Characterization of Cardiopulmonary Interactions and Exploring Their Prognostic Value in Acute Bronchiolitis: A Prospective Cardiopulmonary Ultrasound Study

Round 1

Reviewer 1 Report

Thank you for the opportunity to review this manuscript.

The authors describe a single-centre, observational study of children <1 year of age admitted with acute bronchiolitis. The authors aim to describe the relationship/correlation between clinical, laboratory and cardio-pulmonary ultrasonographic metrics/features. Further, they examine the performance of a number of these laboratory, clinical and CPU features for discriminating between clinical outcomes.

 

The manuscript is well written, interesting and the approach is well described. I have some suggestions for some additional explanations that would be welcome, which I have included below.

Major points:

As the authors point out, the cardiac and pulmonary systems are tightly interconnected and co-dependent. It is therefore perhaps not surprising that a large number of the analyses demonstrate statistically significant results in univariate analyses. I suspect there are many co-linear relationships present across these metrics. Nonetheless the authors have described these univariate relationships well. The number of analyses does raise the issue of multiple testing effects and type 1 error, however given the exploratory nature of the paper, coupled with the strong statistical significance of the results (P often <0.0001) I suspect this is not a critical issue.

 

My biggest comment/critique is regarding the practical applicability of CPU/LUS and whether in real world it’s use would add much compared to more common/easily calculable metrics/tests such as BNP, SaO2, BRODSJOD.

While the authors point out that clinical scores are prone to high inter-observer variability, the score assessed in this study (BROSJOD) demonstrated correlation with outcomes that was strongly competitive with the best ultrasonographic measures. A BROSJOD of >10 also had by far the highest OR for respiratory support and a high AUC-ROC, with the highest sensitivity. BNP and pH also performed strongly as stand-alone tests.

For example, for LOS hospitalisation the five highest correlation coefficients were BNP (0.60), LUS (0.58), RVTX (0.58), LVEI (0.58) and BROSJOD (0.55), all with p<0.001. The significance in distinction between a correlation coefficient of 0.55 to 0.6 is modest, while the expertise required to get these measures varies significantly.

Notwithstanding the limitation of inter-observer scoring, it would likely be easier to train emergency/pediatric doctors to accurately quantify BROSJOD, as opposed to the steep learning curve and technical proficiency required for accurate lung ultrasound, let alone the even more technically challenging requirement for accurate cardiac echo.

 

Based on the results presented I would suspect a composite model which includes components of BROSJOD with BNP, and possibly pH would be strongly predictive and would require little technical up-skilling for front line clinicians. In this regard the conclusions of the authors (that CUS could add value in discriminating high-risk patients), though strictly speaking correct, somewhat gloss over the fact that some of the results suggest that less technically complex tools may perform just as well if not better. Perhaps the authors could comment.

 

 

Aside from this I have some more general comments and suggestions, below.

Materials and Methods

 

1. “The diagnosis and management were made by the pediatrician on charge following current international recommendations” –

Do the authors mean the diagnosis for each case by the admitting paediatrician?  Were members of the study team responsible for the diagnosis in any of these cases or fully independent from establishing the diagnosis?

 

2. “The local Ethics Committee approved the protocol and…”

Please include the ethics committee approval number/ID for this study either here or in the Institutional Board Review Statement.

 

3. Page 3: 2.2.1 Pulmonary assessment:

Who assessed/calculated/recorded the BROSJOD score for each participant? Was this a study team member separate from clinical care? Were multiple assessors included for concordance? As the authors state in the introduction “Clinical scores have proven to be inaccurate in predicting prognosis in acute bronchiolitis, with weak interobserver reliability and limited construct validity”. How did the study team plan for reducing interobserver reliability in the study data collection on this variable (BROSJOD)?

 

4. “We did not evaluate posterior/paravertebral projections due to the difficulty in performing them in these patients, which makes them not very feasible in clinical practice, and furthermore, we believe that these would only provide limited information, since the posterior fields are the ones, most frequently affected in the initial phases of acute bronchiolitis, probably due to the supine position of these patients.”

