Cardiotoxicity in Adult Patients with Relapsed or Refractory Acute Myeloid Leukemia

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper is well written and very interesting.

It would be appropiate to better specify the definition of refractoriness in relation to patients treated with intensive regimens ( it is correct after 2 courses of intensive induction treatment) and the definition you applied for patients treated with less intensive regimens ( such as hypomethylating agents).

How do you explain a median OS of 8,8 months in patients without cardiac events, substantially similar to median OS of 7,6 months of patients with cardiac events of grade 3-4?

Author Response

We thank the reviewer for his comments. We have addressed the reviewer´s remarks on a point by point basis. Changes into the manuscript are highlighted. We hope that after implementing the suggested corrections/changes the manuscript is now suitable for publication
The paper is well written and very interesting.
It would be appropiate to better specify the definition of refractoriness in relation to patients treated with intensive regimens (it is correct after 2 courses of intensive induction treatment) and the definition you applied for patients treated with less intensive regimens ( such as hypomethylating agents).
 Thanks for your comment. The definition for intensive and non-intensive regimens refractoriness is shown in the following paragraph of the M&M section, page 3, line 127: “Refractory AML was defined as a failure to achieve CR/CRi after 1 or 2 courses of intensive-induction chemotherapy treatment, excluding patients with death in aplasia or death due to indeterminate cause. Patients treated with non-intensive chemotherapy approaches were considered refractory when they showed progressive disease, no clinical benefit, or a change in the treatment line. Relapse was defined as AML subjects who achieved a CR/CRi with prior line treatment and had a hematologic relapse.”
How do you explain a median OS of 8,8 months in patients without cardiac events, substantially similar to median OS of 7,6 months of patients with cardiac events of grade 3 4?
 We acknowledge that mOS for the grade 3-4 cardiac event cohort was quite like the non-cardiac event cohort. This is probably due to time bias, as at least some patients had to survive more to develop a grade 3-4 event. This is also evident for the grade 1-2 cardiac event cohort, with 21 months mOS (probably due to this time bias).

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors propose a study on cardiac events post 2L and 3L therapy post R/R AML. The manuscript contain many statistical issues that have to be  corrected by the authors to make the analysis statistically correct and results interpretable.

Major comments:

A) The split of non-fatal and fatal cardiac events is not statistical correct at least for the "non-fatal" cardiac event. Indeed, patients with death post cardiac toxicity are exclude from that group which corresponds to an informative censoring. To do a correct statistical analysis, the authors have to first provide the cumulative incidence of all cardiac event (without any distinction according to the death post cardiac event) and second, provide at the authors did, the cumulative incidence of death related cardiac event (based only on the cause death).

B) The authors provide either the "crude" percentage and cumulative incidence of cardiac events. The point to provide cumulative incidences is the exact reason as cardiac event is sensitive to censoring and competing event, and then the crude percentage is by definition not correct. The authors have to remove the crude percentage of cardiac event.

C) For univariate analysis in competing event setting there is no need and reason to use a Fine and Gray model instead of a simple Gray test. Particularly when the authors use Cox-cause specific model for multivariable analysis.

D) The authors mentioned KM curves that is correct for OS and EFS but not for outcome in the competing risk setting

E) 2L and 3L groups are different but overlapping population. For clarity it is crucial to present in two different tables

F) In the table 5, the authors cannot show the detail of "No cardiac event" versus each grade of cardiac events. Indeed, the cardiac event is a time to event and this is the reason why cumulative incidence have been used. If the authors wants to describe the characteristics of the patients with cadiac event, accordingly, they have to exclude the "No cardiac event group" and calculate every outcomes from the date of cardiac event. Also to delay between 2L and cardiac event have to be included, and response to 2L only if occurring before the cardiac event. The EFS is complicated to calculate in such setting as the events can occur before or after cardiac event.

G) The first column of table 5 must be a separated table given the outcome of the cohort 2L (and a same table for 3L). OS and EFS have to be given at 6m and 5y as the cumulative incidence of cardiac event are given. Indeed, this is crucial to know how important is the death as competing event.

H) The univariable and multivariable analysis have to be redone according to the comments for interpretability

I) The curve in figure 2 is false for several reason: first it is written time from diagnosis. The starting point of the 2L analysis is the date of start of 2L treatment and not diagnosis. If that refers to date of diagnosis of cardiac event, what is the date of "no cardiac event"? Second, cardiac event is a time to event outcome and then, the there is an immortal time bias in favor of having a cardiac event.

