10 (2–182)† The point estimate for NYHA improvement under TTVR is driven by the large RR observed in TRISCEND II and is truncated at the natural ceiling (≤ 100%); the CI range is wide because the underlying baseline NYHA-improvement rate in OMT patients is itself uncertain Absolute risk differences and number-needed-to-treat/harm per 100 patients (baseline OMT rate = pooled 1-year control-arm estimate) † The point estimate for NYHA improvement under TTVR is driven by the large RR observed in TRISCEND II and is truncated at the natural ceiling (≤ 100%); the CI range is wide because the underlying baseline NYHA-improvement rate in OMT patients is itself uncertain Threshold (robustness) analysis Threshold analysis examined the distance of each point estimate from relevant decision boundaries (Table 5; Supplementary Table S3). NYHA and KCCQ-OS benefits were highly robust for both devices: point estimates lay 6–8 standard errors from the null, with robustness buffers of 3–5 SE beyond the 95% significance boundary. For 6MWD, both devices were statistically significant versus the null, but the TEER effect (17.5 m) lay clearly below the 30-m MCID (Z versus MCID = − 1.44), and the TTVR effect (30.9 m) only barely exceeded the MCID at its point estimate (Z versus MCID = + 0.08), with the lower bound of its CI falling well below the MCID. Major bleeding with TEER was borderline significant, with a robustness buffer of just 0.02 SE, indicating that even modest additional data could shift the comparison to non-significance; the TTVR bleeding signal was more robust (buffer 0.76 SE). Pacemaker implantation with TTVR was robustly increased (buffer 0.30 SE), whereas TEER’s pacemaker signal was not statistically different from OMT. Table 5Threshold (robustness) analysis summaryOutcomeTreatmentEstimate|Z| vs. nullRobustness bufferClinical significanceAll-cause mortalityTEERRR 0.990.030.00 SENot significantTTVRRR 0.850.630.00 SENot significantNYHA improvement ≥ 1TEERRR 1.466.394.43 SEHighly robustTTVRRR 3.287.545.58 SEHighly robustMajor bleedingTEERRR 2.401.980.02 SEBorderlineTTVRRR 2.922.720.76 SERobustNew pacemakerTEERRR 1.710.710.00 SENot significantTTVRRR 7.642.260.30 SERobustKCCQ-OS (MCID 5)TEERMD + 11.006.09Z vs. MCID + 3.32Exceeds MCID w/ 95% CITTVRMD + 17.806.95Z vs. MCID + 5.00Exceeds MCID w/ 95% CI6MWD (MCID 30 m)TEERMD + 17.53 m2.02Z vs. MCID − 1.44Below MCIDTTVRMD + 30.90 m2.63Z vs. MCID + 0.08At MCID; CI crossesRobustness buffer = |Z| − 1.96 (for dichotomous outcomes relative to the null); larger values indicate greater resilience of statistical inference to additional data. For continuous outcomes, Z vs. MCID quantifies the distance of the point estimate from the clinically important threshold in SE units Table 5Threshold (robustness) analysis summaryOutcomeTreatmentEstimate|Z| vs. nullRobustness bufferClinical significanceAll-cause mortalityTEERRR 0.990.030.00 SENot significantTTVRRR 0.850.630.00 SENot significantNYHA improvement ≥ 1TEERRR 1.466.394.43 SEHighly robustTTVRRR 3.287.545.58 SEHighly robustMajor bleedingTEERRR 2.401.980.02 SEBorderlineTTVRRR 2.922.720.76 SERobustNew pacemakerTEERRR 1.710.710.00 SENot significantTTVRRR 7.642.260.30 SERobustKCCQ-OS (MCID 5)TEERMD + 11.006.09Z vs. MCID + 3.32Exceeds MCID w/ 95% CITTVRMD + 17.806.95Z vs. MCID + 5.00Exceeds MCID w/ 95% CI6MWD (MCID 30 m)TEERMD + 17.53 m2.02Z vs. MCID − 1.44Below MCIDTTVRMD + 30.90 m2.63Z vs. MCID + 0.08At MCID; CI crossesRobustness buffer = |Z| − 1.96 (for dichotomous outcomes relative to the null); larger values indicate greater resilience of statistical inference to additional data. For continuous outcomes, Z vs. MCID quantifies the distance of the point estimate from the clinically important threshold in SE units Threshold (robustness) analysis summary Robustness buffer = |Z| − 1.96 (for dichotomous outcomes relative to the null); larger values indicate greater resilience of statistical inference to additional data. For continuous outcomes, Z vs. MCID quantifies the distance of the point estimate from the clinically important threshold in SE units P-scores for frequentist ranking (reference) See Table 6. Table 6Frequentist P-scores (for comparison with Bayesian SUCRA above)TreatmentNYHAKCCQ-OS6MWDMortalityBleedingNew PPMTEER0.5000.5070.5790.4260.3270.568TTVR1.0000.9930.9080.6980.1870.056OMT0.0000.0000.0130.3760.9860.876 Table 6Frequentist P-scores (for comparison with Bayesian SUCRA above)TreatmentNYHAKCCQ-OS6MWDMortalityBleedingNew