This retrospective study analyzed consecutive patients who had severe TR and underwent TVR surgery at Beijing Anzhen Hospital between November 1993 and February 2019. Decisions regarding the need for valve replacement and choice of prosthesis were made intraoperatively by the attending surgeon, with a strong institutional preference for valve repair whenever feasible. The inclusion criteria comprised patients with severe TR who underwent isolated or concomitant TVR and had complete preoperative and postoperative clinical data. Exclusion criteria included patients under 18 years of age, those with missing or incomplete clinical data, and individuals undergoing surgery for congenital tricuspid valve disease, as the conditions represent distinct clinical scenarios with unique preoperative profiles and outcomes that are typically treated as separate categories in the literature [1]. Patients who underwent TVR without additional major interventions (e.g., other valve surgery or coronary artery bypass grafting) comprised the isolated-TVR cohort. The primary endpoint was long-term all-cause mortality, while secondary endpoints included in-hospital morbidity and complications such as stroke, renal failure, hepatic failure, bleeding and reoperation. This study was approved by the Ethics Committee of Beijing Anzhen Hospital, Capital Medical University (approval number: 2020101X). Due to the retrospective nature of the study and the use of deidentified patient data, the requirement for informed consent was waived. Data supporting the findings of this study are available from the corresponding author upon reasonable request.

Data were extracted from electronic medical records and included demographic information, preoperative characteristics, intraoperative details, and postoperative outcomes. Right-sided heart failure signs were defined as the presence of clinical symptoms such as peripheral edema, hepatomegaly, spleen enlargements and ascites. The TRI-SCORE was calculated for each patient using the original criteria described by Dreyfus et al. [13] In-hospital mortality was defined as all-cause death occurring during the initial hospitalization. Stroke was defined as a persistent central neurologic deficit, assessed by a neurologist. Acute renal failure was defined as an increase in serum creatinine by $\ge$0.3 mg/dL within 48 hours or a 1.5-fold increase from baseline, or urine output 0.5 mL/kg/h for 6 hours [14]. Liver failure was defined as the presence of coagulopathy (INR $>$1.5) and any degree of encephalopathy in a patient without pre-existing cirrhosis or as acute decompensation in a patient with chronic liver disease [15]. Re-exploration for bleeding was defined as any return to surgery for bleeding within the first 24 postoperative hours.

All patients received postoperative anticoagulation according to valve type and evolving institutional guidelines. Mechanical-valve recipients were started on warfarin targeting an INR of 2.5–3.5, with monthly monitoring at dedicated clinics; aspirin (75–100 mg/day) was added for those with atrial fibrillation or prior thromboembolism. Bioprosthetic-valve recipients received warfarin for 3–6 months (INR 2.0–3.0) before switching to lifelong aspirin, unless contraindicated. Follow-up was completed by scheduled clinic visits and telephone interviews; over a median of 78 months (range up to 22 years), no patient was lost to follow-up.

Continuous variables with a normal distribution are reported as mean $\pm$ standard deviation and compared by two-sample t-test. Non-normally distributed continuous data are presented as median (interquartile range) and analyzed using the Mann–Whitney U test. Categorical data are reported as number (percentage) and compared with Pearson’s $\chi$² test when every expected cell count is $\ge$5, with the Yates-corrected $\chi$² test for 2 $\times$ 2 tables containing any expected count 5, and with the two-sided Fisher exact test when an expected count falls below 1.

Predictors of in-hospital mortality were identified using logistic regression models. Initially, univariable analysis was performed to identify potential predictors during the perioperative period. Variables significant in univariable analysis (p 0.05) were included in multivariable logistic regression models to fine-tune the predictive model and verify the proportional odds assumption.

For long-term follow-up mortality, Cox proportional hazards regression models were used. Univariable Cox regression analysis identified potential predictors, and significant variables (p 0.05) were included in multivariable analysis to derive the best predictive model. The proportional hazards assumption was checked for each Cox regression analysis. Kaplan-Meier survival curves were generated and compared using the log-rank test. To assess potential temporal trends, we divided the study period into three intervals (1993–2003, 2004–2012 and 2013–2019) and performed survival analyses for each time period.

To mitigate potential selection bias and confounding, we conducted sensitivity analyses using inverse probability of treatment weighting (IPTW) and propensity score matching (PSM). PSM between the isolated and concomitant TVR groups was conducted using a multivariable logistic regression model with the following variables: age, sex, body mass index, hypertension, diabetes mellitus, chronic lung disease, chronic liver disease, chronic kidney disease, prior stroke, left ventricular ejection fraction, New York Heart Association (NYHA) Class III/IV, coronary artery disease, previous cardiac surgery, TRI-SCORE, peripheral vascular disease and right-sided heart failure signs. We employed a nearest neighbor matching algorithm with a 1:1 ratio for PSM and applied the same set of variables for IPTW to maintain consistency. The balance of covariates after PSM and IPTW was evaluated using standardized differences, with values below 0.1 indicating well-balanced groups.

