Article Text
PDF
Abstract
Objective To determine the association of pulmonary valve replacement (PVR) with death and sustained ventricular tachycardia (VT) in patients with repaired tetralogy of Fallot (rTOF).
Methods Subjects with rTOF and cardiac magnetic resonance from an international registry were included. A PVR propensity score was created to adjust for baseline differences. PVR consensus criteria were predefined as pulmonary regurgitation >25% and ≥2 of the following criteria: right ventricular (RV) end-diastolic volume >160 mL/m2, RV end-systolic volume >80 mL/m2, RV ejection fraction (EF) <47%, left ventricular EF <55% and QRS duration >160 ms. The primary outcome included (aborted) death and sustained VT. The secondary outcome included heart failure, non-sustained VT and sustained supraventricular tachycardia.
Results In 977 rTOF subjects (age 26±15 years, 45% PVR, follow-up 5.3±3.1 years), the primary and secondary outcomes occurred in 41 and 88 subjects, respectively. The HR for subjects with versus without PVR (time-varying covariate) was 0.65 (95% CI 0.31 to 1.36; P=0.25) for the primary outcome and 1.43 (95% CI 0.83 to 2.46; P=0.19) for the secondary outcome after adjusting for propensity and other factors. In subjects (n=426) not meeting consensus criteria, the HR for subjects with (n=132) versus without (n=294) PVR was 2.53 (95% CI 0.79 to 8.06; P=0.12) for the primary outcome and 2.31 (95% CI 1.07 to 4.97; P=0.03) for the secondary outcome.
Conclusion In this large multicentre rTOF cohort, PVR was not associated with a reduced rate of death and sustained VT at an average follow-up of 5.3 years. Additionally, there were more events after PVR compared with no PVR in subjects not meeting consensus criteria.
- cardiac magnetic resonance (cmr) imaging
- congenital heart disease
- tetralogy of fallot
Statistics from Altmetric.com
Introduction
Pulmonary regurgitation (PR) occurs frequently after surgical repair of tetralogy of Fallot (TOF) and often leads to right ventricular (RV) dilation.1 Repaired TOF (rTOF) patients with progressive RV dilation and dysfunction are at risk for arrhythmias, heart failure and sudden cardiac death, generally beginning in the third decade of life.1 2Pulmonary valve replacement (PVR) is highly effective in treating PR and reducing RV volumes in the short term3 4; yet prognostic benefits of PVR have not been demonstrated in a prospective randomised clinical trial. Moreover, studies comparing outcomes of PVR versus conservative management in matched cohorts found no clear differences in outcomes.5–7 However, these studies were not adequately powered and lacked quantitative assessment of the RV.5 6 Consequently, the role and timing of PVR in rTOF remain a subject of debate.8–10 Current European and American guidelines suggest PVR in asymptomatic patients with worsening haemodynamic parameters, but offer no criteria with specific cut-offs to aid in the timing of PVR.11 12
Several studies reported preoperative RV volume thresholds to achieve postoperative normalisation of RV volumes and/or function following PVR.3 4 13 These observations led some authors to suggest that PVR should be performed early on, using more proactive criteria, to prevent irreversible RV dysfunction and reduce the risk of arrhythmias.14 15 On the other hand, PVR is associated with several late risks including prosthetic valve dysfunction requiring reintervention4 16 and endocarditis.17 Furthermore, patients remain at risk for RV deterioration and clinical events during mid-to-late follow-up after PVR.18 19 Most importantly, patients with limited RV dilation and dysfunction in the setting of PR, who might be considered for early PVR, are at low risk of adverse events or progressive RV deterioration when managed conservatively.20 21 Therefore, optimal timing and specific criteria for PVR remain unknown.
The objectives of this study were to determine the association of PVR with freedom from major adverse outcomes in patients with rTOF and to explore whether there might be a clinical benefit in patients who receive PVR according to consensus guidelines.
