Author s response to reviews Title: Effects of short-term heart rate variability biofeedback on long-term abstinence in alcohol dependent patients - a one-year follow up Authors: Ana Isabel Penzlin (ana.penzlin@ppcr.org) Kristian Barlinn (Kristian.Barlinn@uniklinikum-dresden.de) Ben Illigens (illigens@bidmc.harvard.edu) Kerstin Weidner (Kerstin.Weidner@uniklinikum-dresden.de) Martin Siepmann (Martin.Siepmann@uniklinikum-dresden.de) Timo Siepmann (timo.siepmann@uniklinikum-dresden.de) Version: 1 Date: 22 Jul 2017 Author s response to reviews: We would like thank the editor and reviewers for the opportunity to resubmit our revised paper. The comments made led to the inclusion of additional information and the discussion of additional aspects that, in our belief, have substantially increased the scientific quality of the present manuscript. We revised the manuscript accordingly and responded to all review comments in point-to-point style. Reviewer 1- comments Comment 1: I would like to thank the authors for submitting this well written paper addressing an important topic. Kudos too for doing a follow-up study with a notoriously difficult population to follow-up with. Though differences between groups were not significant in terms of relapse, the study really wasn't powered to find significant effects. Consider reporting effect sizes like Cohen's d, and reframing the manuscript in a more exploratory light. So rather than seeking to show significance, the goal could have been to look for meaningful effect sizes to inform future research. Given HRV-BFB for AUD is so new, and the literature so small, this would be a logical goal for the paper. With this approach you could speak more confidently to the observed
between group differences. Though not significant, differences would be meaningful if you have a medium effect size to speak to. Reply to comment 1: We thank the reviewer for these helpful suggestions. We agree that the exploratory nature of our study should be underscored. Therefore, we have incorporated this thought into the abstract. (page 2, lines 38, 40, 58-9) Furthermore, we modified the introduction to read In this follow up survey study we aimed to determine the relapse rate one year after HRV-biofeedback training and integrative inpatient rehabilitative treatment in order to explore potential treatment effects and acquire first long-term data that could form a basis for confirmatory research in large study populations. (page 3, lines 81-3) Lastly, we modified a paragraph within the discussion to read Although our study was not powered to show significant group differences one year posttreatment, our data might contribute to generating the hypothesis that HRV-biofeedback has beneficial effects on long-term abstinence. (page 8, lines 216-8) In addition, as suggested, we calculated Cohen s effect sizes for craving and anxiety, as these parameters showed a significant difference between intervention and control group in the original RCT. For this analysis, we considered the scores for craving and anxiety for patients who remained abstinent and repeated the analysis for those who suffered from relapse separately. Cohen s effect size value (d=0.44) suggested a small effect of HRV-biofeedback on craving in patients who remained abstinent one year after discharge from rehabilitation. A small effect (d=0.45) of the intervention on craving was also found in patients who suffered relapse after discharge. Similarly, HRV-biofeedback displayed rather small effect sizes with respect to anxiety in patients who remained abstinent at follow up (d=0.06) as well as in those who suffered from relapse (d=0.32). Since effect sizes were presumably biased due to the limited number of survey responders and do not extend interpretability of the non-significant trends reported in our paper, we chose not to add Cohen s effect size values to the present manuscript. However, we will incorporate this valid suggestion into the design of longitudinal follow up research which is currently being designed and will be powered to show effects of HRV-biofeedback on outcomes of long-term abstinence. Comment 2: The authors are doing an intent-to-treat analysis by considering non-responders to have relapsed. I recommend using this term in the manuscript as it succinctly communicates the analytical approach employed.
Reply to comment 2: We agree with the reviewer that this additional clarification is helpful to make the presented data more comprehensible. We highlighted the applied intend-to-treat method throughout the text following this recommendation. (page 5, line 126; page 7, lines 184-5, page 16, line 416) Comment 3: " non-responders were considered relapse" should read " non-responders were considered to have relapsed". Reply to comment 3: We thank the reviewer for this advice. We have changed the sentence within the manuscript to read In this more conservative analysis, non-responders were considered to have relapsed. (page 5, lines 127-8) Comment 4: I think the non-significant nature of the results should be noted in the abstract when describing between group differences in abstinence rates. As is, it is a little ambiguous. I would definitely report effect sizes in abstract though if they turn out to be meaningful. Reply to comment 4 We have added an explanation of the non-significant nature of the observed trends to the abstract. (page 2, line 53) The following sentence was modified: This non-significant trend was also observed when non-responders were additionally included in the analyses and were assumed to have had a relapse (46,7% biofeedback vs. 33.3% controls, p=ns). We have not added effect sizes to the abstract since these have not turned to add substantial insights to the comparison of percentages presented in the current form of the abstract.
Comment 5: Grammar is off in sentence, "Comparisons of these characteristics between responders who had " Reply to comment 5: We thank the reviewer for pointing out this grammatical mistake. We rectified the sentence to read There were no differences in age, gender, tobacco use, neuropathy and number of cases of liver disease between responders who have undergone biofeedback in addition to rehabilitative treatment and those who have undergone rehabilitation only. (page 7, line 175-6). Comment 6: I'm a little confused by Figure 1. Shouldn't the combined bars for survey responders add up to 100%? Ditto for non-responders? Reply to comment 6: We thank the reviewer for the opportunity to clarify this point. In fact, Figure 1 shows the rate of survey responders from each intervention group who remained abstinent one year after discharge from the rehabilitative therapy. Therefore, each study arm represents a separate population in this graph. The red bar indicates survey responders who have undergone HRF biofeedback. The blue bar indicates survey responders who have not undergone the intervention. The height of each bar relates to the percentage of subjects out of each population (study arm) who were abstinent one year post intervention. In order to improve the understanding of the bar graph, we included a brief explanation to the figure legend to read The bar graph shows the rate of survey responders in each study arm of the original RCT who remained abstinent one year after discharge from rehabilitation (left). (page 16, lines 412-3) Comment 7: In the discussion, the authors discussed weak effects of HRV-BFB on HRV following the intervention, and that this may explain the follow-up abstinence results. Though this is of course possible, I suspect this is not the case. HRV-BFB often produces improvements on psychological measures without necessarily effecting major increases in HRV post-intervention.
