October 1st, 2007
Editor, BMC-series Journals
BioMed Central
Middlesex House
34-42 Cleveland Street
London W1T 4LB
Re.: Responses to comments made by reviewer #2 to manuscript submission # MS:
2073626391423293
Dear Sir or Madam,
We recently received a rejection letter from BMC Infectious Diseases on a manuscript
submitted to you as # MS: 2073626391423293
While the comments of reviewer #1 were appropriate and could easily be addressed in
revisions, the manuscript was rejected based on the decision of reviewer #2. Upon
reading the comments made by reviewer #2, I began to question whether they had the
expertise to make such a decision. I found several instances where information was
claimed to be missing that was indeed present in the manuscript. Comments made
suggested that they were unfamiliar with the type of study population (anti retroviral
treated HIV infected patients) and methods (immunological methods) used in the
manuscript. Overall the review was confusing and the nature of their specific objections
unclear beyond what would be related to not understanding the field. Since reviewer # 1
appeared to be comfortable with the study design, study population and methods used, I
conclude that the problem is not solely in how the manuscript is written. At this juncture I
feel strongly that a decision to reject this manuscript based on review #2 is unjustified.
I therefore request that the decision to reject be reconsidered. To support this request I
have addressed point by point the comments made by reviewer #2 below.
Comment 1
General:
The authors are assuming that the readers are very well familiar with their
terminology and notation when reporting the results of their study as well as
familiar with the methods implemented by the authors and the population of
study.
Response 1:
The authors contend that the terminology, methods and study population are standard for
the field and should be understandable to an audience that follows the literature on the
subject matter.
Comment 2:
Abstract:
-What is PBMCs in the abstract, it is not previously referenced.
-what is pre-T1 definition by the authors in this abstract?
-What is IQR and r in this abstract?
-What is the author’s definition of “drug holiday”?
Response 2:
These can be defined in a revised manuscript.
Comment 3:
Background:
Please provide the reference for the SMART trial.
Response 3:
This is already provided in the manuscript as reference #5.
Comment 4:
Methods:
Authors used statistical methods for the description and interpretation of their
findings. Authors used statistical inference techniques to report their findings.
The use of statistical methods in inferential process is based on the assumption
that there is a sample selected from a population of interest. Statistical methods
are also known for samples that were randomly selected from the population of
interest with some known probability to proceed with the inferential process.
The authors lacked to describe the population of interest and how those 13
patients were selected for their analysis. From which country are these patients?
Why these patients represent a particular population? Among how many these
patients are only available? What was the inclusion and exclusion criteria for
these individuals? Why was a retrospective study chosen and from where was
that retrospective information collected? Which format, form, or questionnaire
was used to collect systematical information on these patients? Why there is not
description of the precious drugs used by these patients and why was not
important for the authors? It is unclear how other readers will be able to compare
their patients with the description of the patients provided here, i.e. other
countries? Other hospitals?
Response 5:
1) The patients are Canadian and followed at the Immune Deficiency Treatment Centre
of the McGill University Health Centre in Montreal, Quebec, Canada.
2) The 13 patients studied are all the individuals found at this site who met the inclusion
criteria described in the Materials and Methods section and for whom frozen cell samples
were available on which to perform immunological testing.
3) The inclusion criteria are provided in the Materials and Methods section.
4) A retrospective study was chosen because the material existed and the expertise was in
place to address the following question: Do pre-treatment interruption parameters
determine the extent of viral load rebound and CD4 T cell decline observed in a group of
subjects who were unable to control their viral load before they stopped a therapy
regimen, and who stopped therapy for long enough to experience a viral rebound. These
clinical parameters (viral load and CD4 count change) are surrogate markers of outcome
after treatment interruption.
5) The patients were selected using a database that is maintained in our centre of clinical
testing results collected at each clinic visit.
6) As this was a retrospective study on an HIV clinic based population the patients had
been on multiple treatment regimens. This is typical and representative of a therapy
experienced HIV infected population in the real world. The critical issue is that they were
all experienced with regard to antiretroviral therapy, had been on many regimens and
were experiencing drug failure as defined by detectable viral loads before they stopped
therapy. Listing the regimens could be done but would differ from patients to patient and
would confuse rather than clarify.
7) These patients should be representative of many HIV infected individuals seen by
treating physicians. These are individuals who have been treated for years with multiple
regimens, are failing treatment and for one reason or another wish to stop taking
antiretroviral therapy for a while. This is a situation encountered by physicians treating
HIV infected individuals. Some patients who stop therapy will exhibit a severe decline in
CD4 counts and if an educated guess could be made that this will occur could be
discouraged from stopping therapy. Other patients will remain stable and in their case
stopping therapy will have less negative impact of their disease progression. The
manuscript addresses whether certain tests can aid in making this decision.
