PEER REVIEW HISTORY · PEER REVIEW HISTORY . BMJ Paediatrics Open publishes all reviews undertaken...

PEER REVIEW HISTORY

BMJ Paediatrics Open publishes all reviews undertaken for accepted manuscripts. Reviewers are

asked to complete a checklist review form and are provided with free text boxes to elaborate on their

assessment. These free text comments are reproduced below.

This paper was submitted to a another journal from BMJ but declined for publication following peer

review. The authors addressed the reviewers’ comments and submitted the revised paper to BMJ

Paediatrics Open. The paper was subsequently accepted for publication at BMJ Paediatrics Open.

ARTICLE DETAILS

TITLE (PROVISIONAL) Non-pathologic Bilious Vomiting Complicating Therapeutic

Hypothermia for Hypoxic Ischemic Encephalopathy in Neonates: A

retrospective cohort study

AUTHORS Foster, Jennifer; Hukui, Julie; Jones, Sarah; Coughlin, Kevin; Levin, Simon

VERSION 1 - REVIEW

REVIEWER Burge, David University Hospital Southampton, 1. Department of Paediatric Surgery and Urology Competing Interests: I have no competing interests

REVIEW RETURNED 16-Feb-2017

GENERAL COMMENTS This is a well-written paper. It is a retrospective cohort study using matched controls from an era prior to the use of therapeutic hypothermia and they report their experience with bilious vomiting or drainage. The presence of BVD is indeed a worrying feature that could lead to unnecessary radiological or even surgical intervention as they describe. The results of this study indicating that BVD is quite common in TH infants in the absence of intestinal pathology will help clinicians faced with this scenario. Having said that this reviewer has been closely involved leading a neonatal surgical service based on a combined tertiary neonatal medical and surgical unit during the TH era and I have never been made aware of an infant undergoing TH who has had BVD! I suggest some minor changes: The statement at the beginning of para 2 in the Introduction that 2/5 infants with bile vomiting (BV) have intestinal pathology would best be amended. Not only is Reference 6 somewhat out of date but it has been misquoted - it reported that 38% of infants referred with BV had intestinal pathology, which equates to >1/3 rather than 2/5. There has been a considerable increase in referrals of infants with BV over recent years and there is more recent literature on this issue that might be better cited: Term newborns with bilious vomiting: When should they see a surgeon and how soon? Simon C Blackburn Archives of Disease in Childhood 2014; 100 1-2 Outcomes of full-term infants with bilious vomiting: observational

on February 26, 2020 by guest. P

rotected by copyright.http://bm

jpaedsopen.bmj.com

/bm

jpo: first published as 10.1136/bmjpo-2017-000034 on 31 A

ugust 2017. Dow

nloaded from

http://bmjpaedsopen.bmj.com/

study of a retrieved cohort. Mohinuddin S et al Archives of Disease in Childhood 2014; 100 14-17 The burden of excluding malrotation in term neonates with bile stained vomiting. Drewett M, et al Pediatr Surg Int. 2016 I presume that the control group used comprised infants who also had HIE? This is not entirely clear from the description used and it would be a good idea to formally state this.

REVIEWER Reviewer 2

REVIEW RETURNED 11-Mar-2017

GENERAL COMMENTS Methods: The authors describe a standardised inclusion criteria for cooling. The authors may want to describe their standard practices with respect to their management of gastric tubes and aspirates. Page 5 , line 15: clarify what NG stands for. Page 7, line 7: clarify how sepsis was diagnosed. Results: the following information in the results are missing: the nasogastric tube: (position, size, etc) and about the aspirates. It is important to describe the results of this paper in a wider perspective of the significance of their results. None of the babies had an obstructive cause or a positive surgical diagnosis. I suggest the authors bring this out in the title, their abstract, their discussion and conclusion.

VERSION 1 – AUTHOR RESPONSE

Reviewer 1:

I suggest some minor changes: 1. The statement at the beginning of para 2 in the Introduction that 2/5 infants with bile vomiting (BV) have intestinal pathology would best be amended. Not only is Reference 6 somewhat out of date but it has been misquoted - it reported that 38% of infants referred with BV had intestinal pathology, which equates to >1/3 rather than 2/5. There has been a considerable increase in referrals of infants with BV over recent years and there is more recent literature on this issue that might be better cited: Term newborns with bilious vomiting: When should they see a surgeon and how soon? Simon C Blackburn Archives of Disease in Childhood 2014; 100 1-2 Outcomes of full-term infants with bilious vomiting: observational study of a retrieved cohort. Mohinuddin S et al Archives of Disease in Childhood 2014; 100 14-17 The burden of excluding malrotation in term neonates with bile stained vomiting. Drewett M, et al Pediatr Surg Int. 2016 >>Response: We appreciate the reviewer’s comments and have incorporated a broader discussion of

the incidence of intestinal pathology in the setting of neonatal bilious emesis into the introduction of our paper. This includes the addition of data from 3 references from 2010 onwards. The following is the new paragraph:



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


In the non-HIE population the incidence of true intestinal pathology, such as midgut volvulus or intestinal atresia, in the presence of bilious vomiting varies in recent literature. It depends on the reference population and study site characteristics, the rate of referral for an episode of bilious emesis, and the surgical diagnoses being considered, but is as high as 46%.

