Cardiac Resynchronization Therapy Improves Survival in Selected Patients with Moderately Impaired Ejection Fraction — the Importance of Left Ventricular Remodelling

This article refers to ‘Long-term outcomes of cardiac resynchronization therapy by left ventricular ejection fraction’ by V. Kutyifa et al., published in this issue on pages 360–369. In this issue of the Journal, Kutyifa et al.1 report the impact of cardiac resynchronization therapy with defibrillator (CRT-D) vs. implantable cardioverter defibrillator (ICD) alone on long-term survival in patients with moderately impaired left ventricular ejection fraction (LVEF) > 30%, mild symptoms [New York Heart Association (NYHA) class I or II], and left bundle branch block (LBBB) enrolled in the Multicenter Automatic Defibrillator Implantation Trial-Cardiac Resynchronization Therapy (MADIT-CRT). Three key findings emerge. First, CRT-D vs. ICD reduced composite all-cause mortality and heart failure (HF) events in the population overall, irrespective of baseline LVEF. Second, the mortality benefit was only apparent in patients with significant reverse remodelling after 1 year of CRT. Finally, HF events were reduced irrespective of both baseline LVEF and subsequent remodelling. To unpack these observations we should revisit the MADIT-CRT trial which randomized 1820 patients with LVEF < 30%, QRS duration > 130 ms, NYHA class I/II symptoms, in a 3:2 ratio to CRT-D (n = 1089) vs. ICD (n = 731). The reduction in the primary endpoint of death or HF hospitalization (17.2% vs. 25.3%; hazard ratio 0.66, 95% confidence interval 0.52–0.84) was driven by a 41% reduction in HF hospitalization with no significant difference in mortality.2 Although the inclusion criteria specified LVEF < 30%, the core laboratory assessment identified one third of patients with LVEF > 30%, creating a subgroup with relatively narrow LVEF range (mean 32.2 ± 1.9%). Although small in the trial, this represents a significant proportion of patients with HF in whom LVEF is normally distributed.3 It is also a group for which therapeutic uncertainty exists, excluded from CRT-D trials requiring LVEF < 30%, and CRT-P trials requiring more severe symptoms. However, the prevalence of LBBB declines as LVEF increases, as illustrated by the challenging recruitment in the Multicenter InSync Randomized Clinical Evaluation MIRACLE EF study.4 The true size of the population with mid-range LVEF and LBBB is uncertain. Prior subgroup analyses from both MADIT-CRT and the REsynchronization reVErses Remodeling in Systolic left vEntricular dysfunction (REVERSE) study reported reverse remodelling and reduced risk of the combined endpoint of death or worsening HF associated with CRT over intermediate follow-up.5, 6 The present analysis for the first time demonstrates survival benefit over longer-term follow-up in patients with mild symptoms and higher LVEF, albeit limited to those with significant remodelling using a landmark analysis starting from 1 year post-implantation. Several important limitations merit consideration. Foremost, the population is highly selected, essentially a subgroup (remodelling, n = 204) of a subgroup (higher LVEF, n = 450) of a subgroup (LBBB, n = 1274) within the trial (n = 1820). The apparent treatment effect size is likely magnified by survivor bias, an unusually high definition of remodelling, and selected long-term follow-up centres. Nevertheless, the study also has important strengths: randomized treatment assignment, a large prospective cohort, and long-term follow-up. As with all subgroup analyses, the intention is not to direct clinical practice but generate hypotheses and questions — in this case under what circumstances is CRT associated with improved long-term survival. The translation of early reduction in HF events and composite outcomes into longer-term reduced mortality is fascinating and underscores the complex and incompletely understood mechanisms of CRT-induced remodelling. Early clinical trials assessed response to CRT relatively early, typically at 3 to 6 months. However, multipoint pacing studies have demonstrated late remodelling among patients initially considered ‘non-responders’ in those randomized to standard biventricular pacing.7 This late remodelling was directly observed in REVERSE, where LVEF improvement was maximal at 2 years.6 Two further observations support these findings. First, the survival curves of CRT vs. no CRT continue to diverge in the trials with longer follow-up with no attenuation of hazard ratio during the extension period.2, 8 Second, in statistical modelling studies the lifespan gain continues to accrue far beyond traditional trial follow-up.9 This long-term survival improvement was arguably evident in the Danish Study to Assess the Efficacy of ICDs in Patients with Non-ischemic Systolic Heart Failure on Mortality (DANISH) in which mortality was unexpectedly low in the context of high CRT utilization and optimal medical therapy.10 We must however be cautious in our interpretation of the survival benefit reported by Kutyifa et al. due to the landmark method employed. Only patients surviving to receive