Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper provides a review of advances in hemotologic clinical trials through artificial intelligence (AI), circulating tumor DNA (ctDNA), and real world evidence (RWE).  The review appears useful and timely given how practices are changing in the field.  These three directions each are beginning to have large impact in different ways and great promise for the future.  The work provides a good summary of some of the major challenges of the field, and how AI, ctDNA, and RWE each address some of them.  The material is well referenced, with nice collections of relevant literature in each of these areas.  The material is technically sound, as far as I can judge it, aside from some minor points noted below.  I could see it finding a good audience among clinicians and clinical researchers working on hematologic clinical trials who are looking for an introduction to some promising new directions in the field.  However, there are some points on which I think there may be room for improvement.

 

Major Critiques

 

1. I feel the scope of the review might be better justified.  From the title, I was expecting more discussion of synergy between AI, ctDNA, and RWE, but they were treated as three essentially disjoint contributions, with some limited examples toward the end of work combining more than one of these three threads.  That leaves the question of why consider all three in a single review.  Any one of the three could be the subject of its own review paper and might be expanded significantly if so.  I would therefore want to see a more compelling argument for why consider all three of these directions in a single review as opposed to a more focused review on each one individually.

 

2. I felt the future perspective (Section 9) could use some expansion.  The manuscript essentially repeats the previously described advantages AI, ctDNA, and RWE can bring and suggests they will find greater use in the future because of these.  I think the review would benefit more from discussing where these techniques have unrealized potential that might be developed better in the future.  I would also want to see more informed speculation on synergies between the AI, ctDNA, and RWE that might be better exploited and what that might look like for the future of the field.  A lot of the value for a paper like this, especially for what are still emerging technologies in clinical cancer treatment, is in telling us why the field is excited about them and what is likely to be coming in the future.

 

3. I also felt that some areas were condensed too much to do them justice, perhaps because there is so much material already there in considered AI, ctDNA, and RWE together.  I particularly thought that the regulatory and ethical considerations (Sections 6 and 7) could be much more thorough.  I realize some limit of scope is necessary.  Each of the six subsections of Sections 6 and 7 could also have been its own review paper.  But more depth on these points would be helpful with respect to current regulatory and ethical challenges to bringing these methods more effectively to clinical practice and how they are being addressed, with suitable citations.

 

4. In general, I felt the paper did a good job of describing what AI, ctDNA, and RWE have to offer the field but was lacking in considering limitations of each technology and where research is currently in addressing those limitations.  That is, where are these technologies falling short of their potential so far?  What directions is the field pursuing to address this?

 

Minor Critiques

 

5. Line 160: “can markedly improves” should be “can markedly improve”

 

6. Figure 1 and associated text:  I would dispute the claim that GenAI is a subset of DL.  While references to GenAI today often refer to some of the modern DL techniques (e.g., LLMs and diffusion models), there are older techniques for generative AI that are not DL.  For example, older Bayesian ML methods lent themselves naturally to generative models.  I would suggest either clarifying the particular kind of GenAI that is a subset of DL or note that GenAI and DL are intersecting but neither contains the other.

 

7. Lines 167-168: “related to data quality and algorithmic bias, and interpretability” I believe should be “related to data quality, algorithmic bias, and interpretability”

 

8. Line 232:  The text seems not to be correct grammatically but  I am not completely sure what was intended.  Perhaps “offers” should be “offering”?

 

Author Response

Manuscript Title:

Integrating Artificial Intelligence, Circulating Tumor DNA, and Real-World Evidence to Optimize Hematologic Clinical Trials: Advances in Trial Design and Decision-Making.

Manuscript ID: cancers-4147009

Dear Editor and Reviewers.

We thank the reviewers for their careful evaluation of our manuscript and for their constructive and insightful comments. We have revised the manuscript extensively in response to all suggestions. These revisions have significantly improved the clarity, scientific rigor, and overall impact of the work.

Below, we provide a detailed, point-by-point response to each comment. All changes have been clearly incorporated into the revised manuscript.

Reviewer 1:

Comment 1: I feel the scope of the review might be better justified.  From the title, I was expecting more discussion of synergy between AI, ctDNA, and RWE, but they were treated as three essentially disjoint contributions, with some limited examples toward the end of work combining more than one of these three threads.  That leaves the question of why consider all three in a single review.  Any one of the three could be the subject of its own review paper and might be expanded significantly if so.  I would therefore want to see a more compelling argument for why consider all three of these directions in a single review as opposed to a more focused review on each one individually.

