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. TITLE (PROVISIONAL) AUTHORS REVIEWER REVIEW RETURNED GENERAL COMMENTS ARTICLE DETAILS Impact of Hypoglycaemia on Neurodevelopmental Outcomes In Hypoxic Ischaemic Encephalopathy:A Retrospective Cohort Study Tan, Jason; Minutillo, Corrado; McMichael, Judy; Rao, Shripada VERSION 1 - REVIEW Basu, Sudeepta Children's National Health System, Pediatrics, Neonatology Competing interests:i do not have competing interest. 27-Apr-2017 The manuscript reports a retrospective cohort study indicating association of recurrent hypoglycemia with 2 yr ND outcomes in HIE infants. This is in agreement with reports by several studies in past and definitely adds to the existing knowledge base. This study generates hypothesis whether this could be an association, biomarker, prognostic factor and/or be mechanistically causative and needs more prospectively designed studies to specifically explore this question. Following aspects in the manuscript need revision or rewording: 1) Conclusion statement indicating causative association and benefit of treatment to achieve a specific level Targeting a BGL >3.0mmol/L may improve neurodevelopmental outcomes. is speculative and not definitively proven by retrospective cohort analysis. This study finding is hypothesis generating but not proving causation or treatment benefit. Authors can speculate on this in the discussion section, I would recommend revising the conclusion statements in abstract and body of article. 2) Infants have been categorized as hypoglycemia and normoglycemia based on cut-off of 2.6 mmol. This is not true as the "normoglycemia" population will include infants with hyperglycemia. About 40-50% of HIE infants develop hyperglycemia in first 24 hrs (seems to be similar in this cohort from the glucose distribution figure) and this has been associated with poor ND outcomes. 3) It is interesting to note the lack of significant difference in ND outcomes between normoglycemic infants and those with single or even 2 episodes of hypoglycemia. It would be interesting to see if we can look at hypoglycemia and hyperglycemia separately and compare to "normoglycemia defined as 2.6-7 mmol/dl(or any other acceptable cut-off)". If authors can provide more information on this aspect, it would be helpful. 4) We have to be cautious about specific glucose level cut-offs (2.6, 3.0, etc). These proposed normoglycemic range is mostly speculative in HIE population especially since definition of hypo and hyperglycemia in even healthy term infants is under debate for last 4
REVIEWER REVIEW RETURNED GENERAL COMMENTS decades. 5) The authors report the multiple methods of blood glucose measurement in methods section. Can the authors also report whether treatment of hypoglycemia and/or insulin for hyperglycemia was standardized? Is the information on GIRs or insulin use available? If not, then this should be re-iterated in the limitation section as well. 6) The introduction statement only one of these studies had neonates that received therapeutic hypothermia needs modification. Tam et al study had 12% and CoolCap Study had ~50% infants treated with hypothermia. There are some minor typos in manuscript that need attention. Overall, this study supports the growing concern of deranged glucose homeostasis in HIE infants as being at least a biomarker if not contributory to the brain injury in these infants. This manuscript should be considered worthy of publication with some major revisions. Thoresen, Marianne University of Oslo, Division of Physiology, Institute of Basic Medical Sciences Conflict of Interests: I have no conflict of interest With this work 28-Apr-2017 This is a retrospective database review of 122 infants undergoing therapeutic hypothermia. The authors suggest that the associations between hypoglycaemia and neurodevelopmental outcomes are not well explored in infants with hypoxic ischaemic encephalopathy that receive therapeutic hypothermia and that little information is available on the optimal glucose levels in these infants. Infants were cooled if the fulfilled the ICE trial entry criteria. They have 28% with mild HIE grade 1 in this cohort, which do not comply with ICE (moderate or severe HIE, not mild shoudl be included) Infants were followed up at 24 months using Bayley3. Main outcome scoring: Cognitive or Language Bayley <70 and or GMFCS > 3. Adjusted severity using HIE and cord lactate and ph levels (at birth?) were used in analysis. Why not adjust for Apgar scores at 10 minutes? There is too little detail in the data presented. Glucose was measured 4hrly for the first 24h then 6 hourly. Why not report separately glucose in cord blood or the first glucose? You have measured lactate in cord blood and compare this to glucose measured at any time later during the first 24 h. What nutrition had been administered after birth before the glucose was measured? What was the protocol for managing low glucose? N=122, 104 survivors, 87 were followed up. Of these 87 17 were diagnosed to have moderate or severe disability. 17% is very low, this can be due to a very mild cohort or very strict cutoff for disability. <70 for Bayley 3 is not moderate but very severe disability (Jary S et al 2014). Most people publishing outcome after asphyxia using Bayley 3 have 85 as their cutoff. In results they also use Bayley Motor, this is not suggested in the methods. Important data on glucose levels are the value straight after birth, the duration in hours of low glucose is more accurate than the
