1 Introduction
Tricuspid regurgitation (TR) is common in patients post biatrial orthotropic heart transplant (OHT) [
1]. Risk factors for TR after OHT include endomyocardial biopsies, allograft rejection, mismatch between the donor heart size and pericardial cavity dimensions [
2–
4]. Additionally, several studies identified a biatrial anastomoses technique (vs. a bicaval anastomoses) as independent risk factor for TR after OHT [
3,
5,
6]. Nevertheless, the clinical impact of TR remains unclear, partly because post-OHT TR is a dynamic disease that changes over time in individual patients. Due to these complex characteristics the clinical impact of post-OHT TR cannot be approached using traditional statistical tools. In this study, we aim to elucidate the clinical impact and long-term course of TR, taking into account its dynamic nature, by using novel statistical models to link the course of post-OHT TR to survival and renal function.
2 Materials and methods
2.1 Patients
Consecutive patients that underwent biatrial OHT from 1984 to 2016 in Erasmus MC were included in this retrospective cohort study (n = 687). Patients whose echocardiograms results were not retrievable or had recorded echocardiograms without TR measurements were excluded (n = 115) resulting in 572 patients eligible for analyses. Of note, most patients that died within 30 days did not have a TR measurement on echocardiogram and, therefore, were excluded (Fig. S1). Approval from the local Medical Ethical Committee was obtained to conduct this study (MEC-2017-421).
2.2 Data collection
Baseline characteristics were extracted from our institutional OHT database. Additionally, all echocardiographic measurements and creatinine measurements were collected longitudinally via automated extraction from the electronic patient records. Furthermore, echocardiographic measurements were supplemented with data acquired from paper patient records. The Dutch Municipal Civil Registry was checked for the survival status.
2.3 Study outcome
The main outcome of this study is mortality in relation to the changing TR severity over time. Secondary outcomes include: the evolution of post-OHT TR grade and the evolution of post-OHT creatinine in relation to the changing TR severity.
2.4 Operation
All patients were operated with the biatrial anastomoses technique described in 1960 by Lower and Shumway [
7]. This technique entails an incision in the right atrium from the inferior vena cava toward the right atrial appendage to avoid sino-atrial node injury.
2.5 Statistics
Continuous data are presented as mean ± standard deviation (Gaussian distribution) or median (interquartile range (IQR)) (non-Gaussian distribution). Categorical data are presented as frequencies (percentage).
Logistic mixed-effect models were used to assess probability of TR over time and investigate determinants of the longitudinal evolution over time. These models included random intercept and slope effects to capture the correlation of the repeated measurements in each patient. Natural splines with 2 knots placed at the 1st and 3rd quartiles were used to allow for flexibility of the subject-specific trajectories over time. Splines allow for nonlinear trajectories over time. This is achieved by allowing a different spline-coefficient for each time interval defined by the knots (e.g., two knots define 3 such intervals). Survival probabilities were estimated and visualized by the Kaplan–Meier method. A joint model was developed to investigate determinants of mortality. More specifically, the mixed-effects model of TR and a relative risk model for the hazard of death (e.g., Cox model) were jointly modeled using shared-random effects. The subject-specific estimated longitudinal profiles were included in the relative risk model as predictors. Joint modeling has several benefits, such as the appropriate inclusion of endogenous covariates in relative risk models (TR), reduced bias and increased efficiency, while it can be used to derive dynamic predictions [
8]. At time point
t one can investigate the effect of the current value of TR, the effect of the slope of TR (at which speed probability of TR is changing at time point
t) and the cumulative effect of TR.
Global-local shrinkage priors were used for the regression coefficients of the relative risk sub-model for the selection of the current value of TR as predictor and this is presented in the article (Tables S1−S3).
The longitudinal evolution of TR probability was correlated to the longitudinal evolution of creatinine by multivariate (multiple outcomes) mixed modeling. Correlation tests were done on the random effects D matrix. A P value < 0.05 was considered statistically significant. Statistical analyses were done in R (R core team 2017, Vienna, Austria) with the use of statistical packages “GLMMadaptive,” “splines,” “JointAI,” “survival” and “JMbayes.”
3 Results
In total, 572 patients were included in this study. Baseline characteristics are presented in Tab.1. Given the dynamic character of TR over time, baseline characteristics are not stratified on post-OHT TR grade. On average patients were 50 years old and 74.9% was male. Most frequently cyclosporine/prednisone (26.4%) and tacrolimus/prednisone (21.3%) are prescribed as immunosuppressive maintenance therapy. Median follow-up was 10.4 years (IQR: 6.4−15.3). Two patients were lost in follow-up, resulting in a completeness of 99.5% (C). None of the patients received surgery for TR during follow-up.
