Adverse functional remodelling of the subpulmonary left ventricle in patients with a systemic right ventricle is associated with clinical outcome

Abstract

Aims

Early recognition of adverse remodelling is important since outcome is unfavorable once patients with a systemic right ventricle (sRV) become symptomatic. We aimed assessing prognostic markers linked to short-term clinical evolution in this population.

Methods and results

Thirty-three patients (76% male) with sRV (atrial switch repair for D-transposition of the great arteries and congenitally corrected transposition of the great arteries) underwent detailed phenotyping including exercise cardiac magnetic resonance and were followed over mean follow-up time of 3 years. Mean age was 40 ± 8 (range 26–57) years at latest follow-up. Adverse outcome was a composite of heart failure (HF) and tachyarrhythmia. Descriptive statistics and univariate cox regression analyses were performed. When compared with baseline: (i) most patients remained in New York Heart Association functional class I (76%), (ii) the degree of severity of the systemic atrioventricular valve regurgitation rose, and (iii) more electrical instability was documented at latest follow-up. Six (18%) of a total of 9 events were counted as first cardiovascular events (9% HF and 9% arrhythmia). NT-proBNP, oxygen pulse, left ventricle end-diastolic volume index (LVEDVi), and stroke volume index (SVi) of the subpulmonary left ventricle (LV) both in rest and at peak exercise were significantly associated with the first cardiovascular event.

Conclusion

NT-proBNP was by far the best prognostic marker for clinical outcome. Adverse remodelling with increase of LVEDVi and SVi of the subpulmonary LV at rest and during exercise were associated with worse clinical outcome. We theorize that remodelling of the subpulmonary ventricle might be an early sign of a failing sRV circulation.

Introduction

The morphological right ventricle (RV) supports the systemic high-pressure circulation in patients with complete transposition of the great arteries after atrial switch repair (Mustard or Senning) dextro transposition of the great arteries (D-TGA) and congenitally corrected transposition of the great arteries (ccTGA) resulting in a subaortic or systemic RV (sRV) and a subpulmonary left ventricle (LV).^1,2 The number of patients with sRV reaching adulthood increased steadily^3,4 but the major concerns for long-term outcome in those patients are sRV dysfunction, heart failure (HF), arrhythmias, and sudden cardiac death.⁵ Moreover, sRV dysfunction might compromise the subpulmonary LV function as a consequence of a negative ventriculo-ventricular interaction and pulmonary hypertension.^6,7 Although sRV failure is one of the main contributors to mortality and disability in young patients,⁸ little is known about the early stages of remodelling and its potential impact on short-term outcome. Indeed, haemodynamic deterioration may develop gradually and subclinically for a significant period of time (early remodelling) followed by sudden and sometimes unexpected clinical deterioration.⁸ Therefore, early recognition of haemodynamic deterioration has significant clinical importance. Standard echocardiography provides valuable information on the sRV, the systemic atrioventricular valve (SAVV) function, and residual lesions or post-procedural sequellae.⁹ Cardiopulmonary exercise testing (CPET) might help to determine functional capacity and has a predictive value for outcome in sicker patients.¹⁰ The same is true for the use of biomarkers.⁹ However, all this may be not sufficient enough to identify early signs of an sRV failure. Furthermore, assessment of bi-ventricular morphology and function at rest and during exercise has shown benefits in the work-up of congenital heart disease (CHD) patients. For this, cardiac magnetic resonance (CMR) imaging is a modality of choice, because of high feasibility (e.g. acoustic windows). CMR in CHD patients with RV disease should be performed for quantification of RV volumes and RV ejection fraction (RVEF) amongst other things.¹¹ This study aimed to investigate (i) functional, electrical, and echocardiographic changes over time and (ii) to evaluate whether additional functional and geometric parameters, including chamber size, volumes, and function at rest and during exercise, could add value in predicting clinical outcome (death, occurrence of arrhythmia, and HF). Newly identified variables might contribute to optimize follow-up, to customize treatment, and improve outcome.