PPMTEER0.5000.5070.5790.4260.3270.568TTVR1.0000.9930.9080.6980.1870.056OMT0.0000.0000.0130.3760.9860.876 Frequentist P-scores (for comparison with Bayesian SUCRA above) Discussion In this network meta-analysis of three pivotal randomized controlled trials enrolling 1050 patients, both TEER and TTVR produced consistent, clinically meaningful improvements in functional status and health-related quality of life compared with OMT alone, without affecting short-term mortality. By integrating frequentist and Bayesian frameworks, translating relative effects into absolute risks and numbers needed to treat, performing threshold-based robustness analyses, and formally rating confidence with CINeMA, this analysis provides a more granular and clinically actionable synthesis than prior pairwise reviews [14]. The observed functional and quality-of-life gains can be mechanistically explained by the profound reduction in TR severity achieved by both device platforms. Effective TR elimination alleviates right ventricular volume overload, reduces systemic venous congestion, and improves forward cardiac output, thereby promoting reverse right ventricular remodeling and enhancing hepatorenal perfusion [15−17] These physiological changes translate into better exercise tolerance, reduced symptom burden, and higher patient-reported outcomes, as consistently reported across contemporary registries and meta-analyses [18−21]. Our findings extend prior pairwise meta-analyses in three clinically important ways. First, network comparison shows that TTVR confers numerically larger functional and quality-of-life benefits than TEER, a difference reinforced by near-certain Bayesian ranking (posterior probability that TTVR is best: 100% for NYHA improvement, 99% for KCCQ-OS, 82% for 6MWD). This pattern is biologically plausible: TTVR achieves near-complete TR elimination by replacing the valve, whereas TEER relies on leaflet coaptation and may leave residual moderate TR in a subset of patients with complex anatomy [22, 23]. Second, absolute-risk translation highlights the magnitude of benefit in terms clinicians can use at the bedside: an NNT of 2 for NYHA improvement with TTVR (versus 8 for TEER) is among the most favorable numbers reported for any contemporary heart-failure therapy. Third, threshold analysis reveals an important caveat for 6MWD—although both devices improved 6MWD statistically versus OMT, only TTVR’s point estimate marginally reached the canonical 30-m MCID, and the lower bound of its CI remained below the MCID. This suggests that, beyond statistical significance, the clinical relevance of 6MWD gains is modest and should not be over-interpreted. KCCQ-OS results, by contrast, decisively exceed their 5-point MCID with 95% confidence for both devices. Safety signals also differed between devices. Both TEER and TTVR were associated with increased procedural risks, particularly major bleeding and new permanent pacemaker implantation, yet TTVR carried a numerically higher burden of these complications. Translated into absolute risks, pacemaker implantation carries a particularly striking signal for TTVR: NNH of 10, meaning that for every ten TTVR procedures performed, approximately one additional patient requires a new permanent pacemaker beyond baseline. The TEER pacemaker signal was not statistically robust (robustness buffer 0 SE). Major bleeding was robustly increased by TTVR and borderline-increased by TEER (buffer 0.02 SE)—a fragile result that could reverse with modest additional data. These findings are concordant with large real-world registries and device-specific meta-analyses that have documented higher rates of conduction disturbances and bleeding with orthotopic valve replacement compared with edge-to-edge repair [24−26]. Despite clear symptomatic benefits, neither TEER nor TTVR significantly reduced all-cause mortality or heart failure hospitalization in the short term. Posterior probability that TTVR is best for mortality was 57%, and the effect direction favored TTVR, but the CI was wide and the result is imprecise rather than null; this dissociation between patient-centered outcomes and hard clinical endpoints mirrors the results of the individual pivotal trials and recent meta-analyses [27, 28]. Possible explanations include the advanced disease stage at enrollment, relatively short follow-up duration (typically 1 year), and the presence of multiple comorbidities that continue to drive adverse outcomes irrespective of TR reduction [25]. From a clinical perspective, these