Minor comments:

1) By definition a patient receiving a 3L therapy got a step for a 2L. How many patients with 3L are included in the 2L group?

2) The authors mention to report "means or medians" and did "Mann-Whitney and Student t tests". The authors have to either explain when they used one or the other, or simply use median and Mann-Whitney tests that is valid in any situation.

3) The cumulative incidence function has not been developed by Fine and Gray. Please remove the reference or provide the correct reference.

4) Line 178: the word correlation is not correct and should be replaced by association

5) The curves have to not be shown more than the median follow-up

6) For EMD and biological values, the authors should specify if it is at time of diagnosis or R/R.

7) The reverse KM median follow-up has to be given

Typos:

Line 66: A space is missing between "drugs" and ref 5.

Author Response

We thank the reviewer for his comments. We have addressed the reviewer´s remarks on a point
by point basis. Changes into the manuscript are highlighted.
We hope that after implementing the suggested corrections/changes the manuscript is now
suitable for publication
Comments and Suggestions for Authors
The authors propose a study on cardiac events post 2L and 3L therapy post R/R AML. The
manuscript contain many statistical issu es that have to be corrected by the authors to make the
analysis statistically correct and results interpretable.
Major comments:
A)
The split of non fatal and fatal cardiac events is not statistical correct at least for the "non
fatal" cardiac event. Ind eed, patients with death post cardiac toxicity are exclude from that group
which corresponds to an informative censoring. To do a correct statistical analysis, the authors
have to first provide the cumulative incidence of all cardiac event (without any dis tinction
according to the death post cardiac event) and second, provide at the authors did, the
cumulative incidence of death related cardiac event (based only on the cause death).

While the reviewer`s comments are the most correct, we have performed all these analyses
following a pre defined statistical analysis plan, which was performed and revised several times.
As these analyses were also applied to the previous manuscript of Boluda et al (Boluda, et
al, Incidence and Risk Factors for Developme nt of Cardiac Toxicity in Adult Patients with Newly
Diagnosed Acute Myeloid Leukemia. Cancers (Basel). 2023, 15 (8).) 8).), and to get comparability
between this front line study and our R/R study here, we would prefer to keep as it is .
B) The authors provide
either the "crude" percentage and cumulative incidence of cardiac
events. The point to provide cumulative incidences is the exact reason as cardiac event is
sensitive to censoring and competing event, and then the crude percentage is by definition not
cor rect. The authors have to remove the crude percentage of cardiac event.

In accordance with the reviewer’s comment, we modifi ed it in the manuscript. We
acknowledge that providing the crude incidence could be repetitive. So, we provide now only N
(%) in these columns.
C) For univariate analysis in competing event setting there is no need and reason to use a Fine
and Gray model instead of a simple Gray test. Particularly when the authors use Cox cause
specific model for mult ivariable analysis.

We will correct as this is a typo in the statistical section. For univariate analyses we used the
gray test. The Fine and grey model was used only for multivariate analyses . See statistical
section, page 4 , line 175 , as For the univariate risk factor analyses (using the Gray
test), we considered a significant correlation for p values < 0.05.
D) The authors mentioned KM curves that is correct for OS and EFS but not for outcome in the
competing risk setting

As per reviewer’ s suggestion we have corrected the statistical section stating that the KM
curves were used for OS ad EFS . See statistical section, page 4 , line 173 , as Time to
event analyses were calculated in months (from the first treatment after the index date) and OS
and EFS were summarized using Kaplan Meier (KM) curves.
E) 2L and 3L groups are different but overlapping population. For clarity it is crucial to
E) 2L and 3L groups are different but overlapping population. For clarity it is crucial to present in present in two different tablestwo different tables..

 We acknowledge that they are in part overlapping, however, we prefer to We acknowledge that they are in part overlapping, however, we prefer to keep them atkeep them at the the same table for better comparisons (easy for the reader to understand same table for better comparisons (easy for the reader to understand bothboth))
F) In the table 5, the authors cannot show the detail
F) In the table 5, the authors cannot show the detail of "No cardiac event" versus each grade of of "No cardiac event" versus each grade of cardiac events. Indeed, the cardiac event is a time to event and this is the reason why cardiac events. Indeed, the cardiac event is a time to event and this is the reason why cumulative incidence have been used. If the authors wants to describe the characteristics of the cumulative incidence have been used. If the authors wants to describe the characteristics of the patients with cadiac event, patients with cadiac event, accordingly, they have to exclude the "No cardiac event group" and accordingly, they have to exclude the "No cardiac event group" and calculate every outcomes from the date of cardiac event. Also to delay between 2L and cardiac calculate every outcomes from the date of cardiac event. Also to delay between 2L and cardiac event have to be included, and response to 2L only if occurring before the cardiac event. The event have to be included, and response to 2L only if occurring before the cardiac event. The EEFS is complicated to calculate in such setting as the events can occur before or after cardiac FS is complicated to calculate in such setting as the events can occur before or after cardiac event.event.