Methods

Patient selection

Thirty-three detailed phenotyped sRV patients from a previously conducted study were included in this trial.¹² Baseline data were collected from July 2015 to April 2017 and these patients were thereafter systemically followed-up by the adult CHD care programme of the University Hospitals Leuven, Belgium. The local ethics committee of clinical research UZ/KU Leuven approved the baseline study and the follow-up protocol (S57925). At inclusion, all participants signed informed consent.

Initial data collection

For the baseline study, demographic and clinical data [anatomy, age, gender, body surface area (BSA), body mass index, and New York Heart Association (NYHA) functional class] were retrieved from the electronic patient records. Likewise, standard echocardiographic parameters and NT-proBNP levels were collected in all 33 patients at inclusion.

Data from a CPET were collected including peak VO2 (mL/min/kg), % of predicted peak VO2, anaerobic threshold (% of peak VO2), peak power output (Watt), maximal heart rate (HR, bpm), VE/VCO2 slope, blood pressure at rest and peak exercise (mmHg), and oxygen pulse (mL/beat). Oxygen pulse was defined as the ratio of oxygen consumption (VO2, mL/min) to HR (bpm).

CMR at rest and during exercise was performed in all patients.¹³ HR reserve was calculated by the difference between maximum HR and the resting HR (bpm). Stroke volume (SV, mL) was calculated as the difference between end-diastolic volume (EDV, mL) and end-systolic volume (ESV, mL). Left ventricular ejection fraction (LVEF, %) and RVEF (%) were calculated as ratio of (SV/EDV)*100 (%). Stroke volume index (SVi, mL/BSA) was measured as EDVi—ESV index (ESVi) and cardiac index (CI, L/min*BSA) as the product of SVi and HR (bpm). Total end-diastolic volume index (total EDVi, mL/m²) was defined as the sum of EDVi (mL/m²) from both subpulmonary LV and sRV, similar for total ESVi (mL/m²). Contractile reserve (CR) was obtained by the difference between myocardial contractility at maximal exercise capacity and rest (EF at peak exercise—EF at rest, %). The date when the CMR was performed, was considered as the start of the follow-up.

Follow-up data and endpoints

For the follow-up, all medical records were reviewed from the hospital’s electronic database. Data from follow-up visits and hospitalizations were collected, which consisted of clinical data, age, NYHA functional class, BSA (kg/m²), the electrocardiogram, and standard echocardiography. Electrical instability was specified as loss of sinus rhythm, lowering HR, wider QRS complex, and more prevalent fragmented QRS (fQRS).¹⁴

All cardiovascular endpoints were analysed: HF, arrhythmia, and sudden death. HF was defined as (i) the need for hospitalization with initiation of diuretic therapy, (ii) initiation of standard HF treatment according to the ESC guidelines, or (iii) the presence of clear ventricular dysfunction on transthoracic echocardiography with clinical signs and symptoms of HF. Supraventricular tachycardia was defined as a new episode of small QRS tachycardia captured on a 12-lead electrocardiogram and the need for (i) direct current cardioversion or (ii) adaptation of medical treatment. Non-sustained ventricular tachycardia was defined as a new episode of ventricular tachycardia captured on 12-lead electrocardiography with a HR of at least 120 bpm lasting for at least 3 beats and persisting <30 s. Out of hospital cardiac arrest was defined by failing cardiac mechanical activity and no signs of blood circulation. In the end, HF and tachyarrhythmia’s were considered as combined cardiovascular endpoint.