echocardiography at 1 year were included. This survivor bias excludes an unquantifiable group with significant reverse remodelling who died prior to assessment, thus magnifying the apparent treatment effect. The aforementioned lifespan gained study demonstrated a surprising phenomenon which is particularly relevant to the current analysis.9 In the short term, the lifespan gain is greatest in those with highest compared to lowest baseline risk. However, in the longer term, this situation reverses with greatest lifespan gain in those with lowest baseline risk, such as these patients with milder symptoms and higher LVEF. Not only is this population potentially large as outlined earlier, but also has the greatest potential gain, and thus of great interest from both a patient and health system perspective. This underscores the value of the present analysis, and also the pressing need to assess the impact of CRT in populations with milder HF.4 The reduction in HF events with CRT irrespective of dichotomized baseline LVEF or subsequent remodelling initially appears incongruent with the recognized association between LVEF and HF risk.11, 12 However, closer examination reveals patients with more vs. less remodelling had approximately half the rate of HF events (Figure 3 of analysis),1 consistent with previous analyses in the overall MADIT-CRT population.12 That both groups experienced significantly fewer HF outcomes than ICD alone in part reflects greater statistical power, with three-fold more HF events than deaths in patients with milder HF.2 However, it also highlights the discordant timelines and effects of CRT on left ventricular function, HF clinical events and mortality.13 This returns once more to the limitations of conceptualizing ‘response’ to CRT. Simple dichotomization of outcomes is often misleading, as is dividing patients into ‘responders’ and ‘non-responders’. There is a spectrum of clinical, volumetric and prognostic response with no consensus on threshold values in any domain.13 End-systolic or diastolic absolute or indexed volumes, or ejection fraction, with change defined in absolute or relative terms, have all been employed.12-15 Thresholds of approximately 15% improvement are typically applied,15 whereas the authors define remodelling using the median value of 35% reduction in left ventricular end-systolic volume at 1 year.1 While this and previous MADIT-CRT analyses suggest remodelling is a powerful determinant of subsequent mortality,12 no such relationship was observed with remodelling at 3 months in the Cardiac Resynchronization in Heart Failure (CARE-HF) trial.14 Perhaps this is further evidence that reverse remodelling in response to CRT (and indeed in general) is a longer-term process than we anticipate.6 The aforementioned survivor bias presumably also contributes to the discrepancy between MADIT-CRT and CARE-HF. If the effects of CRT are indeed multifaceted and differ between patients,13 then surviving to 1 year may select patients in whom remodelling is more strongly associated with survival. Methodological considerations aside, the important clinical message is that a subgroup of patients with LBBB and more moderately impaired LVEF derive significant morbidity and mortality benefit from CRT in the longer term. In many ways these findings mirror those of the Biventricular versus Right Ventricular Pacing in Heart Failure Patients with Atrioventricular Block (BLOCK HF) trial.15 The intervention in both studies is CRT. In this MADIT-CRT subgroup the control is LVEF > 30% with LBBB (mean QRS 160 ms); in BLOCK HF the control is right ventricular pacing (typically LBBB) and LVEF ≤ 50%. All therapies balance benefit, risk and cost. The results are therefore particularly relevant as CRT complications have declined with quadripolar leads and improved battery longevity, while leadless technology and His bundle pacing increasingly demonstrate promise. The MIRACLE EF study aimed to implant CRT in patients with LBBB and LVEF 36–50% and randomize to CRT with pacemaker on or off in order to avoid the bias of placebo effect.4 Despite the scientific merit, patients and even ethics committees failed to embrace a control arm with risks of surgery but no benefits of therapy, contributing to premature study termination for enrolment futility. The potential solutions are to go big or go small: either use a non-implanted control arm with the completely objective endpoint of all-cause mortality; or use a surrogate endpoint such as remodelling. This thoughtful analysis by Kutyifa et al.1 provides further evidence to support either strategy. Unfortunately the former may be limited by the eligible population size, while the latter rarely impacts practice guidelines or reimbursement. Until these trials are conducted, we should continue to follow the guidelines and avoid the temptation to treat patients with mid-range ejection and LBBB using CRT. Invasive and costly procedures require the highest level of evidence, and the proverbial road to hell is paved with good intentions. Conflict of interest: none declared.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.