Please remove the comma (,) after “are the ones, …”

 

 

5. Page 3: Please clarify the LUS:

In the methods section the text reads: “Each area was longitudinally explored in a cephalocaudal direction evaluating each intercostal space. In each area, the presence of B-lines and subpleural consolidations was analyzed using a lung ultrasound score (LUS) (Figure 2 & 3).”

However, in figure 2 the scores for B-lines are described by “intercostal space”, and not “area”.

Were <3 vs > 3 B lines needed in every intercostal space in a given area OR was this based on the average of the area OR the highest number of B lines seen within any of the intercostal spaces withing an area? To rephrase my question: how many lines in how many intercostal spaces were used to distinguish a B line score of 0 vs 1? If a patient had (for example) 3 B lines in one intercostal space within area 2, but only 1-2 B lines in the other intercostal spaces in area 2, how would that area score? Please clarify and update figure 2 accordingly.

 

6. Page 4: Figure 3 legend

I suggest changing “Figure 3: This is a lung ultrasound image showing:…”  to “Figure 3: Linear transducer lung ultrasound images showing: ….

 

Also figure 3 legend: please correct “ twoadjacent” to “two adjacent”

 

7. Page 4: Standard techniques to obtain M-mode, two-dimensional and Doppler (colour, pulsed, continuous and tissue Doppler) echocardiograms were performed by the same experienced paediatric cardiologist as recommended in the guidelines for the paediatric echocardiography

Please remove “the” from “for the paediatric echocardiography”

 

8. Page 4: please remove the paragraphed indent between “pressures evaluation” and “included”

 

9. Page 4: “included: Left ventricle (LV) ejection fraction (LVEF), tricuspid annular plane systolic excursion (TAPSE), LV and RV Tei index (LVTX and RVTX) (Figure 3),…”

 

The inclusion of “Figure 3” in this sentence appears incorrect.

 

10. Table 1: please include an explanation for “BS”, “LV” and “RV” in the legend

 

11. Page 7: Statistics

 

Please comment on the decision not to adjust for multiple testing. If the reason was the exploratory nature of the study, then I would suggest inserting a single line comment to that effect (such as “Due to the exploratory nature of our study, significance threshold adjustments for multiple testing were not performed). Please comment on the correlation analysis: were none of the correlations appropriate for Pearson’s correlation (i.e. parametric?)

 

12. Page 8: Table 2:

I suggest changing the title to “Demographic, intervention and outcome data of the….”

I would also suggest changing “nebulizations” to “Nebulised therapies” or similar.

Also, CPAP is not a NIV modality so this would be preferably presented either as CPAP/BiPAP alone (without labelling the two as NIV), or presenting CPAP and BiPAP use separately.

 

13. Page 9: Table 3: Cardiac Variables

Here, for TAPSE < 2 Z score for “BSA” is used as opposed to “BS” in Table 1. I suggest using one for continuity.

 

14. Page 10: table 4 legend:

 please insert a closing bracket “)”at the end of the legend as it currently ends with “(p<0.05”

please insert a reference for the significance of “*” and “**”

Author Response

1. Thank you for the opportunity to review this manuscript. The authors describe a single-center, observational study of children <1 year of age admitted with acute bronchiolitis. The authors aim to describe the relationship/correlation between clinical, laboratory, and cardio-pulmonary ultrasonographic metrics/features. Further, they examine the performance of a number of these laboratory, clinical and CPU features for discriminating between clinical outcomes. The manuscript is well written, interesting and the approach is well described. I have some suggestions for some additional explanations that would be welcome, which I have included below.

• Response: Thank you very much for your time dedicated to reviewing our manuscript and for your helpful suggestions to improve it. We will try to address all of them in the following lines. We noticed your detailed revision and appreciate it. You will notice that some sentences of the methodology were rephrased following the recommendations of the editorial team. We think that the meaning remains similar.

