Good Metabolic Control in Children with Type 1 Diabetes Mellitus: Does Glycated Hemoglobin Correlate with Interstitial Glucose Monitoring Using FreeStyle Libre?

Round 1

Reviewer 1 Report

Thank you for the opportunity to review this manuscript examination the correlation between CGM glucose metrics versus HbA1c 1 year post sensor use in children and adolescents with type 1 diabetes. The authors included a moderately large sample of 191 patients, stratified by HbA1c categories and investigated relationships with time in range (TIR metrics).

I have a number of major concerns for the study, results and methods

i) The baseline characteristics of the patients pre CGM insertion (HbA1c, SMBG frequency etc should be provided). 

ii) Table 1 The percentage sensor use, frequency of isCGM scanning at 1 year was not reported. This has ought to have a greater impact on HbA1c than frequency of SMBG? I am not entirely sure about the relevance of reduction of SMBG frequency during CGM on HbA1c.

iii) Critically, time above range was not reported (level 1 and level 2 hyperglycemia) in Table 1 or in the results or text. Is the lower TIR than expected due to greater TBR or TAR?

iv) Consider treating HbA1c and TIR as continuous variables for correlation analysis

v) Limitations of sensor accuracy and duration of CGM recording should be discussed. Could HbA1c also have been subject to bias due to anemia, haemoglobinapthies ?

Author Response

REVIEW TIR

Reviewer 1

Thank you for the opportunity to review this manuscript examination the correlation between CGM glucose metrics versus HbA1c 1 year post sensor use in children and adolescents with type 1 diabetes. The authors included a moderately large sample of 191 patients, stratified by HbA1c categories and investigated relationships with time in range (TIR metrics).

I have a number of major concerns for the study, results and methods

1. The baseline characteristics of the patients pre CGM insertion (HbA1c, SMBG frequency etc should be provided). 

We have another article pending of publishing in with we speak about the evolution of  these items (basal, 3 months, 6 months and 1 year) . In this article the change of  hbA1c wasn´t the aim of study, but it was used as one variable for classifying the studied subjects in 5 categories to correlate with parameters of CGM

According to SMBG frequency, it is reflected at table 3. (No. capillary blood glucose/day baseline and 1 year after)

1. Table 1 The percentage sensor use, frequency of isCGM scanning at 1 year was not reported. This has ought to have a greater impact on HbA1c than frequency of SMBG? I am not entirely sure about the relevance of reduction of SMBG frequency during CGM on HbA1c.

We have added in inclusion criteria that only those patients who had more than 80% use of the sensor were included to avoid that the percentage of sensor use was considered a possible confounding factor in the interpretation of the data.

It has been added in table 3 the data of the number of scans that do not show significant differences between the different categories of glycosylated hemoglobin.

• Critically, time above range was not reported (level 1 and level 2 hyperglycemia) in Table 1 or in the results or text. Is the lower TIR than expected due to greater TBR or TAR?

We added the explanation in the article.

The time above the range has not been reported. Table 3 reports the TBR, which allows us to observe that the lower TIR is not due to the higher TBR, since in most categories it is close to the recommended one (less than 5%), so, it is the highest time above range which would explain the shortest time in range that we observe in the data.

1. Consider treating HbA1c and TIR as continuous variables for correlation analysis

We have done it and there is a positive correlation as has been documented in other articles. But the objective that we wanted to share in this article is that glycosylated hemoglobin is a variable that overestimates adequate control and patients who are considered well controlled with glycosylated hemoglobin present time values ​​in range insufficient with respect to those proposed by the consensus.

1. Limitations of sensor accuracy and duration of CGM recording should be discussed. Could HbA1c also have been subject to bias due to anemia, haemoglobinapthies ?

The MARD of Freestyle 1 could be a  limitation for the study. It would be usefull to reproduce this study with FreeStyle Libre 2 which has less MARD and check the results.