data reinforce the role of transcatheter therapies as a valuable addition to the heart-team armamentarium for patients with severe symptomatic TR who remain limited despite optimized medical therapy. The Bayesian framework we employed allows clinicians to think in terms of probability of benefit rather than significance thresholds: a patient choosing between TEER and TTVR should be informed that TTVR is very likely to produce the greater symptomatic gain, but also carries a considerably higher pacemaker risk (NNH 10). Device selection should be individualized according to anatomic suitability, right ventricular function, baseline conduction status, and patient preference, with TTVR potentially favored in cases of torrential TR or large coaptation gaps, and TEER preferred when minimizing pacemaker risk is a priority [23, 29]. Since the cutoff of our prespecified data synthesis (1-year primary-endpoint publications through December 2025), 2-year outcomes have become available for all three pivotal trials. The TRILUMINATE Pivotal 2-year analysis (Kar et al., Circulation 2025) reported a statistically significant 28% reduction in recurrent HFH (0.19 vs. 0.26 events/patient-year; HR 0.72, 1-sided upper 95% CI 0.93; P = 0.02). The Tri.Fr 2-year data presented at ACC.26 (Donal, March 2026) demonstrated a 40% relative reduction in HFH (13.8% vs. 23.0%) in a trial uniquely characterized by the absence of crossover, strengthening causal inference. TRISCEND II 2-year intention-to-treat results did not reach statistical significance for individual hard endpoints but showed directionally consistent numerical advantages for TTVR in both mortality (19.1% vs. 25.3%) and HFH (26.8% vs. 32.3%), with crossover-adjusted post-hoc analyses suggesting a mortality benefit. These newer data are directionally consistent with, and biologically coherent with, the sustained TR reduction observed in all three trials, and they suggest that the dissociation we observe at 1 year between functional/quality-of-life benefits and hard clinical endpoints may be primarily a question of follow-up duration rather than a ceiling on the therapeutic value of transcatheter tricuspid intervention. A prespecified update of our synthesis, restricted to 1-year primary publications, remains appropriate; however, readers should interpret our null findings for mortality and HFH in light of this rapidly maturing evidence base. Several limitations of the current evidence base must be acknowledged. The network is star-shaped with no direct TEER–TTVR comparison, making indirect comparisons structurally fragile and precluding statistical testing of incoherence; this was carried forward as ‘some concerns’ in the CINeMA assessment. With only three trials, meta-regression, leave-one-out analysis, comparison-adjusted funnel plots, and Egger tests are not statistically informative and were not performed. Follow-up remains relatively short, and open-label designs introduce potential bias in subjective endpoints. Heart failure hospitalization data could not be pooled because of inconsistent reporting across trials. The NNT/NNH estimates rely on pooled control-arm event rates and should be viewed as a translation aid rather than precise point predictions; they carry wide CIs for low-event outcomes. Future adequately powered randomized trials with longer-term follow-up and direct head-to-head device comparisons are urgently needed to refine patient selection, optimize timing of intervention, and definitively establish the prognostic impact of these therapies [30]. Conclusion This network meta-analysis, extended with Bayesian ranking, absolute-effect translation, threshold robustness analysis, and CINeMA confidence rating, confirms that both TEER and TTVR provide substantial and consistent symptomatic relief in patients with severe TR. While short-term mortality remains unaffected, the magnitude of functional and quality-of-life improvement, particularly with TTVR (NNT = 2 for NYHA improvement; posterior probability of being best for efficacy > 99%), positions transcatheter tricuspid interventions as a meaningful therapeutic advance. These benefits must be weighed against clinically important procedural risks, most notably pacemaker implantation (TTVR NNH = 10) and major bleeding. As experience grows and longer-term data mature, these less-invasive options are likely to become a cornerstone of TR management in appropriately selected patients. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1