2. As the authors point out, the cardiac and pulmonary systems are tightly interconnected and co-dependent. It is therefore perhaps not surprising that a large number of the analyses demonstrate statistically significant results in univariate analyses. I suspect there are many co-linear relationships present across these metrics. Nonetheless the authors have described these univariate relationships well. The number of analyses does raise the issue of multiple testing effects and type 1 error, however given the exploratory nature of the paper, coupled with the strong statistical significance of the results (P often <0.0001) I suspect this is not a critical issue.

- Response: We fully agree with the reviewer with this statistical issue. Consistent with your observation, our main objective was to explore CPI in acute bronchiolitis. As this relationship has not been systematically described in this setting, we designed an exploratory analysis, based on previously published observations of our group (see reference 29). After this comment, we performed a post-hoc Bonferroni ́s type adjustment for multiple correlation analysis. As you said, the P-values maintained their significance for all the comparisons after this adjustment. The maximum comparison made was n=39 (table 6). If you conduct 39 comparisons, only reject the null hypothesis of each comparison if it has a p-value less than 0.00128. Only the correlation of pH, pCO2 and TRJG in table 6 could be considered altered by this adjustment, but because the exploratory nature of the study we considered not relevant for our objective. If you consider that this correction must be present on the text or a different statistical analysis is need, please let me know to perform it and make corrections.

3. My biggest comment/critique is regarding the practical applicability of CPU/LUS and whether in real world it’s use would add much compared to more common/ easily calculable metrics/tests such as BNP, SaO2, BRODSJOD. While the authors point out that clinical scores are prone to high inter-observer variability, the score assessed in this study (BROSJOD) demonstrated correlation with outcomes that was strongly competitive with the best ultrasonographic measures. A BROSJOD of >10 also had by far the highest OR for respiratory support and a high AUC-ROC, with the highest sensitivity. BNP and pH also performed strongly as stand-alone tests. For example, for LOS hospitalisation the five highest correlation coefficients were BNP (0.60), LUS (0.58), RVTX (0.58), LVEI (0.58) and BROSJOD (0.55), all with p<0.001. The significance in distinction between a correlation coefficient of 0.55 to 0.6 is modest, while the expertise required to get these measures varies significantly. Notwithstanding the limitation of inter-observer scoring, it would likely be easier to train emergency/pediatric doctors to accurately quantify BROSJOD, as opposed to the steep learning curve and technical proficiency required for accurate lung ultrasound, let alone the even more technically challenging requirement for accurate cardiac echo. Based on the results presented I would suspect a composite model which includes components of BROSJOD with BNP, and possibly pH would be strongly predictive and would require little technical up-skilling for front line clinicians. In this regard the conclusions of the authors (that CUS could add value in discriminating high-risk patients), though strictly speaking correct, somewhat gloss over the fact that some of the results suggest that less technically complex tools may perform just as well if not better. Perhaps the authors could comment.

- Response: We fully understand and even may agree with your comment. 

Of course, our results show that the BROSJOD score is adequate, but there is controversy about this subject in the literature, and clinical scores vary across different institutions. Therefore we feel difficult to implement a universal clinical score with adequate predictive accuracy. The major issue with clinical scores is that they are excellent tools for the evaluation of clinical state at the moment of the evaluation, not to predict evolution. And as mentioned, they present a variable degree of interobserver and intraobserver variability. Furthermore, the BROSJOD score is used as a criterion for PICU admission and initiation of RS in our institution, and it could explain our results regarding the predictive accuracy for severity. Therefore, we feel that there is a need for adjuvant and less subjective tools to improve the assessment of severity in this setting.