6-9

2. I presume that the control group used comprised infants who also had HIE? This is not entirely clear from the description used and it would be a good idea to formally state this. >>Response: We have included the following phrase in the description of the control population to

ensure clarity on the presence of HIE: The cohort was matched 1:1 for unit of care (NICU versus PICU), GA, and Sarnat score with

historic controls from 2001 (PICU) and 1996 (NICU) who had HIE and did not receive TH. Reviewer: 2 Comments to the Author Methods: The authors describe a standardised inclusion criteria for cooling. The authors may want to describe their standard practices with respect to their management of gastric tubes and aspirates. >>Response: The following description of NG tube management has been included: All nasogastric (NG) tubes were positioned in the stomach and assessed clinically and by x-

ray. Positioning was checked for depth of insertion by the number taped at the nares at the

beginning of each nursing shift. Bilious NG aspirates were detected either by spontaneous

drainage from an open NG or by manual aspiration of green bile.

Page 5 , line 15: clarify what NG stands for. >>Response: We thank the reviewer for noting this omission. “NG” has been defined (nasogastric) in

the first appearance within the manuscript (methods section, paragraph 2, line 1) and has been added to the list of abbreviations in the title page.

Page 7, line 7: clarify how sepsis was diagnosed. >>Response: Positive blood cultures were used as a surrogate for sepsis within this study, as it was

not feasible to retrospectively assess for the consensus group definition of sepsis in each patient. We have made this clearer by replacing the word “sepsis” with “positive blood culture” in several instances, and included the following statement in the methods:

sepsis was defined as positive blood culture.

and the following phrase in the results:

Blood cultures were taken and results available in 75 patients (45 TH patients, 30 control patients)

as a surrogate for sepsis.

Results: the following information in the results are missing: the nasogastric tube: (position, size, etc) and about the aspirates. >>Response: Regarding the NG tube positioning, we have clarified that we considered any NG tube

that did not have positioning confirmed as malpositioned: Incorrect NG position past the pylorus (including NG position not documented), We have provided a broader description of the bilious nasogastric aspirates in the results section:



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


Two patients in the TH group experienced emesis, while all other episodes consisted of bilious NG drainage in both the TH (median volume 15.5mL, IQR 40mL) and control (12mL and 25mL for 2 patients) groups.

It is important to describe the results of this paper in a wider perspective of the significance of their results. None of the babies had an obstructive cause or a positive surgical diagnosis. I suggest the authors bring this out in the title, their abstract, their discussion and conclusion. >>Response: We agree with the reviewer on this comment and have attempted to bring it out by

referring to non-pathologic bilious vomiting and drainage. To this end we have added the word “Non-pathologic” to the title, have added a comment about the clinical results of the BVD to the abstract: Three patients underwent investigation and no patients had surgical pathology.

We have also added a comment in the discussion emphasizing the non-surgical diagnoses from our patients: though it does not appear to be pathologic in our patient population, as no patients demonstrated an anatomic cause,

We have not altered the conclusion as we felt that “Therapeutic hypothermia through whole body cooling appears to be associated with non-pathological bilious vomiting or gastric drainage” adequately indicated that the adverse effect was not caused by a surgical pathology and we feared unnecessarily lengthening the conclusion section with a longer description. We also felt that our research did not provide strong enough justification to allow us to recommend either for or against investigation for the bilious emesis, or for a stronger statement concerning lack of risk of surgical pathology, and so avoided any broader statements than a summary of the study findings.

VERSION 2 – REVIEW

REVIEWER Gale, Chris Imperial College London, United Kingdom Competing interests: None

REVIEW RETURNED 21-Apr-2017

GENERAL COMMENTS This paper reports retrospective analysis of a cohort of infants who underwent therapeutic hypothermia with comparison to a matched, historical control group who did not undergo therapeutic hypothermia. The outcome of interest was bile stained vomiting or nasogastric tube aspirates. The main finding is that bilious aspirates/vomiting was significantly more common among infants receiving therapeutic hypothermia than historical controls. This is an important topic that is relevant to clinicians caring for infants undergoing therapeutic hypothermia. I was particularly interested to see that one baby had an unnecessary laparotomy, showing the potential harms. I have major and minor points: MAJOR: 1. My main concern relates to detection bias, i.e. bilious aspirates being looked for more closely after therapeutic hypothermia was introduced - the authors note "it was our clinical impression that the incidence of bilious vomiting or drainage (BVD) in infants undergoing TH was high" - my worry is that the outcome of interest (bilious vomiting/aspirates) was looked for subconsciously, found, and therefore recorded more in the therapeutic hypothermia epoch. This, rather than any true effect of therapeutic hypothermia, may explain the study findings. I would be more reassured if the authors could