Methods
Subjects
A detailed description of the International Multicenter TOF Registry (INDICATOR), including recruitment protocol, inclusion and exclusion criteria, and data collection and analysis in the core laboratory has been published.22 Briefly, subjects were included by participating centres using the following inclusion criteria: (1) rTOF; (2) cardiovascular magnetic resonance (CMR) completed between 1997 and 2010; (3) 12-lead ECG within 1 year from CMR and (4) clinical follow-up ≥1 year or occurrence of a primary outcome. INDICATOR is being updated at regular intervals and more patients have been included since the trial design and original paper.20 22 Additional inclusion criteria for this analysis included (1) CMR measurable RV volumes from cine steady-state free precession (available since 2002); (2) no history of PVR prior to qualifying CMR and (3) in subjects who underwent PVR (surgical or percutaneous), CMR within 3 years prior to surgery. Subjects who had experienced sustained ventricular tachycardia (VT) or aborted sudden death prior to the qualifying CMR were excluded. Follow-up began at the date of the qualifying CMR.
Data collection
For the INDICATOR registry, deidentified data were sent to the data coordinating centre, which included a data repository, CMR core laboratory and a statistical core.22 Each participating centre updated clinical events until last hospital visit or death, including clinical events that occurred after PVR, using case report forms. A 1.5 T MRI scanner was used for CMR studies by participating centres using a similar imaging protocol.22 Deidentified digital copies of the CMR examinations were sent to the CMR core laboratory for analysis.22 For all CMRs performed, RV and left ventricular (LV) volumes and mass were measured in the ventricular short-axis plane as described by Alfakih et al.23 Biventricular stroke volumes and ejection fractions (EFs) were calculated. Volumes were adjusted for both body surface area (BSA) and BSA1.3 to account for variances across a population that includes paediatric patients, as previously reported.24
Consensus guideline criteria for PVR
Predefined guideline criteria for PVR were characterised as ‘proactive’ or ‘conservative’ approaches based on available literature and by consensus of authors (table 1). These approaches were then used to examine potential associations between PVR and the defined outcomes within guidelines subgroup (criteria met/not met).13 18 20 25
Consensus guideline criteria for PVR
In a total of 252 subjects with missing PR fraction, and 1 with missing body mass index, PVR criteria status could not be obtained. All 170 subjects with PR fraction <25% were classified as not meeting PVR criteria.
Outcomes
The composite primary outcome consisted of all-cause mortality, aborted sudden cardiac death or sustained VT (defined as VT lasting >30 s or requiring cardioversion). The composite secondary outcome consisted of advanced heart failure class (New York Heart Association III or IV), non-sustained VT (<30 s) or sustained supraventricular tachycardia (ectopic atrial tachycardia, atrial flutter or atrial fibrillation). For both outcomes, the time to the earliest occurring component defined the time to qualifying event.
Propensity score
A propensity score for PVR was created to adjust for baseline differences between PVR and non-PVR subjects (see online supplementary tables 1,2 and table notes). A logistic regression model with PVR as the outcome was used to generate propensity scores for all subjects. Variables that are potential arguments for PVR (site, age at repair, era of repair (before or after 1980), age at CMR, RV end-diastolic volume (EDV)/BSA, RV end-systolic volume (ESV)/BSA, RV mass/volume, RV EF, LV EF, QRS duration) were included in the model. All variables in the propensity model were available in >95% of subjects: mean or simple regression imputation was used for missing variables (online supplementary table 2).
Supplementary file 1
Statistical analysis
Categorical data were described as number with frequency and continuous data as median with IQR or mean with SD, as appropriate. Differences in baseline characteristics between PVR and non-PVR subjects and subjects with and without primary and secondary outcome were analysed by independent samples t-test, Wilcoxon rank-sum test, Fisher’s exact test or χ2 test, as appropriate. Time to the occurrence of the primary and secondary outcomes was analysed using multivariable Cox hazards regression analysis with PVR (time-dependent) and propensity score as covariates. In addition, other variables related to the outcome (eg, arrhythmias) after stepwise forward selection were included as time-dependent covariates in the survival model. A secondary analysis was performed using 1:1 matching on propensity score, with difference in score of <0.09 (a quarter of the mean difference) between PVR and non-PVR pairs. A frailty failure time model was used to estimate the association of PVR with outcomes in the 1:1 matched cohort, with adjustment for other risk factors. The frailty model takes the matched design into account by assuming correlation between events within the same pair, rather than assuming independence. Differential association between predefined guideline (proactive and conservative) criteria status and PVR with respect to the primary and secondary outcomes was examined in subjects for whom data regarding PVR criteria were available, using a test of interaction between PVR criteria (met/not met) and PVR status (time-dependent).
Statistical analyses were performed using SAS V.9.3 and R V.3.2.1. A two-sided P value<0.05 was considered to indicate statistical significance.