As you suggested, we previously tested a model with NT-proBNP, the age, and the BROSJOD score with excellent results (AUC 0.945) (ref. 29), so why explore CPU accuracy and why not create the new composite model in this work?? Well, many colleagues and reviewers told us on many occasions that blood extractions in infants with AB are invasive, painful... and they usually expressed concerns about if a painful blood extraction adds enough to improve accuracy to modify the current recommendations of “sitting to watch” what is going to occur with clinical scores. Therefore, this led us to identify non-invasive tools for this purpose. LU has been previously demonstrated to be an accurate tool, but you know that there is not a perfect predictive tool. We think that the combination of several tools is the better approach. Then, we used our previous echocardiographic observations in this setting to increase the pool of tools. Interestingly, no data about predictive accuracy for the severity of any echocardiographic parameter was provided in previous work, and here we provided more reliable parameters that could be easily learned by non-cardiologist in a POCUS, such as pulmonary acceleration time/ejection time ratio or left ventricular eccentricity index. Interobserver variability with CPU exists and it was not tested in this study. However previous studies demonstrated that trained emergency care pediatricians have a good inter-observer variability for at least LU and some echocardiographic parameters (overall ATET that we highlighted in the discussion).

Nevertheless, we fully agree that the time to train clinicians in CPU is a major limitation for this approach. So, what would be the best strategy? Blood extraction adding biochemical markers (pH, NT-proBNP)?? Training pediatricians on CPU or invasive-painful blood extractions?? Even only sitting to watch clinical scores?? We think that there is not a unique, universal, and adequate answer. Maybe different combinations of different tools according to availability at different centers and cost-efficiency issues are the correct answer. Therefore, the publications of works such as this one could provide evidence that there are different tools available for this purpose, and each clinician/institution will choose those that best suit their circumstances. For example, in our institution, we have incorporated NT-proBNP to the blood analysis when a blood extraction is considered by the attending pediatrician. CPU is performed in the work-up of all hospitalized infants requiring supplemental O2. As you could understand is challenging and maybe not cost-efficient (not investigated) to perform in all hospitalized infants.

We discuss it now at the end of the discussion (we used your valuable words for this purpose because you explained it very clear) clarifying that this study does not demonstrate that overpass clinical assessment (not designed for this objective), only suggested a potential role of CPU always in combination with clinical assessment. We added to the limitations the lack of interobserver reliability assessment for CPU and BROSJOD score in this study.

4. “The diagnosis and management were made by the pediatrician on charge following current international recommendations” Do the authors mean the diagnosis for each case by the admitting pediatrician? Were members of the study team responsible for the diagnosis in any of these cases or fully independent from establishing the diagnosis?

• Response: We mean by the admitting pediatrician (added to the methods). All investigators were fully independent of establishing the diagnosis. Admitting pediatrician identified cases of AB and offers the inclusion in the study, and if the parents agreed they notified the investigators that enrolled the patient in the study. After enrollment, the attending pediatrician calculated the BROSJOD score and solicited blood analysis. As you could notice, some of the investigators work at PICU and of course, they take care of some patients included, but always after the inclusion of the patient in the study. Importantly, they never calculated the BROSJOD score at admission, that was the score used for this study. We acknowledge that although we were blinded to the clinical score and blood analysis results of the patient at the time of CPU assessment, the investigators (all pediatricians) could have an approximated idea of the clinical state at the moment of echocardiographic evaluation. This fact could influence the CPU results. To minimize this influence the CPU studies were de-identified and analyzed offline at least 2 weeks after the recording as stated in Methods. Nevertheless, a complete blind in these circumstances is challenging as you could understand.

5. “The local Ethics Committee approved the protocol and...” Please include the ethics committee approval number/ID for this study either here or in the Institutional Board Review Statement.

• Response: We have provided the document of our ethics committee approval to the Editorial Board of the journal. Our local committee does not provide number/ ID, but a document stating that the study fulfills ethical and methodological requirements, and the date (25th October, 2018) when the study was approved. If you want I can send you this document or you can request it to the Editorial board.