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


describe the nursing protocols that were in place at the study sites and if these were consistent before and after the introduction of therapeutic hypothermia - A robust and consistent nursing protocol would be reassuring. 2. Figure 1 is not included so I cannot see the patient inclusion and flow. 3. I would like to know how many babies were eligible before matching, and the baseline characteristics of all potential babies before they placed into matched groups, as well as the ones in Table 1 after matching. 4. The study is relatively small and only represents a single NICU and a single PICU - it would also be useful to see how many infants were cared for in either NICU or PICU as well and the results by site. This might offer some insight into whether detection bias was operating more at one site. 5. Was there a pre-published protocol? 6. Can the authors please provide more detail on the time period that they used when reviewing the notes for bilious aspirates/vomiting? Did they check the notes for the same duration of time for both the hypothermia and the control groups? How long was this period (I expect it was about 132 hours in total to reach 48 hours post rewarming) and can the authors please demonstrate that it was the same in both epochs. 7. Please can the data describing the "risks of bilious aspirates/vomiting" (e.g. NG position, opioid use, muscle relaxant use) be included for both the matched groups and the groups before matching. 8. In table 1, the column for p values should be removed as it is not an appropriate test in this situation, especially for the characteristics that the groups were matched on!! 9. The authors need to include data on mortality for both groups, including when infants died. At the moment it is not clear to me whether any of the babies died within the study period or if this differs between the hypothermia and control epochs (i.e. did more babies die sooner in the pre-cooling period which might mean that there was less time for bilious vomiting/aspirates to be detected). Similarly, it would be good to see if any babies were moved to palliative care (where bilious aspirates may not have been looked for or recorded as consistently). MINOR: 10. The main results would be clearer if included in a table rather than solely in the text. 11. Number needed to treat (to harm in this case) should be presented with 95% confidence intervals

REVIEWER Ojha, Shalini University of Nottingham UK Competing interests: None

REVIEW RETURNED 17-May-2017

GENERAL COMMENTS This is an interesting piece of retrospective analysis – as our experience with therapeutic hypothermia increases, it is valuable to review and periodically analyse any potential issues associated with this extremely beneficial modality. In this respect, the study is useful and highlights the point that clinical assessment of infants undergoing therapeutic hypothermia requires careful consideration.



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


The effects of hypothermia on physiological functions of the different organ systems must be taken into account prior to subjecting the infant to unnecessary interventions. Major comments: 1. A Figure 1 is mentioned in the text but I am unable to see any figure in the manuscript. 2. Page 5, line 52: Why where the historic controls taken from two different years? What was the practice re: NG Tube placement and position checking at these times in the two units? 3. Also, literature form the UK suggests that the practice of investigating infants with bilious vomiting has become more popular over the last decade. It would be useful to know if the authors’ centre has also undergone a similar change in the way infants with bilious vomiting are treated i.e. is it more likely that an infant with bilious vomiting or NG aspirate gets investigated for anatomical pathologies in the epoch of the cases as compared to the time from when the controls have been derived. 4. Only 3 out of 14 cases had bilious vomiting. Most infants had bile drained from the NG tube. Although the authors have mentioned the malposition of the NG tube beyond the pylorus as a potential reason for bilious aspirates, they have not specified how many cases and/or controls had such mal-positioned NG tubes. As it would be perfectly usual to obtain bilious aspirated from an NG tube that was inserted to reach beyond the pylous, I wonder if the primary analysis of the risks and odds of “BVD” should have been calculate after excluding such infants where the NG tube was mal-positioned in this manner. I think this would be more useful that just using this information as a potential confounder. Minor comments: 1. Please review all the numerical values reported in the manuscript and ensure that all numbers are reported to a consistent number of decimal places, such as Abstract Line 34: [95%CI 1.316-42.84] – the CI values are three and two decimal places, receptively. 2. Please provide full forms for all abbreviations, when mentioned for the first time, such as Page 5, line 22: “GA” 3. Table 1 - please give total number of infants in each group - please correct to ensure consistent use of parenthesis for (mean [SD]) - some brackets are missing.