Results
Of the 1358 subjects enrolled in the INDICATOR cohort, 977 met the study inclusion criteria (figure 1 and table 2). A total of 440 subjects underwent PVR (396 (90%) surgical, 44 (10%) percutaneous) 1.02±0.87 years after the qualifying CMR. There were three patients with the primary outcome (two deaths, one sustained VT) within 1 month after PVR. Compared with subjects who did not undergo PVR, those who did were younger at first repair and at CMR, had larger BSA-indexed ventricular volumes, lower mass-to-volume ratio, lower RV and LV EF, and longer QRS duration (online supplementary table 1). The propensity score differentiated adequately (C-statistic 0.85) between subjects with PVR (mean 0.65±0.24) and those without PVR (mean 0.29±0.23) (online supplementary table 2).
Baseline characteristics according to primary outcome (death/ventricular tachycardia (VT)) status
Inclusion flow chart of all subjects included in the INDICATOR cohort and reasons for excluding for analysis in the present study, divided according to pulmonary valve replacement (PVR) status. CMR, cardiovascular magnetic resonance; RV, right ventricular.
Primary outcome
Of the 977 subjects, 41 (4%) experienced the primary outcome during mean follow-up of 5.3±3.1 years. Of those, 30 died (sudden cardiac in 5, heart failure in 3, cardiac other in 1, non-cardiac other in 9 and unknown in 12), 6 experienced sustained VT and 5 had resuscitated sudden cardiac death. Of the 41 subjects who experienced the primary outcome, 15 had PVR prior to the outcome and 26 did not. Table 2 summarises factors associated with the primary outcome. When PVR and propensity score were forced into the final multivariable model, the HR for the PVR group was 0.65 (95% CI 0.31 to 1.36; P=0.25) (table 3), suggesting a non-significantly lower event rate of the primary outcome after adjusting for other factors. In a propensity score-matched analysis, a total of 512 subjects (256 PVR and 256 non-PVR) were included. A total of 184 PVR subjects (with eight events) could not be matched due to higher propensity score. After correction for atrial arrhythmia, RV EF and RV mass/volume, there was a non-significantly lower event rate after PVR in the matched cohort (HR 0.63; 95% CI 0.25, 1.62; P=0.34) (table 4).
Multivariable Cox regression model of primary outcome in full cohort (n=977; 41 events)
Multivariable Cox regression model of primary outcome in 1:1 matched cohort (n=512; 23 events)
Subjects with rTOF and RV-PA conduit (n=162, 17%) had different clinical characteristics than those without a conduit, including increased primary outcome rates (HR 2.34; 95% CI 1.21 to 4.53; P=0.01 in univariate analysis), less PR (mean 28%±15% vs 37%±17%; P<0.001) and increased RV mass-to-volume ratio (0.31±0.13 vs 0.21±0.07; P<0.001). However, in the full cohort, the association of PVR with the primary outcome was similar in subjects with RV-PA conduit compared with those without (interaction P=0.88).
Secondary outcome
A total of 93 subjects who had a history of heart failure (NYHA class III–IV), atrial arrhythmia or non-sustained VT prior to the qualifying MRI were excluded from the secondary outcome analysis. Of the 884 subjects included in the secondary outcome analysis, 88 experienced the outcome (6 had heart failure, 25 atrial flutter, 7 atrial fibrillation and 50 experienced non-sustained VT) during follow-up. The secondary outcome occurred in 44 of the 393 PVR subjects (34 occurred after PVR) and in 44 of the 491 non-PVR subjects.
In univariate analysis, PVR was associated with a higher hazard of the secondary outcome (HR 1.66; 95% CI 1.07 to 2.60; P=0.03). After correcting for propensity score and additional covariates, the higher hazard of event in PVR subjects was no longer statistically significant (HR 1.43; 95% CI 0.83 to 2.46; P=0.19) (online supplementary table 3a). A propensity score-matched cohort (418 subjects) revealed similar results (covariate-adjusted HR 1.28 for PVR-group; 95% CI 0.65 to 2.54; P=0.48) (online supplementary table 3b).
Timing of PVR
In an analysis of the potential clinical benefit according to predefined criteria, 724 subjects with data to assess guideline criteria experienced 25 primary outcomes. A total of 32 subjects with PR <25% underwent PVR, presumably for RV outflow tract obstruction, and were classified as not meeting proactive or conservative criteria. Of 724 subjects, 41% met the proactive criteria definition and 23% met the conservative criteria definition.
