Review Reports - KRAS Subtype Modifies Outcomes with Immunomodulatory Therapy in Advanced Pancreatic Ductal Adenocarcinoma: Evidence from Early-Phase Trials

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this retrospective study, the authors have evaluated the role of KRAS mutations in outcomes to immunotherapy in patients with pancreatic ductal adenocarcinoma (PDAC). The authors show that PDAC harboring KRAS mutation particularly KRASG12D mutation have poor overall response to immunotherapies. The trend was observed for all KRAS mutation sub-type.

Specific comments:

Limitations of this study should be detailed. There seems to be very generic description of the limitations.

The authors should enhance the discussion and incorporate the latest pan-ras inhibitor daraxonrasib findings and how it will impact patient selection for immunotherapies. Currently the discussion is more focused on allele specific agents.

The discussion should emphasize that the TMB and MSI status was low in majority of the cases studied. Why would such patients get ICI or related therapies if they were not pre-screened for TMB and MSI status? This could be a factor in the lack of response to immunotherapies and should be added in the discussion.

Some typos were found that should be checked and corrected

Legends to figures and Table can be enhanced.

Author Response

Specific comments:

Limitations of this study should be detailed. There seems to be very generic description of the limitations.

Thank you for this valuable suggestion. We agree that a more comprehensive discussion of the study limitations would improve the manuscript. We have expanded the limitations section to specifically address the retrospective nature of the study, small sample sizes within KRAS subtype subgroups, heterogeneity of immunomodulatory agents, potential confounding related to treatment line and disease burden, missing molecular data (including MSI, TMB, SMAD4, and CDKN2A), lack of immune microenvironment correlates, absence of adjustment for multiple comparisons in exploratory subgroup analyses, and the inherent selection bias associated with an early-phase clinical trial population.

We thank the reviewer for this insightful suggestion. We have expanded the discussion to incorporate recently reported clinical data with the pan-RAS inhibitor daraxonrasib and discussed how broader RAS pathway inhibition may influence future immunomodulatory treatment strategies and patient selection beyond allele-specific KRAS targeting approaches.

We thank the reviewer for this important observation. The majority of patients in our cohort had microsatellite-stable disease and low tumor mutational burden when these data were available. Importantly, most immunomodulatory therapies evaluated in this early-phase trial population were not conventional immune checkpoint inhibitors but included a broad range of investigational immunomodulatory agents with mechanisms independent of MSI or TMB status. Nevertheless, we agree that the predominance of MSI-low and TMB-low tumors likely contributed to the limited efficacy observed and have expanded the discussion accordingly.

Some typos were found that should be checked and corrected

We thank the reviewer for identifying this issue. The manuscript has been carefully reviewed and edited for grammar, spelling, consistency, and typographical errors throughout the text.

Legends to figures and Table can be enhanced.

We appreciate this suggestion. Figure and table legends have been revised to provide additional methodological details, clarify abbreviations.

Reviewer 2 Report

Comments and Suggestions for Authors

Comments to the Authors

This manuscript addresses a genuinely important and underexplored question: whether KRAS mutational subtype modifies outcomes with immunomodulatory therapy in advanced PDAC. The biological rationale is sound, the study is to the best of my knowledge, the first to examine this stratification in a clinical cohort, and the authors are worthily transparent about the limitations of their data. That transparency is not a weakness; indeed, it is what makes the manuscript credible. My comments intended to help the authors position in this work more accurately, and to address a few methodological points that require clarification before publication.