MAJOR:

1. A Comment on detection bias is required in the discussion; , i.e. bilious aspirates potentially

being looked for more closely after therapeutic hypothermia was introduced - the outcome of

interest (bilious vomiting/aspirates) may have been looked for subconsciously, found, and

therefore recorded more in the therapeutic hypothermia epoch. This, rather than any true

effect of therapeutic hypothermia, may explain the study findings. The authors should

therefore describe the nursing protocols - robust and consistent nursing protocols would be

reassuring. It would also be helpful to have more detail on what the practice re: NG Tube

placement and position checking was the different epochs in the two units?



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


>> Response: We have incorporated a broader and more detailed description of our nursing protocols in both units of care. The following description of NG tube management and feeding has been included, paragraph 2 of methods:

A standard nasogastric or orogastric (collectively abbreviated as NG) tube insertion, maintenance,

and feeding protocol was in place during the study periods in both units. The NG tube was first

inserted as per current Neonatal Resuscitation Program recommendations.(10) The position was

checked by aspiration of stomach contents, auscultation for air entry "pop", and confirmed by x-ray (2

views in the NICU and 1 AP view in the PICU). Incorrectly positioned gastric tubes were repositioned

to sit in the body of the stomach. Final insertion depth was noted at the nares and recorded on the

daily nursing flow sheet. The position of the NG tube was checked once per nursing shift by

confirming depth of insertion, auscultation for air entry and aspiration of air/gastric contents. If

concerned, x-ray confirmation was obtained. For purposes of the analysis, the NG tube was

considered malpositioned any time that its position was not confirmed within the first day of therapy or

its position was not confirmed by nursing. Bilious NG aspirates were detected either by spontaneous

drainage from an open NG or by manual aspiration of green bile at one of the regular position checks.

Infants with HIE in the pre-TH era did not receive enteral feeds for at least the first 48 hours after

birth, and those receiving TH did not receive enteral feeding until rewarming was complete.

Thereafter, feeds for all infants followed a hospital-based protocol that included trophic feeding of

20mL/kg/day continuously with checks for gastric residuals every 4 hours, and advancement of feeds

by 15-20mL/kg/day when gastric residuals were minimal.

This paragraph required a reference, and so reference 10 is now for the Textbook of Neonatal Resuscitation, and all subsequent references in the text have been numbered accordingly. We also acknowledge the real potential for detection bias in the later (hypothermia) era. We began this study fairly soon after it was noticed, specifically because we had noticed it and wanted to determine whether we had been missing something important for several years. We have altered the description of the rational for undertaking the study in the introduction, to more accurately reflect the events leading to the study: In our centre, two recent patients with non-pathologic bilious emesis who received investigation prompted an examination for bilious vomiting as a complication of TH.it was our clinical impression that the incidence of bilious vomiting or drainage (BVD) in infants undergoing TH was high, and may have resulted in unnecessary invasive testing, including surgical exploration We have included the following paragraph describing this in the discussion:

Patients in the later, TH, era may have been more closely observed both in general and in particular

for the presence of BVD if this was unconsciously noted by nursing staff or other physicians prior to

the identifying the need for a study.

2. Figure 1 needs to be included (as it is not) so readers can see the patient inclusion and flow.

>> Response: We thank the reviewer for noting this omission. Figure 1 can be found after the

reference section. The inclusion and flow is as follows:

Figure 1: Patient inclusion and flow



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


3. Information should be provided on how many babies were eligible before matching, and the

baseline characteristics of all potential babies before they placed into matched groups, as well

as the ones in Table 1 after matching.

>> Response: For the cohort group (TH group) all patients who were eligible before matching (n=47)

were matched. This information is provided in Figure 1. Thus, the information provided in Table 1

reflects all eligible patients both before and after matching. We have made this clearer in the title for

Table 1: Baseline characteristics of matched TH and Control patients.

Additionally, we have added the following clarification to the beginning of paragraph 5 (was paragraph

3) in the methods, which describes our matching process and allows the reader to understand why



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


the era of examination for control patients was different in NICU (1996-2008) and PICU (2001-2008),

where unit of care was used as a matching criterion:

Matching proceeded in a retrograde fashion, such that the most recent patients with HIE from the pre-

cooling era were examined for a potential case match first, and we worked backward in time until all

cases were matched.

For the babies from the pre-TH era (matched control group), which we have defined as 2008 and

earlier, demographic information was only obtained on historical controls that were appropriately

matched. From 1996 to 2008, there was a total of 99 and 91 potentially eligible controls (patients with

a diagnosis of HIE) from the PICU and NICU, respectively. If we were unable to match a historic

patient with a study patient based on gestational age, Sarnat score, and unit of care, they were not

considered as a control and the chart was not reviewed further. Thus, we do not have data on the

potential control babies before matching.

4a.The study is relatively small and only represents a single NICU and a single PICU - it would

also be useful to see how many infants were cared for in either NICU or PICU as well and the

results by site. This might offer some insight into whether detection bias was operating more

at one site.