Comment 6:
-What was the purpose of dichotomizes the count of CD4+T cell decline?
Response 6;
The purpose is to view the results in another manner by separating the subjects into the
half that did worse versus better based on median value of the outcome measures (CD4
count change) after therapy and compare the value of markers (%CD4+CD57+ cells and
proliferative capacity to Gag p55 or cytomegalovirus) before therapy interruption in these
2 groups.
Comment 7:
-Which subpopulations were studies? It is unclear in this current version of the
manuscript.
Response 7:
The study population remains constant throughout, i.e. the same 13 individuals.
Comment 8:
-What are the responses evaluated in the figures presented in this paper? They
are described differently than in the text or they are different variables
presented?
Response 8:
The response evaluated in figure 1A are flow cytometry plots of CFSE assay showing
proliferation to the stimuli/controls listed to the left by cells in the CD4 and CD8
compartments for a representative individual who exhibited a good versus poor outcome
following treatment interruption. Figure 1b compiles data for CMV-specific and HIVGag p55-specific proliferation for all individuals tested for CD4+ T cells in the left panels
and CD8+ T cells in the right panels.
Figure 2 shows representative flow cytometry plots of CD4+CD57+ cells for a patient
with low baseline percent CD4+CD57+ T-cells and low CD4 decline during TI (Patient
#2) and a patient with high baseline percent CD4+CD57+ T-cells and high CD4 decline
during TI (Patient #1). Figure 2b compiles the % CD4+CD57+ values for all study
participants versus CD4 changes from baseline to the treatment interruption time point.
This information is in the manuscript and legends. It is also a standard way to show
results where an example of the research results for an individual are shown in one panel
to give a flavor for the quality of the results generated; the results for the study population
is then presented in an additional panel with the statistical analysis.
Comment 9:
-What are the values reported in each contour plot for figures 1, A?
-What was the purpose of the individual contour plots? Why was not described in
the methods of this manuscript?
Response 9:
The value in the contour plots for Figure 1A are the %CFSElow positive cells; for figure
2A the percent of CD4+CD57+ cells. As cells labeled with CFSE proliferate and divide
the CFSE fluorescent marker intensity is halved with each cell division and therefore
becomes dimmer and is detected in a lower fluorescence intensity channel. If this was not
made explicit in the text it is because it is a standard technique used in the field.
Comment 10:
-Why only 13 patients?
Response 10:
See above
Comment 11:
-What was the timeframe for collecting these patients? What are the biases
associated with this?
Response 11:
The timeframe for collecting the samples is 5/20/1997 to 7/31/2001. All eligible
individuals for whom a frozen cell sample was available were tested. I am not certain I
can come up with potential biases in patient selection.
Comment 12:
-Statistical significance is not established by p-values, 0.05 but due to sample
size selection from the population of interest using a type I error level of 0.05.
Authors should correct their statement in the methods section.
Response 12:
I am not sure I understand this comment. To try to respond…. If we had seen no
statistically significant differences it would be valuable to know that our sample size was
large enough to not have missed what was there. Since here we report differences that
are significant at the level of p<0.05 does that not imply that the sample size was large
enough to detect the differences detected?
Comment 13:
-How many observations per patient were used for the analyses presented?
Response 13:
1 observation per patient.
Comment 14:
-How authors handled missing data, specially because this is a retrospective
information collected?
Response 14:
There was no missing data. If samples were not available for time points that met the
inclusion criteria, then the time points were not included in the study. However, due to
the limited amount of cells available for each time point tested, not all immune
parameters could be tested for every time point.
Comment 15:
- Which statistical method was used for testing trends as described in the results
section?
Response 15:
The same statisitical methods. Trends were reported for observations of interest that did
not achieve a p-value of 0.05
Comment 16:
-How was the sample size calculated and power used for selecting the patients
for this analysis?
Response 16:
No sample size was calculated, essentially we studied all subjects meeting the criteria
available for which frozen cells were available.
Comment 17:
Results:
-Unclear what is IQR for these authors. If IQR represents interquartile range, it is
just a number and not two numbers as the authors reported. Are they presenting
minimum or maximum, quartiles? This raised major concerns in the statistical
aspects of this paper.
Response 17:
IQR is interquartile range and as the term implies the 2 numbers represent the 25 and 75th
percentile values surrounding the median. The way IQR is presented in the manuscript
is consistent and accepted with current published journal articles.
Comment 18:
-Authors should report gender in percentages instead of absolute values
Response 18:
We could do this but I am not certain it is an improvement over reporting number given
the small population size.
Comment 19:
-Unclear why, how, and where the Mann-whitney test is reported in figure 1.