 We acknowledge that these analyses are difficult to interpret. However, we believe that We acknowledge that these analyses are difficult to interpret. However, we believe that removing EFS and OS form the manuscript or only making these analyses in patients with removing EFS and OS form the manuscript or only making these analyses in patients with cardiac events since the occurrence of cardiac events, will limit the paper’s intcardiac events since the occurrence of cardiac events, will limit the paper’s interest. We can erest. We can infer that developing a grade 1infer that developing a grade 1--2 cardiac event is not harming the patient prognosis, however 2 cardiac event is not harming the patient prognosis, however developing grade 3developing grade 3--4 does. This information is only possible to display if we do comparisons of 4 does. This information is only possible to display if we do comparisons of nonnon--cardiac vs cardiac. Acknowledging this limitatiocardiac vs cardiac. Acknowledging this limitation, we have included the following sentence n, we have included the following sentence in the discussion (section of limitations of the study, page in the discussion (section of limitations of the study, page 1188 line line 338844: “We acknowledge that : “We acknowledge that comparing the OS and EFS outcomes between patients developing or not cardiac events could comparing the OS and EFS outcomes between patients developing or not cardiac events could be difficult to interpret,be difficult to interpret, as there is a time bias for patients developing cardiac events whichas there is a time bias for patients developing cardiac events which couldcould have inherently more have inherently more time time exposure exposure to develop these complications. However, we can infer to develop these complications. However, we can infer from our data that developing a grade 1from our data that developing a grade 1--2 cardiac event is not harming the patient prognosi2 cardiac event is not harming the patient prognosis, s, while developing grade 3while developing grade 3--4 does.”4 does.”
G) The first column of table 5 must be a separated table given the outcome of the cohort 2L
G) The first column of table 5 must be a separated table given the outcome of the cohort 2L (and a same table for 3L). OS and EFS have to be given at 6m and 5y as the cumulative (and a same table for 3L). OS and EFS have to be given at 6m and 5y as the cumulative incidence of cardiac event are given. Inincidence of cardiac event are given. Indeed, this is crucial to know how important is the death deed, this is crucial to know how important is the death as competing event.as competing event.

 We thank the reviewer for his comments. We thank the reviewer for his comments. We add the new table We add the new table describingdescribing 3L3L (Table (Table 6, page 6, page 1166..
H) The univariable and multivariable analysis have to be redone according to the
H) The univariable and multivariable analysis have to be redone according to the comments for comments for interpretabilityinterpretability. .

 Thank you for your commentThank you for your comment, we will keep this in mind for future studies, we will keep this in mind for future studies. . Due to time limits, Due to time limits, we will include the limitations in the discussion.we will include the limitations in the discussion.
I) The curve in figure 2 is false for several reason: first it is written
I) The curve in figure 2 is false for several reason: first it is written time from diagnosis. The time from diagnosis. The starting point of the 2L analysis is the date of start of 2L treatment and not diagnosis. If that starting point of the 2L analysis is the date of start of 2L treatment and not diagnosis. If that refers to date of diagnosis of cardiac event, what is the date of "no cardiac event"? Second, refers to date of diagnosis of cardiac event, what is the date of "no cardiac event"? Second, cardiac event is a time to event outcomcardiac event is a time to event outcome and then, the there is an immortal time bias in favor of e and then, the there is an immortal time bias in favor of having a cardiac event.having a cardiac event.

 To avoid any misinterpretation, we specify in the figure legend that it is months after To avoid any misinterpretation, we specify in the figure legend that it is months after diagnosis of R/R episode, as followsdiagnosis of R/R episode, as follows (pag(page e 1144)): “Figure : “Figure 33. Overall survival in the . Overall survival in the 2L cohort 2L cohort (intensive salvage regimens only, n=187) according to CTCAE grade of cardiac event (intensive salvage regimens only, n=187) according to CTCAE grade of cardiac event developed during observation period.”developed during observation period.”
Minor comments:
Minor comments:
1)
1) By definition a patient receiving a 3L therapy got a step for a 2L. How many patients with 3L By definition a patient receiving a 3L therapy got a step for a 2L. How many patients with 3L are incluare included in the 2L group?ded in the 2L group?