We added that bias such as anemia or haemoglobinopathies are exclusion criteria in the study.

Reviewer 2 Report

The authors propose a compared analysis of HbA1c and different parameters calculated using common and accessible device for CGM, here FreeStyle Libre, in a pediatric population (4-18y) before and 1-year after they begin to use FSL.

The article is easy and nice to read, in a correct English. 

The data are not very original, but still interesting, especially regarding a pediatric population. My greatest concern is the lack of specific question announced and addressed by the authors. At the end of a quite long introduction, the reader understands that he will hear about HbA1c and CGM parameters, but does not know what is/are the precise issues addressed. Then, the authors presents a series of relevant results, but they deserve a better red line.

Major comments

• Introduction: I strongly recommend to shorten the introduction and to add, at the end, what is (are) the precise question addressed by this paper, and the objectives 
• Method 1: the article lacks a study flowchart (number of patients screened, attrition...). Do we must understand that all included patients performed the 1-year visit?
• Method 2 (statistical analysis): it's strange to read about Student and Mann-Whitney, and, the next page, to find a table with ANOVA or Kruskal-Wallis test (table 3) => please announce all the statistical tests in the Method and repeat it in the legend of the figure (if you keep it - see below)
• Table 3: the table needs improvement.
  • p-values are not very relevant, as you compare not-randomized population which are intrinsically different regarding glucose homeostasis => with enough statistical power, one would except to see significant p-values at each line, so they are not informative. I would advise not to present these "p"
  • in the discussion, you propose very quickly a threshold >70% for the TIR -> I advise to give this binary information in the table 3 and not only the mean of the TIR
  • the numbers precision is too important, please simplify (number of daily capillary blood test : 7.0 and not 7.020, which is both clinically and statistically irrelevant here)
  • TIR and TBR are first expressed as % or hours, and then you calculated a mean (SD)=> please specify the mean of what.
  • HbA1c threshold: please express the borders correctly (< or ≤, so we can know where is a patient with HbA1C = 7%, for example)
• Discussion (lines 222-225) : "No significant differences were found between the greater number of capillary glycemia tests per day and the different glycated hemoglobin groups, which indicates that more frequent capillary glycemia testing is not associated with better metabolic control measured by glycated hemoglobin" : this affirmation is wrong. Your study presents a low statistical power (n = 191) which, additionally, you divided between 5 groups. So, if you fail to show a difference (p-value < 0.05 speaking), you just cannot conclude.

Minor comments

• General: Maybe global consistency could be improved following STROBE guidelines & checklist
• Introduction (line 42): the reference (ADA standards) is too general. Please quote the original article(s)
• Table 1 & 2: please give a proper legend, particularly for abbreviations (NICE, ISPAD, ADA, T1D..)
• Table 2: I had difficulties to read it, especially with the sign "<" added in different cells. I don't now if it's an editing issue but this must be corrected, for example using two cells when there are 2 informations (e.g <25% and < 6h)
• Method 1: please give the inclusion time (dates of first and last patients screened)
• Method 2 (lines 118-121): here, some exclusion criteria are a "mirror" of the inclusion criteria. Please remove redundant exclusion criteria
• Methods/Discussion: how did the authors choose to proposer 5 groups, and this specific thresholds for HbA1c (6.5/7/7.5/8%)? Maybe I missed it but this must appear somewhere in the method or discussion
• "qualitative variable": please use the more common "categorical variable"
• Figure 1 is very basic. If you use function hist() in R, I recommend at least to use argument breaks (=20 or 30, depending on display) or, for a best rendering, the ggplot2 package

Author Response

Reviewer 2

Comments and Suggestions for Authors

The authors propose a compared analysis of HbA1c and different parameters calculated using common and accessible device for CGM, here FreeStyle Libre, in a pediatric population (4-18y) before and 1-year after they begin to use FSL.

The article is easy and nice to read, in a correct English. 