To explore whether outcomes differ between first-time isolated TVR and redo isolated TVR, we conducted sensitivity analyses repeating all outcome models using a three-level exposure variable: (i) concomitant TVR (reference), (ii) primary-isolated TVR (no prior cardiac surgery), and (iii) redo-isolated TVR ($\ge$1 prior cardiac surgery). Covariate adjustment was identical to the main analysis.

Statistical significance was set at a two-tailed p-value of 0.05. All analyses were conducted using R version 4.2.2 (R Core Team, Auckland, New Zealand) and Stata/MP version 17.0 (StataCorp LLC, College Station, TX, USA).

Between 1993 and 2019, a total of 18,496 patients were referred for tricuspid valve surgery at Beijing Anzhen Hospital. Out of these, 245 patients met the inclusion criteria and were included in this study (Fig. 1). Among them, 128 patients (52%) underwent isolated TVR, while 117 patients (48%) underwent TVR concomitant with other major procedures. Patients who underwent isolated TVR had a lower prevalence of male gender and a higher body mass index. They also had more instances of previous cardiac surgery and previous tricuspid valve surgery. Additionally, these patients exhibited lower EuroSCORE II and systolic pulmonary artery pressure compared to those undergoing concomitant TVR. The baseline characteristics of the study population are summarized in Table 1.

Fig. 1.

Study cohort selection for tricuspid valve replacement analysis. Flowchart showing the selection process of the study cohort for TVR at Beijing Anzhen Hospital from 1993 to 2019. The chart details the exclusions and final cohort distribution into isolated and combined TVR groups.

Operative characteristics and clinical outcomes are presented in Table 2. The concomitant TVR group had significantly longer bypass time [153 (123–196) minutes versus 99 (76–133) minutes, p 0.001] and cross-clamp time [99 (78–135) minutes versus 49 (0–66) minutes, p 0.001] compared to the isolated TVR group. In the entire cohort, 97 patients (39.6%) received a mechanical prosthesis, with the most common valve sizes being 31 mm (used in 119 patients). Post-operatively, the isolated TVR group showed generally more favorable outcomes. In-hospital mortality was significantly lower than in the concomitant group (7.8% versus 17.9%, p = 0.017). Acute renal failure occurred less often, although this difference did not reach statistical significance (8.6% versus 16.2%, p = 0.068). Additionally, the isolated TVR group had shorter mechanical ventilation times [20 (14–38) hours versus 23 (16–48) hours, p = 0.016] and shorter intensive care unit (ICU) stays [37 (18–62) hours versus 44 (20–90) hours, p = 0.002].

Factors associated with in-hospital mortality in univariate and multivariate analyses are presented in Table 3. In the univariate analysis, several variables were examined. In the univariate analysis, age, isolated TVR, TRI-SCORE, chronic kidney disease and signs of right-sided heart failure showed significant associations with in-hospital mortality. To mitigate the risk of overfitting in our multivariate analysis, we limited the number of predictors to three, based on the number of observed events (31 deaths). In the multivariate analysis, isolated TVR remained an independent protective factor for in-hospital mortality (OR = 0.40; 95% CI 0.17–0.95; p = 0.037).

Follow-up data were available for all participants, with a mean follow-up duration of 6.8 $\pm$ 5.7 years [median 6.5 years (1.8–9.6), up to 21.9 years]. The isolated TVR group exhibited a significantly lower prevalence of overall mortality compared to the concomitant TVR group (p = 0.018, Fig. 2A). At 1, 5, and 10 years, the survival rates for isolated TVR were 89.1%, 83.3%, and 77.7%, respectively, while for concomitant TVR they were 72.6%, 68.9%, and 60.5%, respectively.

Fig. 2.

Kaplan-Meier survival curves for isolated and concomitant tricuspid valve replacement. (A) Unmatched analysis shows a significant difference in survival between the two groups (p = 0.018). (B) Propensity score matched analysis shows a significant difference in survival (p = 0.023). (C) Inverse probability of treatment weighting analysis indicates a trend towards lower mortality in the isolated TVR group (p = 0.101). TVR, tricuspid valve replacement; IPTW, Inverse probability of treatment weighting.

Univariate Cox regression analysis indicated that overall mortality was associated with age, isolated TVR, chronic kidney disease, TRI-SCORE and previous cardiac surgery history (Table 4). After adjusting for covariates, isolated TVR remained an independent protective factor for overall mortality (HR = 0.49, 95% CI: 0.30–0.81, p = 0.005).

The survival outcomes were analyzed across three time periods: 1993–2003, 2004–2012, and 2013–2019 to assess potential temporal trends (Fig. 3). Overall survival did not differ significantly across the time periods (p = 0.61, Fig. 3A). In the 1993–2003 period, no significant difference in survival was observed between concomitant and isolated TVR (p = 0.48, Fig. 3B). However, during the 2004–2012 period, isolated TVR was associated with significantly better survival compared to concomitant TVR (p = 0.032, Fig. 3C). In the 2013–2019 period, there was no significant survival difference between the two groups (p = 0.29, Fig. 3D). These results suggest that while there may have been a temporal effect in the 2004–2012 period favoring isolated TVR, no consistent trends were observed over the entire study period.

Fig. 3.