 We have now included a consort diagram (Figure We have now included a consort diagram (Figure 11) showing the patients overlapping in 2L ) showing the patients overlapping in 2L and 3L. and 3L. we also describe in the beginning of the results section, as follows (page we also describe in the beginning of the results section, as follows (page 55, line , line 190190): ): ““Overall, 327 patients with first R/R AML Overall, 327 patients with first R/R AML episode (2L cohort), and 189 patients with second R/R episode (2L cohort), and 189 patients with second R/R AML episode (3L cohort) were identified (of those 120 were patients included also in the 2L AML episode (3L cohort) were identified (of those 120 were patients included also in the 2L cohort) (Figure 1)cohort) (Figure 1).”.”
2) The authors mention to report "means or medians" and did "Mann
2) The authors mention to report "means or medians" and did "Mann--Whitney and Student t Whitney and Student t tests". The authors have to either explain when they used one or the other, or simply use tests". The authors have to either explain when they used one or the other, or simply use median and Mannmedian and Mann--Whitney tests that is valid in any situation.Whitney tests that is valid in any situation.

 We used the Mann We used the Mann WWhitney for medians and Student t tests for means. This now specified in hitney for medians and Student t tests for means. This now specified in statistical statistical section, pagesection, page 44, line , line 161161 , as follows, as follows: “: “ChiChi--square with Yates’ correction, Mannsquare with Yates’ correction, Mann––Whitney U (for non parametric distribution variables), and Student’s tWhitney U (for non parametric distribution variables), and Student’s t--tests (for parametric) were tests (for parametric) were used to analyze differences in the distribution of variables among patient used to analyze differences in the distribution of variables among patient subsetssubsets..
3) The cumulative incidence function has not been developed by Fine and Gray. Please remove
3) The cumulative incidence function has not been developed by Fine and Gray. Please remove the reference or provide the correct reference.the reference or provide the correct reference.

 We We have modified as follows, page have modified as follows, page 44, line, line 169169: “The cumulative incidence of cardiac event : “The cumulative incidence of cardiac event was was calculated using the cumulative incidence method. calculated using the cumulative incidence method.
4) Line 178: the word correlation is not correct and should be replaced by association
4) Line 178: the word correlation is not correct and should be replaced by association  We have revised the manuscript and have made that correctionWe have revised the manuscript and have made that correction
5) The curves have to not be shown more than the
5) The curves have to not be shown more than the median followmedian follow--upup

 Thanks for your comment, however, we prefer keeping the curves with more observation Thanks for your comment, however, we prefer keeping the curves with more observation period, even if subjects were censored. period, even if subjects were censored.
6) For EMD and biological values, the authors should specify if it is at time of diagnosis or R/R.
6) For EMD and biological values, the authors should specify if it is at time of diagnosis or R/R.

 Thanks, wThanks, we make a clarification in the manuscripte make a clarification in the manuscript, as follows, page , as follows, page 55, line , line 187187: “: “Median age Median age of 2L patients was 62 years (range 21of 2L patients was 62 years (range 21--87 years), 190 (58%) were male, 115 (35%) were 87 years), 190 (58%) were male, 115 (35%) were secondary or therapysecondary or therapy--related AML, and 38 (12%) had FLT3related AML, and 38 (12%) had FLT3--ITD mutation ITD mutation at the R/R timeat the R/R time
7) The reverse KM median follow
7) The reverse KM median follow--up has to be givenup has to be given

 We We now provide now provide the median the median followfollow--up in the results section, as follows (line up in the results section, as follows (line 216216, page , page 99): ): “The median follow“The median follow--up of patients alive in the 2L cohort was up of patients alive in the 2L cohort was 11361136 daysdays, and 1933 days for the , and 1933 days for the 3L cohort3L cohort.”.”
Typos:
Typos:
Line 66: A space is missing between "drugs" and ref 5.
Line 66: A space is missing between "drugs" and ref 5.  We have revised the manuscript and have made that correctionWe have revised the manuscript and have made that correction..

Author Response File: Author Response.pdf