>> Response: We have included the number of newborns cared for annually in each of the NICU and

PICU to provide an estimate of volume. This is provided in the methods section. We had originally left

out the explanation because, with a hospital merger that occurred in 2012 that altered patient

destinations and therefore volumes, the explanation is somewhat long. We have added the following

paragraph to the methods:

The neonatal intensive care unit (NICU) of St. Joseph’s Health Centre cared for approximately 850

neonatal patients annually until 2011, when it moved to LHSC and assumed care for all inborn and

premature patients (900-1000 annually). The pediatric intensive care unit (PICU) of LHSC cared for

approximately 150 neonatal patients annually until 2011, and 50-100 since.

The results by site have been added to Table 2. The following sentence has been added to the

results, at the end of paragraph 2 in the “patient outcomes” subsection:

The unit of care (PICU versus NICU) was not significantly associated with the proportion of patients

who experienced BVD (see Table 2) in either the TH (p=0.33) or control (p=0.21) groups.

b. The historical data for the neonatal group is very old (1996 and from a different date than in

PICU 2001). The age of the data and the differences merit a mention in the discussion. Much

could have changed in regard to feeding and other aspects of care in this time frame (breast

feeding, expression, inotrope use, etc).

>>Response: This is a salient point, and one we should have noted in the first submission. We have

added the following to the discussion (paragraph 7, after sentence 2):



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


It bears mentioning that the control population for the PICU was taken from 2001-2008, while we had

to continue back an additional 5 years, to 1996, to find matching patients for the NICU control

population. Although the management of feeding tubes was protocolized and consistent across the

eras, differences may have existed in feeding practices, ventilation strategies, sedation and pain

management, and numerous other care practices.

5. Was there a pre-published protocol; if not this should be mentioned.

>> Response: There was no pre-published research protocol. We have included a sentence to this

effect in paragraph 1 of the methods section:

The study protocol has not been previously published.

6. Can the authors please provide more detail on the time period that they used when

reviewing the notes for bilious aspirates/vomiting? Did they check the notes for the same

duration of time for both the hypothermia and the control groups? How long was this period (I

expect it was about 132 hours in total to reach 48 hours post rewarming) and can the authors

please demonstrate that it was the same in both epochs.

>> Response: The time periods that were reviewed for the cases were based on the time spent in

each of the cooling phases. The time periods reviewed for the control group was based on 6 hours

pre-cooling, 72 hours of cooling, 8 hours rewarming, and 48 hours post-rewarming, resulting in a total

of 134 hours. We made the following modifications to the methods section to clarify this point:

The presence, time, and volume of BVD was documented for the pre-cooling hours of life (hours 0 to

6 in the control group), during cooling in 12-hour intervals for 3 days (hours of life 7-78 for controls),

rewarming (hours of life 79-86 for controls), and in 12-hour intervals for 48 hours after rewarming was

complete (hours 87-134 for controls).

In addition, we have added the mean observation time for both the control and study groups into the

results section, to address concerns about the possibility of increased risk of finding BVD if one

examines a population for longer. The following phrases have been added to the “patient outcomes”

section of the results:

Based on data availability, the mean (standard deviation [SD]) observation time for patients was 129.8

(11.6) hours for the TH group and 125.5 (25.9) hours in the control group (p=0.3), Figure 1.

7. Please can the data describing the "risks of bilious aspirates/vomiting" (e.g. NG position,

opioid use, muscle relaxant use) be included for both the matched groups and the groups

before matching.

>> Response: We have included this data in the new Table 2. As the entire study population was

matched, we did not include a pre- and post-matching description.



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


Table 2. Incidence of BVD and presence of potential confounders in TH and control groups

Patient exposure TH group

(n=47)

Control group

(n=47)

p value

BVD 12 (26) 2 (4) 0.04

PICU care (n=22) and BVD, n(%) 4 (18)* 2 (9)* 0.7

NICU care (n=25) and BVD, n(%) 8 (32)* 0 (0)* 0.004

NG position abnormal or not

documented on first day, n (%)

5 (11) 15 (32) 0.06

NG abnormal and BVD, n(%) 2 (14) 1 (8) 1.00

Opioid used at any point, n(%) 40 (85) 24 (51) 0.0004

Opioid used and BVD, n(%) 11 (28) 1 (4) 0.02

MR used at any time, n(%) 11 (23) 11 (23) 1.00

MR use and BVD, n(%) 3 (27) 0 (0) 0.21

Blood culture (n=45 TH, n=30 control)

positive, n(%)

3 (7) 1 (3) 0.53

Blood culture positive and BVD, n(%) 0 (0) 0 (0) n/a

*Comparison of proportion of patients with BVD between unit of care (PICU versus NICU) was not

significant in either the TH (p=0.33) or control (p=0.21) groups.