Response 19:
Figure 1 A shows a representative plot of 1 patient each in the good and poor CD4 group.
No Mann-Whitney test was reported for the data presented in Figure 1, rather a
Spearmans correlation test was used to assess the strength of the association between
CD4 count change and proliferation to CMV and HIV Gag p55. However, a MannWhitney test was used to compare the dichotomized results for proliferation to Gag p55
and percent CD4+CD57+ cells comparing the patients groups with good vs. poor CD4
change after treatment interruption. These results were reported in the text in the Results
section but refer to a different way of looking at the results than the correlation analyses
presented in Figure 1B.
Comment 20:
-Unclear paragraph on page 8, what are the comparison groups and the test
presented for comparing what with what?
Response 20:
We dichotomized the 13 subjects by the median CD4+ count loss into the good vs. poor
CD4 group and then we proceed to compare the various immune parameters between
these two groups. This is stated in the text of the Result section.
Comment 21:
-What is the definition of the authors for immune predictors?
Response 21:
We would define an immune predictor as an immunological parameter that can be tested
before stopping therapy that would have a predictive value in determining whether the
patient will experience poor outcome or not after stopping therapy. Such a test could be
further validated with a goal of eventually being used in clinical decision making on
whether or not to stop therapy.
Comment 22:
Discussion:
-Authors should minimize their discussion substantially for example authors
claimed that their results presented “SHOWED” clear evidence for their findings,
however there is not a clear population study or description that they have
enough sample size for their conclusions or hypothesis of interest.
Response 22:
See above and we can do this
Comment 23:
-Author lack to motivate readers to compare their findings with other because it is
impossible to know from where these patients come from to use their results,
especially because authors indicate that this is key for those using “drug
holidays”. Implications and links to real clinic is lacking in this version of the
manuscript.
Response 23:
This manuscript presents results that are a first step to identifying markers that may
predict outcome following treatment interruption. We are not suggesting these markers
should be used in the clinic but rather that they could be explored in larger studies that
would be powered to validate their use in the clinic.
Comment 24:
-Authors lack to evaluate all the potential biases found in their sample in terms of
the epidemiological components.
Response 24:
This comment is vague. I don’t know what epidemiological components that the
reviewer wants us to address.
Comment 25:
-What are the limitations of this research and your findings?
Response 25:
Small sample size.
Comment 26;
Figure 1
-Please review titles.
Response 26:
This can be been done
Comment 27:
-Are the correlations across multiple measures in time for these patients?
Response 27:
No, the correlations are for one measure in time within 12 months of treatment
interruption.
Comment 28:
-What does the rectangle measures in figure 1 section A?
Response 28
See response to comment 9.
Comment 29:
-Why are graphs presented for selected patients instead of all patients? What is
the message they want to communicate to readers? Why are these contour
graphs important in measuring the correlation?
Response 29:
See response 8…
This is a standard way to show results. Figure 1A shows representative raw data and
figure 1B shows all the data for measures of interest versus change in CD4 which is
defined in the text as the difference between CD4 count before and after treatment
interruption.
Comment 30:
-Why there is a different notation in this graphs and the text? What is CD4 count
change??? In time???
Response 30:
There is no difference in the notation between the graphs and the text. CD4 count change
is the change in CD4 count between the pre-TI baseline time point to the lowest CD4
count during treatment interruption (TI).
Comment 31:
-It seems authors think that a not statistical significant p-value for a spearman
correlation coefficient indicates lack of correlation. However, their graphs
indicated that they are measuring is linearity of their correlation which is usually
the case with Pearson correlation. Authors need to review what are their testing
when reporting each p-value in their manuscript and figure.
Response 31:
The test performed was Spearman correlation. The line was added to show linearity.
This is commonly done as seen in other publications. However, we could remove the
line to avoid confusion with Pearson correlation as the reviewer suggested.
Comment 32:
-Graphs in section B do not include 13 patients, why? What the line represents in
these graphs? How the assumptions for the line provided here were evaluated?
-Why the correlation coefficients are different between text and figure?
-When were these variables measured?
Response 32:
1. 2 graphs in section B do not include 13 patients because we did not have
sufficient cells to test that parameter. The line is to show the linearity. See
Response 30 for further clarification.
2. The correlation coefficients are the same between the text and Figure 1. This has
been verified. The only discrepancy is between the p-value in Figure 2 where it is
0.0095 in the figure but rounded up to 0.001 in the text. This could easily be
corrected.
3. These variables were measured with the cells from the one pre-TI baseline time
point for each patient.
Comment 33:
-What are the socio-demographic characteristics of these patients?
Response 33:
This not relevant for our study and does not influence the conclusions we draw.