Kaplan-Meier survival curves comparing overall survival by time period and between isolated and concomitant TVR groups. (A) Overall survival across three time periods: 1993–2003, 2004–2012, and 2013–2019, showing no significant differences in survival across the time periods (p = 0.610). (B) Survival comparison between concomitant and isolated TVR in the 1993–2003 period, showing no significant difference (p = 0.480). (C) Survival comparison between concomitant and isolated TVR in the 2004–2013 period, with isolated TVR showing significantly better survival (p = 0.032). (D) Survival comparison between concomitant and isolated TVR in the 2014–2019 period, showing no significant difference (p = 0.290). TVR, tricuspid valve replacement.

To account for baseline differences and validate our findings, we conducted matching between patients in the isolated and concomitant TVR groups. After PSM, isolated TVR group showed significantly lower in-hospital mortality (3.9% versus 18.2%, p = 0.005) and long-term follow-up overall mortality (19.5% versus 37.7%, p = 0.013) compared to the concomitant TVR group. However, the smaller sample size post-PSM may have diminished the statistical power of these comparisons. To mitigate this limitation, we additionally applied IPTW using the same set of covariates as in PSM.

After both PSM and IPTW adjustments, the standardized differences for nearly all covariates were below 0.1, except left ventricular ejection fraction for PSM, indicating good balance between the groups (Fig. 4). After IPTW, the isolated TVR group showed lower in-hospital mortality (6.6% versus 17.1%, p = 0.013) and long-term follow-up overall mortality (24.5% versus 37.0%, p = 0.064) than the concomitant TVR group. Table 5 presents logistic regression results for in-hospital mortality. Both PSM analysis and IPTW analysis confirms lower in-hospital mortality for isolated TVR [(OR = 0.09, 95% CI: 0.02–0.42, p = 0.003) and (OR = 0.28, 95% CI: 0.11–0.70, p = 0.007)]. Table 6 shows the Cox regression results for long-term overall mortality. Fig. 2B,C illustrate the survival for overall mortality in the PSM and IPTW analyses for the isolated and concomitant TVR groups. Detailed characteristics of the study cohort after PSM and IPTW are shown in Supplementary Tables 1–4.

Fig. 4.

Standardized mean differences for covariates across unmatched, matched, and weighted groups. Standardized mean differences for various covariates in the unmatched, propensity score matched, and inverse probability of treatment weighting groups. The dashed line at SMD = 0.1 indicates the threshold for acceptable balance. The figure shows improved balance post-matching and weighting, with nearly all variables achieving SMD below the threshold.

Among the 55 primary-isolated, 73 redo-isolated, and 117 concomitant cases, in-hospital deaths occurred in 0%, 13.7%, and 17.9%, respectively (two-sided Fisher’s exact p = 0.001). In the multivariable logistic model restricted to patients with at least one event (redo-isolated plus concomitant), TRI-SCORE remained independently associated with mortality (adjusted OR = 1.51, 95% CI 1.21–1.88; p 0.001), whereas CKD and the redo-isolated indicator were not significant (Supplementary Table 5).

During follow-up, overall mortality was 10.9% (6/55), 31.5% (23/73) and 40.2% (47/117) in the primary-isolated, redo-isolated and concomitant cohorts, respectively (Fisher’s exact p 0.001). Primary-isolated TVR conferred a significant long-term survival advantage over concomitant TVR (adjusted HR = 0.35, 95% CI 0.14–0.82; p = 0.017), whereas redo-isolated TVR showed no statistical difference (adjusted HR = 0.72, 95% CI 0.43–1.19; p = 0.194) (Supplementary Tables 6,7). A higher TRI-SCORE remained an independent predictor of late mortality (HR 1.28 per point, 95% CI 1.12–1.45; p 0.001). Supplementary Fig. 1 illustrate the survival for overall mortality in primary-isolated, redo-isolated and concomitant TVR.

This study has several limitations. First, the research was conducted at a single institution which may limit the generalizability of the findings to other settings or populations. Second, the retrospective design of the study may introduce selection bias, despite the use of PSM and IPTW to mitigate this issue. Third, the study included only patients who underwent TVR, excluding those who had tricuspid valve repair, which may limit the applicability of the results to the broader population of patients with tricuspid valve disease. The relatively small sample size (245 patients) might affect the statistical power of the study and the robustness of the conclusions drawn. Formal anatomic staging of tricuspid disease was not available in this retrospective cohort, necessitating reliance on surrogate markers and TRI-SCORE for disease-severity assessment. Additionally, detailed cause-specific mortality data were not available in our retrospective cohort, preventing formal adjudication of operative deaths. Detailed rates of prosthetic valve failure were unavailable in this retrospective series, limiting our ability to assess long-term valve durability. Furthermore, the 25-year study period encompasses significant advancements in surgical techniques and postoperative care, introducing variability based on the treatment era. Lastly, the heterogeneity in patient populations regarding previous cardiac surgeries may result in residual confounding, despite our efforts to adjust for these factors. Addressing these limitations in future research will be crucial to corroborate our findings and enhance the understanding of TVR outcomes.