3.1 Tricuspid regurgitation evolution
In total, 8826 echocardiograms were collected (range: 1−50, mean: 15.4) and all echocardiograms are used in the analyses. The model predicting the evolution of TR over time is presented in Tab.2. Probability of TR changed over time, as indicated by the significant times estimates (Tab.2). The evolution of the probability of moderate-to-severe TR over time, as estimated by the mixed-model, is presented in Fig.1. On average, approximately 32% of patients have moderate-to-severe TR immediately after surgery. However, this declines to approximately 11% after 5 years and 9% after 10 years of surgery. Pre-implant mechanical support was significantly associated with lower probability of moderate-to-severe TR during follow-up (Tab.2). Additionally, a worse LV function at the time of the TR measurement was significantly associated with a higher probability of moderate-to-severe TR (Tab.2). Strikingly, the number of rejections in the first year was not associated with a higher probability of moderate-to-severe TR.
3.2 Mortality
During follow-up 357 patients died of which 5 (0.9%) were within 30 days. Survival at 1, 5, 10, 20 years was 97% ± 1%, 88% ± 1%, 66% ± 2% and 23% ± 2%, respectively (Fig.2). The presence of moderate-to-severe TR during follow-up was associated with higher mortality (Tab.3). Tab.3 presents the estimates of the joint model. A higher age, the presence of pre-OHT diabetes, recipient female sex and dialysis were significantly associated with mortality during follow-up. Moderate-to-severe TR remained significant a sensitivity analyses in which left ventricular dysfunction was incorporated in the Cox model as time-varying covariate (Table S4).
Fig.3 presents a dynamic survival probability plot for two patients. The first patient developed moderate-to-severe TR after approximately 3 years. At this moment, the survival probability of 10 years later is estimated to be 77% (Fig.3). The second patient did not develop moderate-to-severe TR at 3 years, and the survival probability of 10 years later for this patient is estimated to be 81% (Fig.3).
3.3 Kidney function
Creatinine was collected at 4426 times simultaneously with an echocardiogram. The longitudinal evolution of creatinine is presented in Fig.4 as estimated by a mixed-model containing only the variable time with a spline function.
The random slope of moderate-to-severe TR was highly positively correlated to the slope of creatinine (R = 0.45), meaning that if the probability moderate-to-severe TR increases, creatinine levels also increase in an individual patient. The intercept (starting point) of TR was not highly correlated the intercept (starting point) of creatinine (R = 0.04). The correlation matrix is shown in Table S5.
The current value of post-OHT moderate-to-severe TR was found to be predictive for dialysis dependence (HR 1.21 95% CI (1.04 to 1.44), P = 0.012) as estimated by a simple joint-model adjusting for baseline creatinine, sex and age (Table S6).
4 Discussion
This study investigated the long-term course of moderate-to-severe TR and its impact on mortality and renal function. We found that moderate-to-severe TR during follow-up was associated with higher mortality and progressive decline of renal function. Specifically, moderate-to-severe TR was found to be a risk factor for dialysis. To the authors’ knowledge, this is the first study that accounts for the dynamic nature of TR during follow-up.
4.1 TR evolution
The etiology of TR after OHT is multifactorial in nature [
3]. In older studies higher pulmonary pressures after OHT and endomyocardial biopsies were mainly found to be associated with TR [
3,
9–
11]. Furthermore, the biatrial surgical technique is found to be associated with more TR in multiple studies [
12].
In our study, left ventricular dysfunction at the time of TR measurement was significantly associated with the higher probabilities of moderate-to-severe TR, probably because worse LV function causes higher pulmonary pressures, subsequently leading to RV dysfunction and dilatation, and leading to functional TR. Moreover, we noted that patients who have mechanical assistance (LVAD, ECMO, IABP) prior OHT have a lower probability of post-OHT moderate-to-severe TR. It has been observed that left ventricular assist devices effectively unload the left ventricle and reduce pulmonary pressures [
13]. Hence, patients with pre-OHT mechanical assistance will probably have lower pulmonary pressures, resulting in less right ventricle dysfunction, annulus dilation, and secondary TR immediately after OHT.
Other studies noted initially a decrease in TR severity after OHT, but a relative increase later in follow up, or even a gradual increase in TR over time [
5,
11,
14]. This study did not replicate these results. Nevertheless, change over time was not significantly decreasing over time later in follow-up. The results of prior studies can partly be explained by the used methodology, which does not take into account the correlations within patients vs. between patients nor does take into account the dropout of patients (either due to death or censoring), whereas the joint modeling framework does take these phenomena into account.