Statistical analysis

Continuous variables were tested for normal distribution with the Kolmogorov–Smirnov test and presented as mean ± standard deviation (SD) or as median (minimum–maximum range) as appropriate. For categorical variables, frequencies and percentages were used. To compare frequencies, a χ² test or Fisher’s exact test was performed. For paired frequencies, a McNemar test was applied. Differences in the same group were calculated using paired t-test, where differences between groups for continuous variables were analysed with an unpaired t-test, Kruskal–Wallis H test, or Wilcoxon–Mann–Whitney test where applicable. Spearman’s test was applied to assess the correlation between the indices SVi during exercise and the degree of SAVV regurgitation. For outcome analysis, baseline characteristics at the time of CMR were associated with the first event occurring. All statistical tests were two-sided, and significance was defined as P < 0.05. For all the comparative variables with P-value <0.1, univariate cox regression analysis was performed. Multivariate cox regression analysis was not performed because of the low number of events and the small sample size. These analyses were performed using IBM SPSS Statistics, version 26 (Armonk, NY: IBM Corp) for Windows.

A more complex multidimensional method that can handle small sample sizes, called a support vector machine (SVM), was also conducted.¹⁵ As input served all variables with a P < 0.05 resulting from the univariate cox regression. An SVM model searches for the hyperplane in a multidimensional space that best segregates the two classes. When unable to separate linearly, the model transforms the data to a higher dimension space in order to find a linear separation. Several kernels exist to achieve this transformation. They each define a different shape of the decision boundary in the original space. In this study, the linear, Gaussian, and polynomial kernels were tested. To validate the SVM models, the leave-one-out cross-validation method was applied. Accuracy was used as performance metric to compare the different models. The SVM analysis was implemented in Python 3.6 and documented in Jupyter Notebook.

Results

Patient characteristics

Thirty-three patients with a male preponderance (76%) and mean age of 40 ± 8 (range 26–57) years at latest follow-up were observed. Mean follow-up time since CMR was 3.0 ± 0.6 (range 1–4) years.

Follow-up data

At latest follow-up, average functional class improved (more patients in NYHA I compared to baseline) and no patients were in NYHA functional class IV. During follow-up, body weight increased. At latest follow-up, ECG variables were significant different from baseline with a slower HR, loss of sinus rhythm, longer QRS duration, and more prevalent fQRS were found. Echocardiography showed that both tricuspid annular plane systolic excursion (TAPSE, mm) and right ventricular fractional area change (RV FAC, %) were significantly higher when compared to baseline. Likewise, more severe regurgitation of the SAVV was found. CPET—analyses of this group at baseline illustrated an overall well-preserved exercise capacity for sRV patients. These data are summarized in Table 1.

Outcome

The maximal follow-up time was 4 years. The total number of events for the entire cohort was 9 (27%), of which 6 (18%) were counted as a first cardiovascular event. Specifically, 9% suffered from an episode of HF and in 9% had an episode of tachyarrhythmia. The occurrence of HF and/or arrhythmias were considered as combined clinical endpoint. Three patients had recurrent events. Data are summarized in Table 2. No patients died during follow-up or needed heart transplantation. Mean age at first event was 40 ± 11 (range 25–52) years and mean time between first and second event was 1 ± 0.6 (range 0–2) years.

Patient characteristics in the event and non-event group are summarized in Table 3. Neither TAPSE nor RV FAC differed between groups. Percentage of predicted peak VO2 and the VE/VCO2 slope was similar in both groups. However, oxygen pulse was significantly higher in the event-group. CMR measures at baseline showed that the mean HR was slower in the event group at rest and during peak exercise. Compared to the event-free patients, LV EDVi and SV LVi at rest were significantly larger. At maximum exercise, for those patients who suffered from a cardiovascular event, RVEF and LVEF or CR did not vary, but significantly higher values for left ventricle end-diastolic volume index (LVEDVi), SV LVi, and total ESVi were observed.

