Overall statement Dear authors of Development of Heart Failure From Transient Atrial Fibrillation Attacks in Responders to Cardiac Resynchronization Therapy. This is an interesting study looking at intermittent AF patients with CRTs and investigating the AF related HF admissions and how that affects the biventricular pacing percentage and how intermittent AF and AF attacks are associated with worse clinical outcomes.

Strengths and impact The strength of the study is its novelty, length of follow-up, and its clinical assessment. The impact is very relevant because it concludes that the CRT benefits are attenuated in patients with intermittent AF. However there are some concerns the authors should try to address:

Major points.

1. Page 3. Statistical analysis The authors states they counted the cumulative event rates by the Kaplan Meier method for death, HF hospitalization, and ICD shocks. That is sufficient for HF/Death outcome. But not for ICD shock outcomes because of the competing risk of death. I suggest analyzing the ICD shocks as cumulative incidence with Nelson Aalen plots, which takes into account the competing risk of death. See more here https://www.ncbi.nlm.nih.gov/pubmed/17255278

2. Page 3. Clarification of the Cox model: Was the Cox proportional hazard model assumptions met? The authors should state whether the Cox model assumptions were met. The test for linearity and proportionality should be included in the method section.

3. Page 3. The authors adjusted the Cox model for age, sex, ECG morphology, HF severity, creatine levels, and beta-blocker use. Can the authors clarify the justification for these adjustments - Did the authors consider also to adjust for ischemic heart disease, diabetes, chronic pulmonary obstructive disease, and digoxin use (other possible confounders).

4. Page 3, baseline table. Could it be made clearer whether the baseline table is at CRT implantation date or 3-months after CRT implantation?

5. Page 5, results: The analysis of AF + BVIP% <90% AF compared to BVIP% >90% during AF attack is interesting. Could the authors provide any information on the key baseline differences between these two groups?

6. Page 6, discussion. There is mentioning of a subgroup analysis of ablation in the CRT patients in the study. The authors should present this subgroup analysis in the result section.

7. Page 6, discussion. The authors state their findings supports that a small burden of AF deteriorates CRT outcomes. In this study, was there any risk difference between intermittent AF and permanent AF on the outcomes? For my perspective, the data shows that there was no difference in permanent /intermittent AF on the outcomes, but outcomes depended on the BVIP% in the AF attacks. Could the authors elaborate on this point?

8. Page 7, limitations. It is written that the patients with intermittent AF had more severe HF which could influence the outcomes. This could be addressed by an interaction analysis to see if severe HF was an effect modifier of the outcome.

1) etc.. They can only refer to associations, 2) I worry about the small sample size; many of the interesting questions you raised in your review cannot be addressed due to the small number of pts and/or events. 3) I worry that their model was overfitted due to the small number of events, 4) I agree that competing risk of death when looking at other outcomes should be considered but this can also be done using the Fine-Grey methodology, and 5) another weakness is that events were not adjudicated centrally.

Minor points

1. Abstract: The method section of the abstract should be clearer. how the statistical method of comparison was performed what was the primary and secondary outcomes. The BVIP% is stated as significant difference with a p-value, I would prefer to see the percentage difference instead of a p-value. It gives better information than a p-value.

2. Introduction: The authors write there are limited studies on the subject and could clarify what these studies have shown so far.

3. Introduction: The specialized term for an "AF attack" is unclear. This term should be defined.

4. Table 1. Baseline characteristics. If the results are shown after 3 months of CRT, could the authors provide baseline BVIP% during the first 3 months for all the groups?

5. Page 5, discussion: The authors state that "Our findings suggest that AVJA even for the patients with intermittent AF may be reasonable to improve their outcomes." In this study the AVJA patients were excluded. Therefore it is unknown if intermittent AF patients will benefit from AVJA. The authors should make it clearer that this is a hypothesis for future research (and is not supported by the findings).

6. Limitations should include the non-randomized nature of the study, non-causality, residual confounding, and selection bias.