The data are not very original, but still interesting, especially regarding a pediatric population. My greatest concern is the lack of specific question announced and addressed by the authors. At the end of a quite long introduction, the reader understands that he will hear about HbA1c and CGM parameters, but does not know what is/are the precise issues addressed. Then, the authors presents a series of relevant results, but they deserve a better red line.

Major comments

• Introduction: I strongly recommend to shorten the introduction and to add, at the end, what is (are) the precise question addressed by this paper, and the objectives 

Introduction has been shortened and the objectives has been added.

• Method 1: the article lacks a study flowchart (number of patients screened, attrition...). Do we must understand that all included patients performed the 1-year visit?

We added that we have not got any attrition along the study.

• Method 2 (statistical analysis): it's strange to read about Student and Mann-Whitney, and, the next page, to find a table with ANOVA or Kruskal-Wallis test (table 3) => please announce all the statistical tests in the Method and repeat it in the legend of the figure (if you keep it - see below)

We added it to Methods and table 3.

• Table 3: the table needs improvement.

We have tried to improve it.

• p-values are not very relevant, as you compare not-randomized population which are intrinsically different regarding glucose homeostasis => with enough statistical power, one would except to see significant p-values at each line, so they are not informative. I would advise not to present these "p"

We have removed no significtive p-values

• in the discussion, you propose very quickly a threshold >70% for the TIR -> I advise to give this binary information in the table 3 and not only the mean of the TIR

It is a very interesting suggestion. We have added this data in results.

• the numbers precision is too important, please simplify (number of daily capillary blood test : 7.0 and not 7.020, which is both clinically and statistically irrelevant here)

We have simplified daily capillary blood test.

• TIR and TBR are first expressed as % or hours, and then you calculated a mean (SD)=> please specify the mean of what. We have specified that it was %
• HbA1c threshold: please express the borders correctly (< or ≤, so we can know where is a patient with HbA1C = 7%, for example) We have changed it.
• Discussion (lines 222-225) : "No significant differences were found between the greater number of capillary glycemia tests per day and the different glycated hemoglobin groups, which indicates that more frequent capillary glycemia testing is not associated with better metabolic control measured by glycated hemoglobin" : this affirmation is wrong. Your study presents a low statistical power (n = 191) which, additionally, you divided between 5 groups. So, if you fail to show a difference (p-value < 0.05 speaking), you just cannot conclude.

The statement is withdrawn after consideration by the proofreader.

Minor comments

• General: Maybe global consistency could be improved following STROBE guidelines & checklist

STROBE Statement—Checklist of items that should be included in reports

• Introduction (line 42): the reference (ADA standards) is too general. Please quote the original article(s)

We have added reference to this line

• Table 1 & 2: please give a proper legend, particularly for abbreviations (NICE, ISPAD, ADA, T1D..)

We have added abbreviations.

• Table 2: I had difficulties to read it, especially with the sign "<" added in different cells. I don't now if it's an editing issue but this must be corrected, for example using two cells when there are 2 informations (e.g <25% and < 6h)

We divided information in several cells.

• Method 1: please give the inclusion time (dates of first and last patients screened)
• We have added it to methods.
•  
• Method 2 (lines 118-121): here, some exclusion criteria are a "mirror" of the inclusion criteria. Please remove redundant exclusion criteria

We removed redundant exclusión criteria

• Methods/Discussion: how did the authors choose to proposer 5 groups, and this specific thresholds for HbA1c (6.5/7/7.5/8%)? Maybe I missed it but this must appear somewhere in the method or discussion

We have chosen those groups according to cut-off points proposed by the different scientific societies of good control.

• "qualitative variable": please use the more common "categorical variable"

We have changed it.

• Figure 1 is very basic. If you use function hist() in R, I recommend at least to use argument breaks (=20 or 30, depending on display) or, for a best rendering, the ggplot2 package

We would remove the figure, saying that it is redundant with the information included in the text