Response to Reviewer 1 number of measurements. VERSION 1 AUTHOR RESPONSE Conclusion statement indicating causative association and benefit of treatment to achieve a specific level Targeting a BGL >3.0mmol/L may improve neurodevelopmental outcomes. is speculative and not definitively proven by retrospective cohort analysis. This study finding is hypothesis generating but not proving causation or treatment benefit. Authors can speculate on this in the discussion section, I would recommend revising the conclusion statements in abstract and body of article. Thank you for your advice. We agree that the findings of our study can help to generate hypotheses. We have made alteration to the conclusion statements of the abstract and body of the article. Infants have been categorized as hypoglycemia and normoglycemia based on cut-off of 2.6 mmol. This is not true as the "normoglycemia" population will include infants with hyperglycemia. About 40-50% of HIE infants develop hyperglycemia in first 24 hrs (seems to be similar in this cohort from the glucose distribution figure) and this has been associated with poor ND outcomes. We have now replaced the word normoglycaemia with no hypoglycaemia throughout the manuscript. It is interesting to note the lack of significant difference in ND outcomes between normoglycemic infants and those with single or even 2 episodes of hypoglycemia. It would be interesting to see if we can look at hypoglycemia and hyperglycemia separately and compare to "normoglycemia defined as 2.6-7 mmol/dl(or any other acceptable cut-off)". If authors can provide more information on this aspect, it would be helpful. Thank you for your comments and suggestions for analyses. We were able to look at the true normoglycaemic group versus recurrent hypoglycaemic group (excluding any with hyperglycaemia). We found that the socioemotional subscale for BSID-III was significantly lower in the recurrent hypoglycaemia group (mean difference=23; p=0.007). The cognitive (mean difference=16), language (mean difference=12) and motor subscales (mean difference=7) were similar between groups. Unfortunately with these restrictions, there were only four infants with 3 or more episodes of hypoglycaemia and so these results should be interpreted with caution. We did not analyse for other adverse neurodevelopmental outcomes, as the numbers were so small.
We also looked at normoglycaemic infants versus hyperglycaemic infants (excluding hypoglycaemia). We found that the language subscale for BSID-II was significantly lower in the recurrent hyperglycaemia group (mean difference=22, p=0.008). Cognitive (mean difference=23), socioemotional (mean difference=19) and motor subscales (mean difference=15) were similar between groups. There were only 5 infants with 3 or more episodes of hyperglycaemia (without hypoglycaemia) and as such we are cautious about interpretation of these results. Given the small numbers we did not analyse for other adverse neurodevelopmental outcomes. Since these analyses were based on extremely small sample size we have not added these results into the manuscript. We were able to include hyperglycaemia in the multivariate binary logistic regression models. We have to be cautious about specific glucose level cut-offs (2.6, 3.0, etc). These proposed normoglycemic range is mostly speculative in HIE population especially since definition of hypo and hyperglycemia in even healthy term infants is under debate for last 4 decades. We have discussed the limitations of the current definition of hypoglycemia as well as arbitrary cut-off values in the manuscript. The authors report the multiple methods of blood glucose measurement in methods section. Can the authors also report whether treatment of hypoglycemia and/or insulin for hyperglycemia was standardized? Is the information on GIRs or insulin use available? If not, then this should be re-iterated in the limitation section as well. We have included details of the general management in both our units (KEMH and PMH) of hypoglycaemia and hyperglycaemia in the methods section. Both units function as a single clinical care unit and follow the same management protocols. Neonates with HIE were started on a maintenance intravenous infusion of 10% Dextrose at 50 ml/kg/day without enteral feeds. A BGL below 2.6 mmol/l was managed with an intravenous bolus of 2ml/kg of 10% Dextrose and the maintenance infusion increased by 10-20ml/kg/day with a repeat BGL within 60 minutes. If the total fluid rate reaches 80-90mL/kg/day then the concentration is incrementally increased by 5%. Hyperglycaemia was managed by reducing glucose infusion rates; insulin was not administered to any infant in this study. We did not record the glucose infusion rate for all our infants however our protocol for both units is to aim for between 6-8mg/kg/min. The introduction statement only one of these studies had neonates that received therapeutic hypothermia needs modification. Tam et al study had 12% and CoolCap Study had ~50% infants treated with hypothermia.