4.2 Mortality and morbidity
In this cohort we only included patients with a follow-up echocardiogram, as the focus was on evolution of TR. Previously we reported the outcomes of the entire cohort [
15]. In patient who die early an echocardiogram may not be performed or TR in this echocardiogram is not recorded, explaining the low 30-day mortality (0.9%) in this subset of the entire cohort (Fig. S1).
Previous studies noted that TR at discharge was associated with impaired late mortality [
5,
16]. Two other studies examined late TR and noted contradicting results in regard to the association with mortality [
14,
17]. This study models the dynamic nature of TR over time and the association with mortality. During follow-up developing TR is associated with higher probability of mortality. The dynamic predictions estimated that developing TR at 3 years after OHT is paired with a 4% reduction in survival 10 years later compared to a patient who does not develop TR, given that all the other variables are similar.
The observed association of TR with mortality does not inherently imply a causal association. An important factor in this interplay is right ventricular dysfunction. We did not analyze right ventricular function in this study, because of high rate of missing data (it was only reported in 1216 of the 8826 echocardiograms) without possibilities for review of older echocardiograms, and it had a tendency to be reported more if significantly impaired (missing mechanism: not at random). Moreover, it is complicated to make causal inference in regard to right ventricular dysfunction and TR due to their circular relationship; TR leads to right ventricular dysfunction, which leads to dilatation, in turn leading to more TR. One needs to backtrack which phenomena starts first and starts the negative spiral, which is difficult to do in retrospective studies. Nevertheless, previous studies claim that it is the TR that may lead to right ventricular dysfunction [
1,
14,
18].
Moreover, we could also link the longitudinal evolution of TR probabilities to the longitudinal evolution of creatinine. Previous studies also found an association between renal function and TR [
17]. It is still debatable whether it is the TR or right ventricular dysfunction causing the renal dysfunction, however TR may contribute to renal dysfunction by increasing venous congestion [
19] and the combination of TR and right ventricular dysfunction is found to predictive of impaired renal function [
20]. Furthermore, in a recent study Karam et al. noted stabilization of renal function and improvement of liver function in patients undergoing transcatheter tricuspid valve repair, suggesting a beneficial effect of eliminating TR [
21].
4.3 Clinical implications
In most cases TR is managed with diuretics, but in refractory cases a surgical intervention becomes necessary [
1]. Literature regarding surgery for TR after OHT is scarce. Nevertheless, it has been shown that surgery in these patients can be performed safely [
22,
23]. The authors who linked discharge TR to impaired survival suggest to surgically intervene if TR is not resolved by discharge [
5]. However, our data shows that it may be reasonable to wait longer, as probability of TR continues to decrease after discharge, and TR usually remains asymptomatic for years [
4]. Notwithstanding, our data shows that after approximately five years post-OHT the decrease probability of moderate-to-severe TR negates. In patients with persistent TR at five years post-OHT surgical intervention may be most beneficial, assuming the association of TR and mortality/renal function is causal in nature.
A small randomized clinical trial (
n = 60) in which patients received either prophylactic tricuspid annuloplasty vs. no annuloplasty concomitant to OHT noted a better cardiac survival in the annuloplasty groups, if they combined early and late deaths [
18]. No overall survival difference was noted. Furthermore, opportunities arise with emerging trans-catheter devices to treat TR. However, these devices still need to be validated in this complex subgroup of patients. Other authors advocate the use of bicaval anastomosis to prevent TR in the first place [
6,
24].
4.4 Strengths and limitations
The major advantage of this study is that we were able to collect 8826 echocardiograms, enabling us to use advanced statistical methods to model the dynamic nature of TR and making less biased inference of the impact of TR on mortality. Several limitations apply to this study common in retrospective analyses. We were not able to analyze right ventricular function and pulmonary artery pressure due to limited and unreliable data. We did not consider cardiac allograft vasculopathy (CAV) explicitly in this study since CAV is diagnosed by coronary angiography and, therefore, there is a delay between development and diagnosis of CAV. Nevertheless, CAV manifests as LV dysfunction, which we were able to analyze, hence CAV is implicitly considered. Patients who died without an TR measurement on echocardiogram were excluded, which can introduce selection bias. Assessing TR remains challenging, but we dichotomized this variable to create a more robust measurement. Moreover, it was not possible to determine the cause of TR (e.g., biopsy related vs. functional). Lastly, LV function, pacemaker and dialyses were incorporated in the models as a time-varying exogenous variable, while in fact these variables are more likely to be endogenous.
5 Conclusions
TR during follow-up is significantly associated with higher mortality and progressive decline of renal function/ end-stage renal failure. Nevertheless, probability of TR is the highest immediately after OHT and decreases thereafter. Therefore, it may be reasonable to refrain from surgical intervention during early phase after OHT with bi-atrial anastomoses.
PDF
(1169KB)