On univariate Cox analysis NT-proBNP [HR 11.02 (95% CI 1.296–93.662), P = 0.028], oxygen pulse [HR 1.202 (95% CI 1.012–1.428), P = 0.037], LVEDVi in rest [HR 1.046 (95% CI 1.002–1.092), P = 0.041], and during exercise [HR 1.035 (95% CI 1.002–1.069), P = 0.038], SVi of the subpulmonary LV in rest [HR 1.154 (95% CI 1.005–1.322), P = 0.038], and at peak exercise [HR 1.065 (95% CI 1.007–1.125), P = 0.026] were significantly associated with the first cardiovascular event (Figure  1A and B). Table 4 shows the average accuracies obtained for the different SVM models with the leave-one-out cross-validation method. For an SVM, with a Gaussian kernel the highest accuracy, namely 0.91, was found. In Table 5, the real vs. the predicted points of this best model are listed. In conclusion, this model with complex multidimensional method predicted 4 out of 6 six events with SVM model with a Gaussian kernel and high accuracy of 91%.

Discussion

This short follow-up study with 33 detailed phenotyped sRV patients including exercise CMR, observed that clinical outcome is associated with baseline (i) elevated NT-proBNP levels, (ii) increased oxygen pulse on CPET, and (iii) adverse remodelling with increased subpulmonary LV EDVi and SVi on CMR. Following findings did not align to the evolution of the study cohort: (i) the functional capacity improved during follow-up and (ii) there was no decline of the sRV function measured by transthoracic echocardiography. In contrast, TAPSE and RV FAC even increased. Only the degree of the severity of the SAVV regurgitation rose. Remarkably, the study results put us in a dilemma. Despite no clinical deterioration nor reducing sRV systolic function, events did occur during follow-up. This data suggest that standard daily practice parameters are insufficient to predict pre-clinical deterioration. We wondered whether detailed phenotyping of patients with an sRV circulation might have an added value in predicting outcome. Biomarkers in combination with mechanical factors at rest and during exercise could be of added value. This data implies the importance of the frequently overlooked subpulmonary LV.

This study indicates that during follow-up, the proportion of patients with severe SAVV regurgitation increases, which is in line with previous studies.¹⁶ Moreover, quite a number of patients experienced electrical instability over the years. It is, however, of interest that functional status and the sRV function remained unchanged also found in other studies.¹ Similarly, parameters reflecting sRV function were stable or even improved during follow-up, despite the occurrence of events.

Over a follow-up time of 3 years, almost 1 in 5 patients experienced a significant clinical event, either tachyarrhythmia (9%) (1 out of hospital arrest in cc-TGA, 1 patient with nsVT, and 1 with an atrial flutter, both in D-TGA) or HF episode (9%) and these findings are comparable with previous studies.¹⁷ The occurrence of HF and/or arrhythmias were considered as combined clinical endpoint. Other events such as atrial baffle problems (leakage and/or obstruction) and/or the development of pulmonary arterial hypertension⁵ that may occur in sRV patients were not found, potentially due to the small sample size of the study. So, despite reasonable and stable functional status and sRV function, based on standard follow-up criteria, a significant proportion (18%) encounters a clinical event, underscoring the need for better risk stratification. It was questioned whether parameters from deeper phenotyping would have an added value.

This data indicated that NT-proBNP, oxygen pulse, LVEDVi, and SV LVi at rest and at peak exercise were significantly associated with worse clinical outcome. Log NT-proBNP reflects neurohormonal activation and has shown in prior studies to be strongly related with outcome in patients with an sRV.^7,9,18