The association between PVR and the primary outcome did not vary according to meeting versus not meeting either conservative or proactive criteria subgroup (interactions P=0.17 and 0.28, respectively) (figure 2). However, we found non-significantly higher event rates after PVR when conservative criteria (HR 1.92; 95% CI 0.66 to 5.61; P=0.23) or proactive criteria (HR 2.53; 95% CI 0.79 to 8.06; P=0.12) were not met and no significant difference when criteria were met (figure 2).
Event rate according to proactive and conservative criteria and pulmonary valve replacement (PVR) status Event rates according to prespecified PVR and proactive/conservative criteria subgroups. *The PVR propensity-corrected HR of event (PVR compared with no PVR) for all patients is displayed as the overall estimate for the primary and secondary outcome. The unadjusted HR of event within subgroups is also displayed in this figure including the interaction test P value.
For the analysis of PVR criteria and the secondary outcome, 655 subjects were included and experienced 61 secondary outcomes. Patterns were similar to those for the primary outcome (figure 2). Although there was low statistical power to detect a differential impact of PVR in subjects who did and did not meet guidelines criteria (ie, non-significant interaction terms, P=0.09 for conservative and P=0.36 for proactive criteria), there were subgroup findings of clinical interest. In subjects who did not meet conservative criteria, there was a higher hazard of the secondary outcome after PVR compared with no PVR (HR 2.35; 95% CI 1.24 to 4.44; P=0.009). In contrast, there was no association between PVR and the secondary outcome when conservative criteria were met (HR 0.91; 95% CI 0.35 to 2.39; P=0.85). Similarly for proactive criteria, there was no significant difference in secondary outcome events rates after PVR compared with no PVR when criteria were met, and there was a higher event rate after PVR when criteria were not met (HR 2.31; 95% CI 1.07 to 4.97; P=0.03). These findings were robust to adjustment for propensity score and other relevant covariates.
Discussion
In this large multicentre cohort of mostly young adults with rTOF, PVR was not associated with a reduced risk for adverse clinical outcomes during a follow-up period of 5.3±3.1 years. In exploratory analysis investigating different strategies of PVR timing, we found a weak, statistically non-significant signal suggesting a lower rate of death/sustained VT in patients meeting conservative criteria. Our results underscore the observation that the rate of death or sustained VT in this population is low and even a large cohort such as INDICATOR is insufficient to provide conclusive evidence to support a proactive versus a conservative approach to PVR. Specifically, it is conceivable that with a substantially larger cohort and longer follow-up, evidence of treatment benefit or harm may emerge.
Our results are in line with several, smaller previous reports, in which PVR was not associated with improved outcomes.5–7 Other studies observed more supraventricular tachyarrhythmias following PVR (15% vs 6%)5 and an increased rate of non-sustained VT after percutaneous Melody implantation.26
To our knowledge, the finding that ‘early’ PVR, in patients not yet fulfilling proactive criteria, was associated with increased risk of heart failure, atrial arrhythmia and non-sustained VT has not been previously reported. One may speculate that the higher event rate is due to several factors: (1) patients without PVR with similar haemodynamic status had a very low clinical event rate, suggesting there is no clear clinical need for valve implantation. In these patients, RV function may be sufficient to maintain adequate cardiac output; (2) although PVR reduces RV volumes, its effects on RV EF, dyssynchrony and fibrosis in patients with a modest disease burden are likely small or non-existent.4 27 28 These factors may be more important determinants of clinical outcomes in patients with limited RV dilation; (3) surgery is associated with risks4 and may have detrimental short-term impact on RV function.29 Furthermore, myocardial incisions may also create new arrhythmia substrates; and (4) bioprosthetic valves used for PVR typically deteriorate over time,16 which may be associated with progressive RV dilatation and dysfunction,19 with likely multiple reinterventions needed in patients with early referral for PVR as observed in this cohort. Percutaneous PVR may avoid some of these factors, but haemodynamic response is likely similar. The number of patients with percutaneous PVR was still limited in our study; future studies may investigate whether a percutaneous approach may alter timing of PVR.17 26
It is worth noting that our study aimed to evaluate clinical outcome after PVR compared with no PVR. Although we observed no statistically significant differences, we did not determine the effect of PVR on subjective symptoms or exercise capacity.4 PVR can be considered in patients with symptoms attributed to RV volume overload, according to current guidelines.10–12