8. In table 1, the column for p values should be removed as it is not an appropriate test in this

situation, especially for the characteristics that the groups were matched on. The data should

simply be presented without p values

>> Response: We have removed the comparisons from table 1, which no longer has a column for p

values.

9. The authors need to include data on mortality for both groups, including when infants died.

At the moment it is not clear whether any of the babies died within the study period or if this

differs between the hypothermia and control epochs (i.e. did more babies die sooner in the

pre-cooling period which might mean that there was less time for bilious vomiting/aspirates to

be detected). Similarly, it would be good to see if any babies were moved to palliative care

(where bilious aspirates may not have been looked for or recorded as consistently).



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


>> Response: We thank the reviewer for pointing this out as we had not considered it and it was a

significant omission.

We have added the following paragraph to the results:

Two study patients and 6 control patients died during the period of observation (p=0.3), none of whom

experienced BVD. Two study patients and 3 control patients died from non-gastrointestinal causes

after the period of observation and prior to hospital discharge.

While we have mortality data and timing, unfortunately we do not have complete data on movement to

palliative care, and so have made note of this in the discussion as a possible source of error:

Also, while we noted the time to death, it is possible that some of the patients who died either during

observation or shortly after had been made palliative with no examination for gastric aspirates, which

would represent missed cases of BVD.

While examining the data and re-examining charts for mortality data, it became clear that we had

made a significant error in data entry. Our research assistant had entered a string of zeros into all

data fields after a patient had died, rather than leaving the fields blank. This had an effect on the

logistic regression analysis, as several fields involving presence or absence of opioids, muscle

relaxants, and correct NG positioning had a value (0) in them rather than being blank. We sought

statistical consultation and have redone the calculations. The odds ratios as well as the variance and

reliability of the regression models has changed, paragraphs 3 and 4 of the “patient results”

subsection of the Results section. We apologize for this original error, and are grateful that the

reviewer’s comment forced a re-examination of the data.

Paragraph 3:

Accounting for the first 3 of these potential confounders, the increased risk of BVD with TH persisted,

OR=7.8 [95%CI 1.4-43.3]. This model accounted for between 10.7% and 18.8% of the variance in the

presence of any BVD and was found to be a reliable model (p=0.77, Hosmer Lemeshow test p>0.05

significant) that accurately predicted 75.6% of cases.

Paragraph 4:

When including presence of positive blood cultures in the linear regression model, with only 75

patients in the model, the OR for BVD was 11.6 [95%CI 1.2-115.0]. The model accounted for between

15.0 and 25.7% of the variance in the presence of BVD, and had good reliability in terms of the

agreement between the outcome and the predictors (Hosmer & Lemeshow test, p=0.84).

10. It is suggested in the introduction that TH might result in unneccessary invasive

investigation and potentially surgical exploration. The investigative approach taken in these

cohorts should be described to reassure the reader that diagnoses may not have been missed.



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


It is noteworthy that only 3 patients received a surgical consult.

>> Response: We appreciate the reviewer’s comments on the concern for possible missed pathology,

and agree that it is possibly concerning that only 3 patients received surgical consult, or at least

noteworthy that only 3 patients were deemed concerning enough by the clinical service to warrant

investigation. Ideally, all of the patients would have received some form of investigation to reassure

the reader that the majority of these cases are benign. Unfortunately, as a retrospective study the

clinical and investigative approach taken was completely out of our control. However, in lieu of having

the investigations, we do have information on the presence or absence of clinically significant

pathology to the end of the birth hospital admission, and have included this information with changes

to the last paragraph of the results section:

There was no formal hospital protocol for assessment of patients after experiencing BVD. Of patients

with BVD, 3 patients in the TH group (both patients with vomiting, 1 patient with drainage) received

surgical consult and subsequently had an upper gastrointestinal contrast series performed. One study

patient who had concerning radiographic findings received an exploratory laparotomy with no

pathology identified. None of the remaining 11 patients with BVD had investigations performed, and

there were no notable sequelae during the birth admission to indicate missed pathology.

The following sentence has been added to the end of the second last paragraph of the discussion:

It is noteworthy that only 3 patients in this study were deemed concerning enough by the clinical

service to warrant investigation, though no patients ultimately had surgical pathology.

11. Only 3 out of 14 cases had bilious vomiting. Most infants had bile drained from the NG

tube. The authors should specify how many how many cases and/or controls had such mal-

positioned NG tubes. The authors should consider undertaking the primary analysis of the

risks and odds of “BVD” after excluding such infants where the NG tube was mal-positioned

as a sensitivity analysis. Alternatively they should justify why this approach was not taken.