Response:
We have revised the Introduction and Section 9 to clarify that AI, ctDNA, and RWE represent complementary and synergistic components of next-generation adaptive clinical trials. Additional text has been included to emphasize their integration into a unified framework. (Line 69-79, 457-511)

Comment 2: I felt the future perspective (Section 9) could use some expansion.  The manuscript essentially repeats the previously described advantages AI, ctDNA, and RWE can bring and suggests they will find greater use in the future because of these.  I think the review would benefit more from discussing where these techniques have unrealized potential that might be developed better in the future.  I would also want to see more informed speculation on synergies between the AI, ctDNA, and RWE that might be better exploited and what that might look like for the future of the field.  A lot of the value for a paper like this, especially for what are still emerging technologies in clinical cancer treatment, is in telling us why the field is excited about them and what is likely to be coming in the future.

Response:
Section 9 has been substantially expanded to include forward-looking scenarios, unrealized opportunities, and explicit examples of synergy among AI, ctDNA, and RWE.
(Line 526-538, 547-557, 562-565)

Comment 3: I also felt that some areas were condensed too much to do them justice, perhaps because there is so much material already there in considered AI, ctDNA, and RWE together.  I particularly thought that the regulatory and ethical considerations (Sections 6 and 7) could be much more thorough.  I realize some limit of scope is necessary.  Each of the six subsections of Sections 6 and 7 could also have been its own review paper.  But more depth on these points would be helpful with respect to current regulatory and ethical challenges to bringing these methods more effectively to clinical practice and how they are being addressed, with suitable citations.

Response:
Sections 6 and 7 have been expanded to include FDA and EMA perspectives, EU AI Act implications, and deeper discussion on data governance, bias, and privacy.
(Line 361-377, 399-404,  411-417,  423-429).

Comment 4: In general, I felt the paper did a good job of describing what AI, ctDNA, and RWE have to offer the field but was lacking in considering limitations of each technology and where research is currently in addressing those limitations.  That is, where are these technologies falling short of their potential so far?  What directions is the field pursuing to address this?

Response:
A new subsection (Section - 8) has been added addressing limitations of AI, ctDNA, and RWE, along with current research efforts to overcome these challenges. (Line 431-454).

Minor comments:

All grammatical corrections have been implemented, and the GenAI definition has been clarified.
We thank the reviewer for the comment. We agree that the original wording oversimplified the relationship between generative artificial intelligence (GenAI) and deep learning (DL). We have revised Figure 1 associated text to reflect a more accurate conceptual framework. While many contemporary GenAI approaches (e.g., large language models and diffusion models) are based on deep learning architectures, earlier generative models “such as probabilistic and Bayesian approaches” are not inherently dependent on DL (Line 159-164).

Sincerely,

Prof. Abdurraouf Mokhtar Mahmoud

Reviewer 2 Report

Comments and Suggestions for Authors

The work and idea conceptualization of the authors is good. Even though some modifications are required in this manuscript to attain our journal standards. Some of them I mentioned below. Kindly request the author to include this
1. Problem identification and the gap of the study were not mentioned clearly.
2. A comparative table was required to improve the readability.
3. Please implement the practical framework. No need for conceptual work.
4. The stepwise workflow model was not mentioned in your study
5. Need a clear description of the dataset and why you chose this algorithm.
6. Did you check your dataset with this algorithm? Did any algorithmic bias happen in your study?
7. Please rewrite and improve the conclusion to make it easy to read.
8. Too many references are in the manuscript.

Comments on the Quality of English Language

After these corrections I kindly request our beloved editor to give accepatnce for this paper

Author Response

Manuscript Title:

Integrating Artificial Intelligence, Circulating Tumor DNA, and Real-World Evidence to Optimize Hematologic Clinical Trials: Advances in Trial Design and Decision-Making.

Manuscript ID: cancers-4147009

Dear Editor and Reviewers.

We thank the reviewers for their careful evaluation of our manuscript and for their constructive and insightful comments. We have revised the manuscript extensively in response to all suggestions. These revisions have significantly improved the clarity, scientific rigor, and overall impact of the work.

Below, we provide a detailed, point-by-point response to each comment. All changes have been clearly incorporated into the revised manuscript.

Reviewer 2:

Comment 1: Problem identification and the gap of the study were not mentioned clearly.

Response:
The Introduction has been revised to clearly define the limitations of conventional trials and the gap addressed by integrating AI, ctDNA, and RWE. (Line 69- 78)

Comment 2: A comparative table was required to improve the readability.

Response:
A new comparative table has been added summarizing roles, strengths, and limitations of AI, ctDNA, and RWE. (Line 187)

Comments 3–6: Framework, dataset, algorithm.

Response:
As this is a narrative review, no original dataset or algorithm was used. However, a conceptual workflow model has been added to illustrate practical implementation.