6. Page 3: 2.2.1 Pulmonary assessment: Who assessed/calculated/recorded the BROSJOD score for each participant? Was this a study team member separate from clinical care? Were multiple assessors included for concordance? As the authors state in the introduction “Clinical scores have proven to be inaccurate in predicting prognosis in acute bronchiolitis, with weak interobserver reliability and limited construct validity”. How did the study team plan for reducing interobserver reliability in the study data collection on this variable (BROSJOD)?

• Response: As mentioned above, the BROSJOD score was assessed by the attending pediatrician on charge of the patient (only 1 person, not multiple assessors). This pediatrician could be located at the emergency department or at the pediatrics ward of our hospital, but always was not an investigator. This varied in function of the availability of the investigators to perform CPU and nursery to perform blood extractions simultaneously (very close).

• We consider reducing the interobserver reliability a very interesting issue to solve in the future, but it was not our objective in this study. We try with our line of research to identify potential tools to reduce clinical scores limitations when used in combination. Therefore, in this study, we did not plan anything to solve this limitation. This comment was added to the limitations.

7. “We did not evaluate posterior/paravertebral projections due to the difficulty in performing them in these patients, which makes them not very feasible in clinical practice, and furthermore, we believe that these would only provide limited information, since the posterior fields are the ones, most frequently affected in the initial phases of acute bronchiolitis, probably due to the supine position of these patients.” Please remove the comma (,) after “are the ones, ...”

• Response Done.

8. Page 3: Please clarify the LUS: In the methods section the text reads: “Each area was longitudinally explored in a cephalocaudal direction evaluating each intercostal space. In each area, the presence of B-lines and subpleural consolidations was analyzed using a lung ultrasound score (LUS) (Figure 2 & 3).” However, in figure 2 the scores for B-lines are described by “intercostal space”, and not “area”. Were <3 vs > 3 B lines needed in every intercostal space in a given area OR was this based on the average of the area OR the highest number of B lines seen within any of the intercostal spaces withing an area? To rephrase my question: how many lines in how many intercostal spaces were used to distinguish a B line score of 0 vs 1? If a patient had (for example) 3 B lines in one intercostal space within area 2, but only 1-2 B lines in the other intercostal spaces in area 2, how would that area score? Please clarify and update figure 2 accordingly.

- Response: Each area is classified based on the highest score obtained from the intercostal spaces. For example, if in area 1, in an intercostal space 2 points were obtained and in another intercostal space of the same area 3 points were obtained, area 1 was scored with 3 points. We have clarified the explanation of figure 2 in the text: "Each area was longitudinally explored in a cephalocaudal direction evaluating each intercostal space. In each area, the presence of B-lines and subpleural consolidations was analyzed using a lung ultrasound score (LUS). Each area was scored based on the highest scores obtained from the exploration of each intercostal space of the same area. For example, if in area 1, in an intercostal space 2 points were obtained and in another intercostal space of the same area 3 points were obtained, area 1 was scored with 3 points (Figure 2 & 3)."

9. Page 4: Figure 3 legend. I suggest changing “Figure 3: This is a lung ultrasound image showing:...” to “Figure 3: Linear transducer lung ultrasound images showing: ....

• Response: Done.

10. Also figure 3 legend: please correct “ twoadjacent” to “two adjacent”

• Response: Done.

11. Page 4: Standard techniques to obtain M-mode, two-dimensional and Doppler (colour, pulsed, continuous and tissue Doppler) echocardiograms were performed by the same experienced paediatric cardiologist as recommended in the guidelines for the paediatric echocardiography. Please remove “the” from “for the paediatric echocardiography”

• Response: Done.

11. Page 4: please remove the paragraphed indent between “pressures evaluation” and “included: Left ventricle (LV) ejection fraction (LVEF), tricuspid annular plane systolic excursion (TAPSE), LV and RV Tei index (LVTX and RVTX) (Figure 3),...”

• Response: Done.

12. The inclusion of “Figure 3” in this sentence appears incorrect.

• Response: You are right. Deleted. It is Figure 5.

