Page 166 - Simplicity is Key in CRT

166
very reason for LBBB not showing up as independent predictor of CRT response. Unfortunately, meta-analyses may thus be misleading and should be disregarded in clinical guidelines if essential mechanistic or clinical information is lacking from the studies the meta-analysis is based upon.
When reviewing the methods of classifying ECGs as LBBB in the various trials, it becomes clear that there has been appallingly little attention for this topic. The early landmark trials MIRACLE [5], MUSTIC [6] and CARE-HF [7], and the more recent REVERSE trial [8] do not mention LBBB morphology at all. The COMPANION [9], MADIT-CRT [10] and RAFT [11] trials mention the use of LBBB morphology, but failed to define it. Only subanalyses of some of these trials, specifically addressing the issue of LBBB morphology, accurately define LBBB. However, all these subanalyses (REVERSE [12], MADIT-CRT [13] and RAFT [14] trials) classified LBBB morphology differently. This also becomes apparent from the percentage of patients classified as LBBB, ranging from 60 to 79 percent in these trials. These differences support the results from the analyses in this thesis. While the percentage of patients classified as LBBB according to landmark trial definitions ranged between 75 and 78%, larger variations were found with guideline (ESC and AHA/ACC/HRS) recommended definitions (18 and 75%) and the newly introduced Strauss definition (69%).
The analysis in a large multicentre CRT cohort in the Netherlands in chapter 5 further shows that in each of the four analysed LBBB definitions there is at least one redundant ECG feature that does not have any association with outcome in CRT. Moreover, only three of the eleven ECG features included in any of the LBBB definitions together are independently associated with outcome. This may be of importance as we show in chapter 3 that inter-observer variability was largest in the most complex definition (AHA/ACC/HRS), which was attributed to features that are more sensitive to subjective interpretation, i.e. notching/slurring of the R wave in different leads.
An important consideration in this observation, is that most of the definitions of LBBB have never been developed to predict outcome in CRT, but to merely describe a conduction abnormality. The results in chapter 5, show that only a few ECG features suffice to predict CRT response. It may be worthwhile to investigate the development of a ‘CRT- outcome’ specific marker from the 12-lead ECG. In this regard, modern machine learning techniques may be useful.
In general the conclusions from chapter 3 and chapter 5 stress the need for uniformity and ‘simplicity’ in the definition of LBBB in clinical practice. However even when guidelines and trials would recommend and use a uniform definition of LBBB, it remains a collection of morphological patterns on the 12-lead ECG sensitive to subjective interpretation in clinical practice. Future research should therefore focus on finding a simpler, objective marker for patient selection in CRT.
Predicting outcome in CRT with LBBB
This thesis presents two retrospective studies that investigate the value of LBBB as predictor of CRT response. In chapter 4 of this thesis we show that in a multicentre, international cohort of 316 CRT patients, only the ESC 2009 and 2013 definitions were significantly associated with clinical outcome. In contrast, in chapter 5, analysis in the aforementioned ‘MUG’ cohort consisting of 1492 patients, all LBBB definitions used were associated with the combined clinical endpoint. Although this study confirmed the major differences in populations classified as LBBB by different definitions, it showed that a patient classified as LBBB has a significantly better outcome in CRT than patients classified as non-LBBB, irrespective of the definition used. The difference in final conclusion between the