Oxygen pulse is considered as a surrogate marker of the effective SV in patients with normal arterial oxygen content. This was confirmed on CMR where increased SVi at rest and peak exercise related to worse outcome. Moreover, a higher LVEDVi was associated with worse outcome. It is remarkable that in this small group of relatively stable sRV patients with a short follow-up period, factors related to the subpulmonary ventricle predict outcome rather than sRV dysfunction itself. Indeed, authors have shown that subpulmonary LV dysfunction relates to worse outcome in this group of patients.¹⁹ Similarly, a ‘normal-looking’ subpulmonary LV in sRV patients is often indicative of elevated pulmonary artery pressures. Therefore, it is maybe not unexpected that higher LVEDVi was associated with worse outcome in this group of sRV patients reflecting early subpulmonary LV remodelling prior to overt subpulmonary LV dysfunction. The finding that a higher SVi (which is confirmed by a higher peak oxygen pulse on CPET) relates to worse outcome may be counter intuitive. This either reflects the dilating subpulmonary LV and/or chronotropic incompetence, since CI remains unchanged. In contrast to prior studies, there was no relationship between peak oxygen uptake, SAVV regurgitation, CMR LVEF, RVEF, CR, and outcome,²⁰ related to sample size, shorter follow-up, and/or better overall clinical status. Despite small sample size, more complex multidimensional method that can handle small sample sizes predicted 4 out of 6 events with SVM model with a Gaussian kernel and high accuracy of 91%. To our knowledge, there are no previous studies associating oxygen pulse, LVEDVi, and LV SVi as predictors for outcome in sRV patients.

In summary, we know that sRV dysfunction can progress subclinically during a long period of time and then result in a rapid and unexpected development of congestive HF.⁸ The findings of this study illustrate that even before the rise of symptoms (NYHA and peak oxygen uptake), adverse remodelling of the subpulmonary LV is correlated with worse outcome. In other words, morphological changes in EDVi and SVi of the subpulmonary LV at rest and during exercise could have a prognostic value. Improvement of functional class is remarkable but better lifestyle behavior can be translated in measurable parameters. In a structurally normal heart, the RV was forgotten until recent studies showed evidence that the RV is a barometer for cardiac outcomes across a range of pathologies.^21,22 Similarly, the focus in the evaluation of the sRV has centered on the sRV physiology and function. Until now, the subpulmonary LV is somewhat ignored. Therefore, we conclude that adverse remodelling of the subpulmonary LV might be the first stage of a failing sRV circulation even before developing symptoms and noticeable sRV dysfunction (Figure 2). That is why daily practice parameters have not always a predictive value in early disease progression and that it is crucial to remind that the remodelling of the subpulmonary LV is associated with ventricular dysfunction, arrhythmias, and poor prognosis.²³ Nevertheless, biomarkers seem to be most accurate tool to detect pre-clinical deterioration. According to the guidelines, regular measurements of the biomarkers is recommended. However, this research data open the perspective that the increase of the biomarkers do not find their origin in the sRV, but in the subpulmonary LV.

Study limitations

This was a single-institution cohort study. Adults with sRV are only a small proportion of all CHD patients, making large studies and subgroup analyses difficult and is the main limitations of the work. We did not have systematically assessed NT-proBNP, other biomarkers, or CPET during follow-up. Invasive haemodynamics were not systematically conducted in all patients. fQRS was not significantly higher in the event group although previous research showed that appearance of QRS fragmentation late after Mustard/Senning repair is associated with adverse outcome.¹⁴ Presumably, this is secondary to the small sample size. Same for TAPSE and RV FAC that not differed between the two groups. Cox regression analysis of log NT-proBNP showed a broad confidence interval. Furthermore, we did not discuss the diastolic function of the ventricles. Although strain analysis and fractional area change of the subpulmonary LV might be much more sensitive to predict early deterioration of the sRV, we did not include these variables¹⁹ since a recent study showed only limited value of strain analysis.²⁴ Exclusive to this study was that we associated in detail CMR measurements at rest and during exercise with outcome but there was no follow-up CMR conducted.

Conclusions

NT-proBNP was the best prognostic marker for clinical outcome. Adverse remodelling of the subpulmonary LV with LV dilatation and increase of LVEDV index and SV index at rest and during exercise are associated with worse clinical outcome. We hypothesize that adverse remodelling of the subpulmonary LV could be the first stage of the failing sRV circulation. Daily practice parameters have poor predictive value.

Acknowledgement

The authors would like to thank all the co-authors for their valuable input and expertise.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Funding

This research received project funding by KU Leuven (grant no. C2-1700417).

Conflict of interest: none declared.

References