13. Table 1: please include an explanation for “BS”, “LV” and “RV” in the legend

• Response: Done. BS is BSA (body surface area), LV (left ventricle) and RV (right ventricle).

14. Page 7: Statistics. Please comment on the decision not to adjust for multiple testing. If the reason was the exploratory nature of the study, then I would suggest inserting a single line comment to that effect (such as “Due to the exploratory nature of our study, significance threshold adjustments for multiple testing were not performed).

• Response: As you mentioned previously, we only aimed to explore in this study. Thank you very much for your suggested comment. We added this line tho the methods.

15. Please comment on the correlation analysis: were none of the correlations appropriate for Pearson’s correlation (i.e. parametric?)

• Response: You are right, many variables presented a normal distribution and parametric test should be applied. However, all variables selected as end-points presented skewed distribution and therefore, when comparing a normal with a non-normal variable in the correlation analysis (tables 4 and 6) we decided to use Spearman test. After your comment we tested also with Pearson test for comparisons where at least one variable presented normal distribution and the significance remained similar. However if you consider that our analysis is not correct please let us to know to make the changes needed. Now the correlation coefficient (CC) is stated for each variable in tables as Spearman or Pearson.

16. Page 8: Table 2: I suggest changing the title to “Demographic, intervention and outcome data of the....”

• Response: Done.

17. I would also suggest changing “nebulizations” to “Nebulised therapies” or similar.

• Response: Done.

18. Also, CPAP is not a NIV modality so this would be preferably presented either as CPAP/BiPAP alone (without labelling the two as NIV), or presenting CPAP and BiPAP use separately.

• Response: Done.

19. Page 9: Table 3: Cardiac Variables. Here, for TAPSE < 2 Z score for “BSA” is used as opposed to “BS” in Table 1. I suggest using one for continuity.

• Response: Done.

20. Page 10: table 4 legend: please insert a closing bracket “)”at the end of the legend as it currently ends with “(p<0.05”) please insert a reference for the significance of “” and “”

• Response: Done. Spearman and Pearson.

Reviewer 2 Report

I do appreciate the authors for this well-written article, however, I still have some comments and questions.

    Please introduce abbreviations before using them such as LOS, AB, and BA.
    You have stated, "Cardiopulmonary ultrasonography could be a good strategy to easily identify high-risk population for a complicated acute bronchiolitis hospitalization". However, I believe CPU is rather a tool than a "strategy".
    In the exclusion criteria, it is written "poor-quality ultrasound images" were excluded. Doesn't it lead to a selection bias by a probable inclusion of the only typical cases?
    It is written this study was conducted in an "a tertiary University hospital in Spain", please specify where exactly.
    Finally, this study of yours is very similar to your former study entitled "Left Ventricular Dysfunction and Plasmatic NT-proBNP Are Associated with Adverse Evolution in Respiratory Syncytial Virus Bronchiolitis". How do your findings in the current study make it novel compared to your previous study?

Author Response

1. I do appreciate the authors for this well-written article, however, I still have some comments and questions.

• Response: Thank you very much for your time reviewing our manuscript and your helpful comments. You will notice that some sentences

2. Please introduce abbreviations before using them such as LOS, AB, and BA.

• Response: Done

3. You have stated, "Cardiopulmonary ultrasonography could be a good strategy to easily identify high-risk population for a complicated acute bronchiolitis hospitalization". However, I believe CPU is rather a tool than a "strategy".

• Response: Ok. We fully agree with this comment and we changed this expression along the text.

4. In the exclusion criteria, it is written "poor-quality ultrasound images" were excluded. Doesn't it lead to a selection bias by a probable inclusion of the only typical cases?

• Response: CPU could be well done in all the patients who accepted to be included. We acknowledge are all experts in echo and this could be possible because our expertise. One of the limitation of the application of our results to the clinical practice is the need of training in order to obtain adequate images in a relatively short time in infants usually irritable and with poor acoustic window. We added this comment to the discussion.