Comment 7: Please rewrite and improve the conclusion to make it easy to read.

Response:
We thank the reviewer for this helpful suggestion. The Conclusion section has been fully revised to improve clarity, readability, and overall flow (Line 567-597).

Comment 8: Too many references.

Response:
The reference list has been revised and prioritize key studies.

Sincerely,

Abdurraouf Mokhtar Mahmoud

Reviewer 3 Report

Comments and Suggestions for Authors

This article innovatively integrates artificial intelligence, circulating tumor DNA, and real-world evidence to build an adaptive hematology clinical trial framework. Through AI-driven patient stratification, dynamic ctDNA monitoring, and RWE external controls, it aims to enhance trial efficiency and personalization. The following five questions are raised for a better version: 

1. Has the generalizability of the AI model across multicenter heterogeneous data been sufficiently validated?
2. Are the clinical validation standards for ctDNA as a surrogate endpoint consistent, and how can differences in technical sensitivity be avoided?
3. When using RWE to construct synthetic control arms, how can bias from unmeasured confounders be effectively controlled?
4. Is the regulatory compliance pathway for the integrated framework clear, particularly concerning the approval challenges of dynamic AI decision-making?
5. How can the balance between data sharing and patient privacy be managed at the ethical level, ensuring algorithmic fairness?

Author Response

Manuscript Title:

Integrating Artificial Intelligence, Circulating Tumor DNA, and Real-World Evidence to Optimize Hematologic Clinical Trials: Advances in Trial Design and Decision-Making.

Manuscript ID: cancers-4147009

Dear Editor and Reviewers.

We thank the reviewers for their careful evaluation of our manuscript and for their constructive and insightful comments. We have revised the manuscript extensively in response to all suggestions. These revisions have significantly improved the clarity, scientific rigor, and overall impact of the work.

Below, we provide a detailed, point-by-point response to each comment. All changes have been clearly incorporated into the revised manuscript.

Reviewer 3:

Comment 1: Has the generalizability of the AI model across multicenter heterogeneous data been sufficiently validated?

Response:
We added discussion on multicenter validation and federated learning approaches (Line 123-132, 361-377, 431-454, 583-597).

Comment 2: Are the clinical validation standards for ctDNA as a surrogate endpoint consistent, and how can differences in technical sensitivity be avoided?

Response:
We expanded discussion on assay variability and need for standardization (Line 361-377).

Comment 3: When using RWE to construct synthetic control arms, how can bias from unmeasured confounders be effectively controlled?

Response:
We included advanced causal inference methods such as propensity score matching and inverse probability weighting. (Line 325-336)

Comment 4: Is the regulatory compliance pathway for the integrated framework clear, particularly concerning the approval challenges of dynamic AI decision-making?

Response:

We clarified regulatory challenges related to adaptive AI systems. (Line 126-135)

Comment 5: How can the balance between data sharing and patient privacy be managed at the ethical level, ensuring algorithmic fairness?

Response:
We expanded ethical considerations including privacy-preserving AI and fairness. (Line 399-404, 411-417, 423-429)

Final Statement:

We believe these revisions have significantly strengthened the manuscript. We thank the reviewers for their valuable feedback

Sincerely,

Abdurraouf Mokhtar Mahmoud

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

All of my substantive concerns have been sufficiently addressed.  As before, I believe this is an interesting review that makes a useful contribution to the literature.  I have no more major concerns.  I noticed just a couple of minor grammatical errors in the revised text:

On line 525, there is at least a missing word.  I would suggest rephrasing as "This approach --- AI, ctDNA, and RWE --- is expected..."

I think the sentence from lines 529-534 does not flow grammatically as written.  I would suggest splitting it into two sentences at line 532:  "... and advancing personalized medicine.  This approach will increasingly..."

Author Response

Manuscript Title:

Integrating Artificial Intelligence, Circulating Tumor DNA, and Real-World Evidence to Optimize Hematologic Clinical Trials: Advances in Trial Design and Decision-Making.

Manuscript ID: cancers-4147009

Dear Editor and Reviewers.

We thank the reviewers for their careful evaluation of our manuscript and for their constructive and insightful comments. We have revised the manuscript in response to all suggestions. These revisions have improved the clarity, and overall impact of the work.

Below, we provide a detailed, point-by-point response to each comment. All changes have been clearly incorporated into the revised manuscript.

Minor comments: 

On line 525, there is at least a missing word.  I would suggest rephrasing as "This approach --- AI, ctDNA, and RWE --- is expected..."

I think the sentence from lines 529-534 does not flow grammatically as written.  I would suggest splitting it into two sentences at line 532:  "... and advancing personalized medicine.  This approach will increasingly..."

All grammatical corrections have been implemented.