Thank you for your comment. We have amended our introduction as well as discussion sections to reflect that these two studies included infants that received therapeutic hypothermia. There are some minor typos in manuscript that need attention. We have proofread the manuscript to correct these typographical errors. Response to Reviewer 2 Infants were cooled if the fulfilled the ICE trial entry criteria. They have 28% with mild HIE grade 1 in this cohort, which do not comply with ICE (moderate or severe HIE, not mild should be included) Even though the inclusion criteria for the ICE trial was moderate to severe HIE, the authors of the ICE trial reported that 42 out of 221(19%) study infants had stage 1 HIE. 1 Overall in our study there were 23% infants that had mild HIE. It reflects the difficulties in accurately staging the HIE especially when the infants are born in remote regional centres. Similar observations have been made in recent publications from Canada and the US.2,3 We have updated our manuscript to include this explanation. Infants were followed up at 24 months using Bayley3. Main outcome scoring: Cognitive or Language Bayley <70 and or GMFCS > 3. Adjusted severity using HIE and cord lactate and ph levels (at birth?) were used in analysis. Why not adjust for Apgar scores at 10 minutes? There is too little detail in the data presented. Since majority of the study infants were out-born, there were significant numbers of missing values for ph, lactate and Apgar scores at 10 minute. Hence we could not include them in the multivariate analysis. We have added it as a limitation of our study. We have now included stage of HIE, hyperglycaemia and time of starting therapeutic hypothermia in the updated adjusted analyses. Additionally, we have also included analyses of hypoglycaemia within the first 12 hours of life. We have included these analyses into our results sections and amended the methods and discussion section to reflect these changes. We have also included two tables. Glucose was measured 4hrly for the first 24h then 6 hourly. Why not report separately glucose in cord blood or the first glucose? You have measured lactate in cord blood and compare this to glucose measured at any time late during the first 24 hr.
We agree that it would be useful to examine the relationship between early hypoglycaemia and adverse neurodevelopmental outcomes. There is data to suggest early hypoglycaemia is related to poorer long-term neurodevelopmental outcomes. We performed analysis using the first measured glucose. We found that on univariate and multivariate analysis, first measured BGL was not associated with poorer neurodevelopmental outcomes. We analysed the association between early hypoglycaemia (first 12 hours of life) similar to Basu et al. (reference 16 in manuscript). We found that there was an association between early recurrent hypoglycaemia and adverse neurodevelopmental outcomes. We have updated the methods, results and discussion to reflect this analysis. What nutrition had been administered after birth before the glucose was measured? What was the protocol for managing low glucose? Infants received intravenous dextrose infusion initially. After the first 24 hours, infants were commonly commenced on total parenteral nutrition. No infants received enteral feeds whilst receiving therapeutic hypothermia. Insulin was not used in any infants in this study. This has been updated in the methods section. N=122, 104 survivors, 87 were followed up. Of these 87 17 were diagnosed to have moderate or severe disability. 17% is very low, this can be due to a very mild cohort or very strict cutoff for disability. <70 for Bayley 3 is not moderate but very severe disability (Jary S et al 2014). Most people publishing outcome after asphyxia using Bayley 3 have 85 as their cutoff. We do acknowledge that our rate of disability is low. We agree that this may be due to 1) inclusion of infants with mild HIE in analysis, and 2) reporting on severe disability. We did not include infants with a BSID score between 70-85 (between 1-2 standard deviations from the mean) because we wanted to focus on severe disability. We understand that infants that fall in this range may have clinically relevant delays. There are studies that have identified this range as still being within normal limits with around 13.5% of the general population falling within this cutoff. We have updated the manuscript to highlight that infants with mild HIE were included and that we were focused on severe disability. In results they also use Bayley Motor, this is not suggested in the methods. Thank you for you suggestion, this has been updated in the methods section. Important data on glucose levels are the value straight after birth, the duration in hours of low glucose is more accurate than the number of measurements. The recent study by Basu et al (secondary analysis of Coolcap trial) used blood glucose levels within the first 12 hours of life for analysis. We have used similar approach. Since we do not use continuous glucose monitors, we do not have information on the duration of hours with low glucose levels.