5. It is written this study was conducted in an "a tertiary University hospital in Spain", please specify where exactly.

• Response: The study was carried our at the city of Cadiz (Puerta Del Mar University Hospital)

6. Finally, this study of yours is very similar to your former study entitled "Left Ventricular Dysfunction and Plasmatic NT-proBNP Are Associated with Adverse Evolution in Respiratory Syncytial Virus Bronchiolitis". How do your findings in the current study make it novel compared to your previous study?

• Response: From the last 5 years our main line of research has been the investigation of cardiac strain (and maybe subclinical dysfunction) as a risk factor for complicated courses of several non-primary pediatric cardiac diseases. The current study served us as an internal validation of our previous investigations (with a similar “cardiac design”) where we observed again similar results previously obtained (Cardiac strain measured by NT-proBNP or echocardiography is associated with severe bronchiolitis in healthy infants). These results led us to include this year these tools in our local protocol for the management of bronchiolitis. In our institution, we have incorporated NT-proBNP to the blood analysis when a blood extraction is considered by the attending pediatrician, and CPU is performed in the work-up of all hospitalized infants requiring supplemental O2. As you could understand is challenging and maybe not cost-efficient (not investigated) to perform in all hospitalized infants.

• We agree that some results could be similar to that study, but there are substantial differences that deserve to be communicated:

– Here we tested the hypothesis extracted from this previous paper. There, we suggested CPI in AB, but did not test them. This is the main novelty of this work because, we previously focused only on intra-cardiac (adverse LV–RV) interactions not on CPI, and on NT-proBNP as a predictive tool for LV dysfunction and also severity, not on echocardiographic parameters.

• We added for the first time LU to our study protocol. The pulmonary variables did not include LU in the previous work and therefore the more detailed pulmonary status could be delineated here, also according to current clinical practice where LU is increasingly included in the work-up of these patients. We think that simultaneous CPU led us to evaluate CPI better than blood analysis. We would like to remark that simultaneous CPU is most easy to do than simultaneous blood extraction and ultrasound in clinical practice (usually delayed).

• One of the main critiques by colleagues and reviewers during the review process of our previous articles was that blood extractions in children with AB are invasive, painful... and do not add enough to modify the current recommendations of sitting to watch what is going to occur. Therefore, we tried to identify non-invasive tools for this purpose. LU has been previously demonstrated to be an accurate tool, but you know that there is not a perfect predictive tool, so we used our previous echocardiographic observations to increase the pool of tools for this purpose. Interestingly, No data about predictive accuracy (ROC curves values) for severity of any

10

echocardiographic parameter was provided in previous works and this was one of the main novelties of this study. At that study we assessed LV Dysfunction with TEI index by Tissue Doppler echo that require deep training and large experience (performed by cardiologists) to be adequate applied. Here we provided more reliable parameters that could be easily learned by non-cardiologist in a POCUS, such as pulmonary acceleration time/ejection time ratio or left ventricular eccentricity index.

• Taking all together, with the observations from these 2 works, we suggest that in previously healthy infants a worse respiratory state induced raised pulmonary pressures, increased RV strain and subsequent LV strain. These cardiac alterations could be easily screened by NT-proBNP, and the presence of these alterations are related with complicated bronchiolitis. We finally suggest that CPU could be an interesting non-invasive tool to use in this setting with adequate accuracy.

• Finally, What would be the best strategy? Invasive/painful blood extraction adding biochemical markers (pH, NT-proBNP)?? Training pediatricians on non invasive CPU (the time to train clinicians in CPU is a major limitation for this approach) ?? Even only sitting to watch clinical scores?? We think that there is not a unique, universal, and adequate answer. Maybe different combinations of different tools according availability at different centers is the correct answer. Therefore, the publications of works such as this one could provide evidence that there are different tools available for this purpose, and each clinician/institution will choose those that best suit their circumstances.

• We comment on this issue in the revised manuscript (at the end of the discussion).

Round 2

Reviewer 2 Report

The provided answers by the authors seem to be justifying and enough.