>> Response: The reviewer raises a good point. We, too, were concerned about the risk of

malpositioned NG tubes in the duodenum causing an increase in apparent bilious drainage, or in the

esophagus causing a decrease in the apparent bilious drainage. We collected information for every

patient on initial confirmation of NG position, and on confirmation of NG position within each of the

time periods assessed (first 6 hours, next 72 hours in 12-hour increments, 8 rewarming hours, and

final 48 hours in 12 hour increments). We acknowledge that we had not clearly described the patients

with incorrectly positioned NG tubes and how this was determined, and so this has been addressed

both by incorporating the actual numbers into Table 2, and by better describing that we considered

any instance where the correct position of the NG tube was not confirmed during the time period of

interest as an incorrectly-placed NG tube.

We have made it clearer in the methods section that we looked at regular documentation of correct

NG positioning:

For each of these time periods we also documented verification of NG positioning, use of opioids, and

use of muscle relaxants.



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


We did not find cases of the NG malpositioned beyond the pylorus, a finding that makes clinical sense

given the poor gut motility expected in these patients. However, for purposes of analysis we

considered the NG tube malpositioned whenever the location was not documented within the first day

of treatment. We have made this analysis assumption clearer in the methods, paragraph 4:

For purposes of the analysis, the NG tube was considered malpositioned any time that its position

was not confirmed within the first day of therapy.

In our analysis, we accounted for incorrectly positioned NG tube (or NG position not documented) by

incorporating it into the regression model reported in the results section.

MINOR:

1. The main results would be clearer if included in a table rather than solely in the text.

>> Response: The main result has been included in Table 2.

2. Number needed to treat (to harm in this case) should be presented with 95% confidence

intervals.

>> Response: We thank the reviewers for noting this oversight. We have now reported the NNH as

follows:

The number needed to treat to create one episode of BVD attributable to TH (number needed to

harm) was 5 (95% CI 3-14).

3. Please review all the numerical values reported in the manuscript and ensure that all

numbers are reported to a consistent number of decimal places, such as Abstract Line 34:

[95%CI 1.316-42.84] – the CI values are three and two decimal places, receptively.

>> Response: We have reviewed the manuscript and verify that, taking significant figures into

consideration, all numbers are now reported to an appropriate and consistent number of decimal

places for the given variable.

4. Please provide full forms for all abbreviations, when mentioned for the first time, such as

Page 5, line 22: “GA”.

>> Response: We appreciate the reviewer noting this omission. We have made the following changes

to the manuscript:



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


Criteria for TH were defined in previously-published guidelines:1 gestational age (GA) ≥ 36 weeks; ;

5. Table 1 - please give total number of infants in each group

- please correct to ensure consistent use of parenthesis for (mean [SD]) - some brackets are

missing.

>> Response: These errors have been corrected in Table 1.

ADD the word retrospective to the title before cohort

>> Response: The title is now as follows:

Non-pathologic Bilious Vomiting Complicating Therapeutic Hypothermia for Hypoxic Ischemic

Encephalopathy in Neonates: A retrospective cohort study

Additional changes:

In order to shorten the manuscript to come closer to a 2500 word count, we have removed several

sentences in the discussion. These have been indicated with track changes.

VERSION 3 – REVIEW

REVIEWER Gale, Chris Imperial College London, United Kingdom Competing interests: None

REVIEW RETURNED 04-Jul-2017

GENERAL COMMENTS The authors are to be congratulated on the thorough responses given to the reviewers comments. The paper is much improved by the additional information and more complete discussion of the limitations. My only minor comments are: 1. In the tables and figure please can all the abbreviations be defined in the key 2. "APGAR" is not an acronym and should be "Apgar"(Figure 1)

REVIEWER Ojha, Shalini University of Nottingham UK Competing interests: None


GENERAL COMMENTS Thank you for the reponse to the comments.

REVIEWER Hilliam, Rachel Mary The Open University



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


UK Competing interests: None


GENERAL COMMENTS This is an interesting paper, with a good study design and well analysed. It would be helpful to note in the abstract that the test used for the positive blood culture in the regression model is the odds ratio (this is stated in the paper but worth making clear in the abstract also). The first sentence in the second paragraph of the introduction doesn't quite make sense as it currently stands. The one to one matching is clearly explained. It would be useful to have an idea of the number that were matched and not matched due to gender. Given the explanation I'm assuming that it is by chance that there are 24 males in the TH group and 25 in the control group as it suggests that gender matching was dropped for a number, so I think table 1 may give a false impression here. What is of interest is the number of males that were correctly matched in the proccess, not the total number of males. The statistical analysis is appropriate and clearly explained. There is one section however which is not quite correct. Hosmer Lemeshow is a goodness of fit test which gives an idea of how well the logistic regression fits the data. This means that the null hypothesis is that the observed and expected proportions are the same. You are correct in the fact that your p value gives no reason to suggest the model does not fit the data, but you cannot state that p>0.05 is significant (this is not correct). p<0.05 is significant, but in this particular test significant (ie p<0.05) means there is evidence to suggest the model does not fit. This needs re-wording before going to print. It would be helpful to the reader to specify what test p=0.3 is referring to in the final sentence of the results section. It would be worth going through the paper and correcting the degree celsius, sometimes this is given with a subscript 'o' and sometimes a subscript '0', I would suggest the former should be used throughout. The paper is well written and should be of interest to the readers of this journal.