References 1. Jacobs SE, Morley CJ, Inder TE, Stewart MJ, Smith KR, McNamara PJ, et al. Wholebody hypothermia for term and near-term newborns with hypoxic-ischemic encephalopathy: a randomized controlled trial. Arch Pediatr Adolesc Med. 2011 Aug;165(8):692-700 2. Gagne-Loranger M, Sheppard M, Ali N, Saint-Martin C, Wintermark P. Newborns referred for therapeutic hypothermia: association between initial degree of encephalopathy and severity of brain injury (what about the newborns with mild encephalopathy on admission?) Am. J. Perinatol., 33 (2016), pp. 195-202 3. Massaro AN, Murthy K, Zaniletti I, Cook N, DiGeronimo R, Dizon M et al. Short-term outcomes after perinatal hypoxic ischemic encephalopathy: a report from the Children's Hospitals Neonatal Consortium HIE focus group. Journal of Perinatology. 2015 Apr;35(4):290-6 REVIEWER REVIEW RETURNED VERSION 2 REVIEW Bagkeris, Emmanouil University College London, UK Competing interests: No competing interest 31-Jul-2017 GENERAL COMMENTS 1. Refer to multiple regression as multivariable and not multivariate. Multivariate analysis is a type of analysis with more than one outcome variable. 2. Add as a limitation of the study its case-control design. A cross-sectional design does not allow for prospective conclusions. 3. While table 1 suggests that Mann-Whitney U test was performed for non-normal continuous data, the main text acknowledges Student t-test only. Please revise the statistical analysis section accordingly. 4. For each variable in table 1, indicate whether mean (SD), median (IQR), n (%) is reported as you did for the Apgar score, follow the same principle for all tables as appropriate. Response to reviewer 1 VERSION 2 AUTHOR RESPONSE 1. Refer to multiple regression as multivariable and not multivariate. Multivariate analysis is a type of analysis with more than one outcome variable. Thank you for this comment. We have renamed multivariate analysis as multivariable analysis. 2. Add as a limitation of the study its case-control design. A cross-sectional design does not allow for prospective conclusions. This was a retrospective cohort study and we have changed the title to reflect this as per editor s suggestion. We have also added it as a limitation of our study in the discussion section.
3. While table 1 suggests that Mann-Whitney U test was performed for non-normal continuous data, the main text acknowledges Student t-test only. Please revise the statistical analysis section accordingly. We have updated the statistical analysis section in our methodology section. 4. For each variable in table 1, indicate whether mean (SD), median (IQR), n (%) is reported as you did for the Apgar score, follow the same principle for all tables as appropriate. We have updated all our tables to include the appropriate measures used. We have also updated table 3 and 4 to include 95%CI. 5. Add "a retrospective cohort study" to the title We have updated our title as suggested. In addition to the above changes, we have also rounded the variables that are presented as percentages to whole numbers and removed the single decimal point in our tables given our small sample size. bmjpo: first published as 10.1136/bmjpo-2017-000175 on 18 September 2017. Downloaded from http://bmjpaedsopen.bmj.com/ on 24 April 2019 by guest. Protected by copyright.