Reply to comment 7: We thank the reviewer for pointing out this valid point. We incorporated this useful explanation into the discussion and added a relevant citation. The new paragraph added to the discussion reads: In fact, treatment with HRV-biofeedback can improve psychometric measures such as subjective fatigue independent from changes in autonomic cardiac function.[13] (page 9, lines 231-3) Comment 8: Was amount of past-year drinking assessed at follow-up? If so would be good to include this in the analyses? Similarly, were participants asked if the continued to practice HRV-BFB in the year following the intervention? If so this could be a useful covariate in the models. Reply to comment 8: Although we agree that the questions raised by the review would be of interest to answer, we have not assessed the amount of past-year drinking. Therefore we weren t able to add these data to the manuscript. However, we will incorporate this valid idea into the design of follow research that is currently under way. In our study, patients weren t asked to continue paced breathing after the intervention and therefore we haven t assessed the rate of those still practicing post discharge. This is actually a crucial point which might explain why in our study observed trends of increase in abstinence have not reached statistical significance. We added this important limitation to the discussion: Furthermore, patients weren t asked to continue paced breathing after the intervention and therefore we haven t assessed the rate of those still practicing post discharge. This might, in part, explain why in our study observed trends of increase in abstinence have not reached statistical significance. (page 10, lines 268-75) Comment 9: "In both studies the duration and frequency " I recommend rewording this sentence to improve clarity. It sounds like the Eddie et al. study had 6 sessions when it had 3. Reply to comment 9: We rephrased this sentence in order to improve clarity. The modified sentence reads: In both studies, the duration and frequency of HRV-biofeedback training was limited, corresponding to 6
treatment sessions over a period of 2 weeks in our RCT and 3 sessions during 3 weeks in the work of Eddie et al.. (page 9, lines 232-3) Comment 10: "A possible explanation for changes." Grammar correction needed in this sentence. Reply to comment 10: We thank the reviewer for this comment. We rephrased this sentence to read The observed effect of HRV-biofeedback on autonomic cardiac function and craving might be explained by the physiological mechanism of action of this technique. (page 9, lines 229-31) Comment 11: "Although the exact neurophysiological interaction between CAN and AER are " 'Are' should be 'is' in this sentence. Reply to comment 11: We thank the reviewer for pointing out this mistake which we have corrected in the revised manuscript. The revised sentence reads: Although the exact neurophysiological interaction between CAN and AER is poorly elucidated, their common capacity of altering HRV might explain why improvement of HRV could also lead to stabilization of craving and anxiety in alcohol dependent patients. (page 10, lines 254-7) Comment 12: Are the post-hrv-bfb assessments reported from the assessment following the final HRV-BFB training session? I recommend clarifying this. Forgive me if this is stated and I missed it! Reply to comment 12: We thank the reviewer for pointing out this imprecise description. Indeed, all assessments were performed post-final training. In order to clarify the time points of post-biofeedback assessment,
we rephrased the corresponding sentence in the methods session to read We included in our analyses psychometric scores at baseline and those obtained immediately following the final HRV-biofeedback training session (post-biofeedback or post-observation period for patients in the control group). (page 5, lines 134-7) Reviewer 2- comments Jong-Min Woo (Reviewer 2): This is a very well-written manuscript dealing with important clinical issues. I have read through the paper with joy. This study has two major weaknesses in terms of short period/low intensity of biofeedback intervention and insufficient number of study participants. However, these limitations do not hinder the merit of the manuscript. Only suggestion on my end is editing the tables to be more visible. Comment to reviewer 2: We thank the reviewer for the positive evaluation of our work. We agree that the limited number of study participants as well as the comparatively short of the intervention time pose limitations to our data. We have changed parts of the discussion to acknowledge that these limitations might have contributed to the fact that observed trends toward improved long-term abstinence after biofeedback have not reached statistical significance. ( Although our study was not powered to show significant group differences one year post-treatment, our data might contribute to generating the hypothesis that HRV-biofeedback has beneficial effects on long-term abstinence. (page 8, lines 212-4) It might also important to point out that in our study patients were not instructed to continue paced breathing after the completion of the intervention period, which might have further weaken the observed trends to increased abstinence at follow up. This important limitation of our study also addresses comment eight of reviewer one. We have added this limitation to the limitations section of the revised discussion. Additionally, we have weakened the wording for interpreting our findings within the abstract (page 2, lines 38, 40, 58-9) and discussion to read Although our study was not powered to show significant group differences one year posttreatment, our data might contribute to generating the hypothesis that HRV-biofeedback has beneficial effects on long-term abstinence. (page 8, lines 216-8) We agree that a more visible design of the tables would be helpful. We would like to point out that we have generated the manuscript including tables and figures to comply with the journal s instructions for author. We increased font sizes and line thickness of our tables in the revised
manuscript and would like to recommend further optimizing visibility of the tables during editorial processing according the journal s house style.