The outcomes of different PVR timing strategies should be confirmed over a longer period of time, as the beneficial haemodynamic effects of PVR may translate into better clinical outcomes after a longer follow-up duration.4 18 Clinically significant (35% reduction in risk in the overall cohort), although not statistically significant, trends after 5 years of follow-up may suggest that beneficial effects of PVR in the more severely affected patients (fulfilling conservative criteria) could manifest after a longer period than in our study. However, this needs to be evaluated in future studies as unfavourable effects of bioprosthetic valve deterioration, endocarditis and reinterventions are also likely to increase in the PVR group.16–19
Limitations
This study is limited by its observational cohort design. Although we employed propensity score adjustment to minimise bias in causal inferences, such inferences should still be treated with caution and ideally validated in a randomised trial. The cohort is further restricted to subjects who have undergone CMR; this limitation is partially mitigated by the routine use of CMR in this patient group at the participating centres. However, inclusion of subjects with CMR allowed correction for baseline differences using a PVR propensity score. The PVR propensity score was also used to account for differences between centres in referral for PVR. The event rate was low (4% for the primary outcome), which limits statistical power, particularly for the detection of differential associations of PVR and outcome according to treatment guidelines subgroup. An additional limitation is that there is no standardised protocol for diagnostic testing at these centres. Specifically, exercise testing and echocardiography were not available in many subjects, and PR fraction data were missing in some. We could not ascertain whether pulmonary stenosis (PS) or endocarditis was also present in some patients, but sensitivity analysis excluding patients with limited (<25%) PR revealed similar results. We did not investigate PVR according to symptoms as this is recommended by current guidelines and not necessarily dependent on clinical benefits.11 12 Because this is one of the largest cohorts available for this population, we have presented all subgroup effect estimates with their CIs and P value for use in future hypothesis generation, regardless of interaction test results.
Conclusions
In this large multicentre cohort of patients with rTOF with an average follow-up of 5.3 years, PVR was not associated with a reduced risk for adverse clinical outcomes, including death and sustained VT. With regard to different strategies of PVR timing, we found more clinical events if PVR was performed without meeting consensus treatment criteria, but this result should be confirmed over a longer period of time to inform optimal timing for this intervention.
Key messages
What is already known on this subject?
Pulmonary valve replacement (PVR) is highly effective in treating pulmonary regurgitation and reducing right ventricular volumes in patients with repaired tetralogy of Fallot. However, it is not known if PVR is also associated with improved clinical outcomes.
What might this study add?
In this large cohort study, PVR was not associated with a reduced rate of death or sustained ventricular tachycardia in a propensity-adjusted analysis during approximately 5 years of follow-up. Additionally, there were more events after PVR in subjects not meeting consensus criteria for PVR.
How might this impact on clinical practice?
Our findings suggest a higher event rate after PVR in patients not meeting treatment criteria. Importantly, this result should be confirmed over a longer period of time to inform timing of this intervention.
References
Footnotes
Contributors All authors attributed to both the conception, design, critical revision and final approval of this manuscript. LAS and ML performed statistical analyses. JPB interpreted the data and drafted the manuscript under supervision of senior authors BJMM, AMV and TG.
Funding This work was supported by the Netherlands Heart Institute (NL-HI) and Nuts Ohra foundation. The work described in this study was carried out in the context of the Parelsnoer Institute. Parelsnoer Institute is part of and funded by the Dutch Federation of University Medical Centers. This work was also supported by the Higgins Family Noninvasive Research Fund at Boston Children’s Hospital, the Lerner Research, NIH/NHLBI 1 R01HL089269-01A2, and British Heart Foundation (SVB-N; FS/11/38/28864). SVB-N, REW and MAG were supported by the NIHR Cardiovascular Biomedical Research Unit of Royal Brompton and Harefield NHS. RMW was supported by the Canadian Institutes of Health Research MOP 119353. Foundation Trust and Imperial College London. This report is independent research by the National Institute for Health Research Biomedical Research Unit Funding Scheme.
Competing interests None declared.
Ethics approval Boston Children’s Hospital and participating centers.
Provenance and peer review Not commissioned; externally peer reviewed.