Response to Letter to the Editor: “Clinical Implication of an Elevated Circulating Galectin-3 Level on Predicting the Occurrence of Ventricular Arrhythmias”

The Authors Reply We thank Nakagawa et al. for their letter identifying several important concerns regarding our study (1), namely the inclusion of secondary prevention patients and patients who received cardiac resynchronization therapy (CRT) in the cohort and the lack of a time-dependent receiver operating characteristic (ROC) analysis.

The pros and cons of enrolling secondary prevention and CRT patients were discussed at the planning stage of this study. If serum galectin-3 (Gal-3) reflects ongoing ventricular remodeling, such as myocardial inflammation and fibrotic processes (2), it may be possible to predict the prognosis independently of history of ventricular arrhythmia and CRT. In fact, the results of the 2014 MADIT-CRT substudy suggested that CRT did not reduce Gal-3 levels compared to implantable cardioverter-defibrillator (ICD) (3). Therefore, to investigate the usefulness of Gal-3 as a biomarker with an absolute value, the decision was made to include these patients (4). However, to exclude the effects immediately after CRT implantation and in the subacute phase following the occurrence of ventricular arrhythmias, only patients for whom at least six months had passed since implantation were enrolled.

As shown in the results, Gal-3 was an independent prognostic factor, even if the history of ventricular arrhythmias was included as a factor in the multivariable Cox analysis. Although CRT was not listed in the table, CRT was not a significant predictor of the new occurrence of ventricular arrhythmias or the exacerbation of heart failure.

Predicting the occurrence of ventricular arrhythmias in patients with heart failure is not easy. At present, ICD implantation is the main treatment. However, it is also known that patients with recurrent sustained ventricular arrhythmias have a poor prognosis (5). Based on this knowledge, it is worth identifying cases with an increased risk of developing ventricular arrhythmias, even after ICD implantation, in order to optimize treatment. The clinical implication of this study was not the avoidance of ICD implantation in patients classified as low risk by Gal-3. The purpose was to examine the additional usefulness of Gal-3 in risk stratification.

However, we did not assess whether or not the threshold of Gal-3, which predicted the increased incidence of ventricular arrhythmias or heart failure hospitalization, differed between primary and secondary preventive patients and between CRT-treated and non-CRT-treated groups. We agree that this is a limitation of this study and that larger-scale research is required to further evaluate this.

This study investigated the relationship between the Gal-3 concentration and the incidence of clinical events during long-term follow-up. As suggested by Nakagawa et al., a time-dependent ROC analysis may be more appropriate than the analysis employed here. A time-dependent ROC analysis using the timeROC package (R environment for statistical computing version 4.1.3 on macOS) yielded the following results for Gal-3 values and appropriate ICD therapy (Figure) (6): the most efficient cut-off Gal-3 values at 381, 510, and 551 days, which correspond to the 1st quartile, median, and 3rd quartile follow-up periods, were 13.23, 13.13, and 13.13 ng/mL (area under the curve: 0.82, 0.82, and 0.79, respectively). These cut-off values corresponded closely with those obtained by the non-time-dependent ROC analysis. For heart failure hospitalization, the cut-off values obtained by the time-dependent ROC analysis (12.41, 12.41, and 13.13 ng/mL, respectively) were also consistent with those obtained by the non-time-dependent ROC analysis.

Figure.

Time-dependent receiver operating characteristic (ROC) curves for galectin-3 and appropriate ICD therapy. The ROC curves at 381, 510 and 551 days of follow-up (1st, 2nd and 3rd quartile of follow-up) are shown.

We thank Nakagawa et al. for their constructive suggestions.

The authors state that they have no Conflict of Interest (COI).