Reviewer #1:

1. In the tables and figure please can all the abbreviations be defined in the key >>We have provided a key for each of the tables and figures. In table 2 we have spelled out Muscle Relaxant, as this was not previously defined in the key for the paper.

2. "APGAR" is not an acronym and should be "Apgar"(Figure 1) >>This has been changed in the Figure. Reviewer #2:



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


1. It would be helpful to note in the abstract that the test used for the positive blood

culture in the regression model is the odds ratio (this is stated in the paper but worth making clear in the abstract also).

>> We have added this into the abstract. The sentence indicating this now reads: The group exposed to TH was more likely to experience BVD compared to the control group with a relative risk of 6.0 [95% CI 1.4-25.4], even after accounting for improper or unchecked nasogastric position, opioids, and muscle relaxant use, OR=7.8 [95%CI 1.4 - 43.3], and when positive blood culture was included in the regression model, OR=11.6 [95%CI 1.2 - 115.0].

2. The first sentence in the second paragraph of the introduction doesn't quite make sense as it currently stands.

>>We have changed the wording and order of the second paragraph to improve readability and flow. The changes are indicated in the marked manuscript under “track changes”. It currently reads as follows: In our centre, two recent patients had non-pathologic bilious emesis requiring investigation while undergoing TH. In the general neonatal population with bilious vomiting, the incidence of true intestinal pathology, such as midgut volvulus or intestinal atresia, is as high as 46%.(6-9) Bilious vomiting or drainage (BVD) has not previously been described as a side effect of TH and the incidence in this population is unknown.

3. The one to one matching is clearly explained. It would be useful to have an idea of the number that were matched and not matched due to gender. Given the explanation I'm assuming that it is by chance that there are 24 males in the TH group and 25 in the control group as it suggests that gender matching was dropped for a number, so I think table 1 may give a false impression here. What is of interest is the number of males that were correctly matched in the proccess, not the total number of males.

>> We agree with the reviewer about this point. We have added an explanation to the key of Table 1, which described the numbers of males unmatched for gender in each group, which reads as follows: *One male in the TH group and 2 in the control group were not matched for gender.

and have changed the paragraph describing our handling of the gender matching (or inability to match in some cases) to read as follows: We also attempted to match 1:1 for gender to ensure similar baseline characteristics and risk. It

was agreed that this characteristic would be least likely to affect baseline risk of BVD and so

when matching was not possible the requirement was eliminated rather than excluding a member

of the interventional cohort, resulting in 3 pairs unmatched for gender.

4. The statistical analysis is appropriate and clearly explained. There is one section

however which is not quite correct. Hosmer Lemeshow is a goodness of fit test which gives an idea of how well the logistic regression fits the data. This means that the null hypothesis is that the observed and expected proportions are the same. You are correct in the fact that your p value gives no reason to suggest the model does not fit the data, but you cannot state that p>0.05 is significant (this is not correct). p<0.05 is significant, but in this particular test significant (ie p<0.05) means there is evidence to suggest the model does not fit. This needs re-wording before going to print.

>> We thank the reviewer for this point on the statistical analysis, which improves the accuracy of our report. We have altered the relevant section in the results to read as follows: This model accounted for between 10.7% and 18.8% of the variance in the presence of any BVD and was found to be a reliable model (p=0.77, Hosmer Lemeshow test of event rates observed versus



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


rates predicted by the model, where p<0.05 indicates a significant difference) that accurately predicted 75.6% of cases.

5. It would be helpful to the reader to specify what test p=0.3 is referring to in the final sentence of the results section.

>>This has been clarified as follows: Two study patients and 6 control patients died during the period of observation (p=0.3, Fisher’s exact 2-tailed test)

6. It would be worth going through the paper and correcting the degree celsius, sometimes this is given with a subscript 'o' and sometimes a subscript '0', I would suggest the former should be used throughout.

>>We thank the reviewer for noting this discrepancy and have corrected it to read oC throughout the

manuscript.



jpaedsopen.bmj.com

/bm


ugust 2017. Dow

nloaded from


PEER REVIEW HISTORY · PEER REVIEW HISTORY . BMJ Paediatrics Open publishes all reviews undertaken...

Documents

Transcript of PEER REVIEW HISTORY · PEER REVIEW HISTORY . BMJ Paediatrics Open publishes all reviews undertaken...