Major comments

Reforrmulation as exploratory/hypothesis-generating. The study should be positioned explicitly in the abstract, introduction, and conclusions as an exploratory, signal-generating analysis. The primary value of these findings is not to establish that immunomodulatory therapy is detrimental in KRAS G12D-mutant PDAC, but to provide a clinically grounded rationale for prospective validation. Several passages in the discussion currently overreach what the data supports. The conclusion should close with a concrete proposal: a prospective, multicenter trial with KRAS subtype stratification pre-specified at the design stage, and with immunomodulatory agent class as a stratification or randomization factor.
Heterogeneity of immunomodulatory agents. The grouping of checkpoint inhibitors, innate immune activators (BNT411, IMSA101), and myeloid/TME modulators (VT1021, Q702) into a single binary exposure variable is the study's most important analytical limitation and should be explicitly acknowledged as such not merely listed alongside sample size and retrospective design. These agents operate through mechanistically distinct pathways, and their pooling precludes any inference about which class of immunomodulation may be driving the observed signal. The authors should discuss whether any agent-class-specific trends are discernible, even at the descriptive level, while being clear that formal subgroup analyses by agent class are underpowered.
Confounding by treatment line. Table 3 shows an imbalance: 70% of patients not receiving immunomodulatory therapy were enrolled first-line, versus only 5.9% of those who did receive it. This is the primary confounding structure of the study, and while the multivariable model addresses it partially, the attenuation of the immunotherapy HR from 1.78 (univariate) to 1.61 (multivariable, p=0.06) suggests residual confounding cannot be excluded. This should be discussed more prominently not to invalidate the findings, but to frame them correctly.
Selection of Fleming-Harrington weights. The use of different FH weighting schemes across subgroups (FH(1,0) for G12D, FH(1,1) for G12V and other variants) requires justification. If weights were selected based on observed KM curve patterns, this should be stated explicitly and the potential for inflated type I error in exploratory subgroup analyses acknowledged. Pre-specification of a single weighting scheme or reporting all FH variants alongside the standard log-rank for each subgroup, would be more transparent.

Minor comments

The missing data on MSI, TMB, SMAD4, and CDKN2A status (30–38% unknown across variables) limits co-mutation analysis. This is an unfortunate circumstance but an understandable consequence of the study's temporal span NGS was not uniformly available in 2014–2018 and does not require apology but warrants an explicit acknowledgment that these variables may be important effect modifiers in future analyses.
Figure legends for Figures 2–4 should state the number of patients per arm and the specific FH weighting used.
The abstract reports the multivariable HR as 1.61 with 95% CI 0.98–2.66 and p=0.06, but Table 5 shows HR 1.613 (0.977–2.663) with p=0.062. Minor discrepancy, please harmonize.

Overall recommendation

Major revision. In my opinion, the scientific question is relevant, the biological rationale is well-supported in the literature, and the authors' intellectual honesty about their data's limitations which is really appreciated. With appropriate reframing and the methodological clarifications above, this work provides a meaningful contribution to the field and a clear rationale for the prospective studies this question ultimately requires.

Author Response

Major comments

Reforrmulation as exploratory/hypothesis-generating. The study should be positioned explicitly in the abstract, introduction, and conclusions as an exploratory, signal-generating analysis. The primary value of these findings is not to establish that immunomodulatory therapy is detrimental in KRAS G12D-mutant PDAC, but to provide a clinically grounded rationale for prospective validation. Several passages in the discussion currently overreach what the data supports. The conclusion should close with a concrete proposal: a prospective, multicenter trial with KRAS subtype stratification pre-specified at the design stage, and with immunomodulatory agent class as a stratification or randomization factor.

We thank the reviewer for this important observation. We agree that the study should be positioned as an exploratory, hypothesis-generating analysis. Accordingly, we have revised the Abstract, Introduction, Discussion, and Conclusion sections to consistently emphasize the exploratory nature of the study and the need for prospective validation. We have also modified language that could be interpreted as implying causality and now frame the findings as signals that warrant further investigation.

Heterogeneity of immunomodulatory agents. The grouping of checkpoint inhibitors, innate immune activators (BNT411, IMSA101), and myeloid/TME modulators (VT1021, Q702) into a single binary exposure variable is the study's most important analytical limitation and should be explicitly acknowledged as such not merely listed alongside sample size and retrospective design. These agents operate through mechanistically distinct pathways, and their pooling precludes any inference about which class of immunomodulation may be driving the observed signal. The authors should discuss whether any agent-class-specific trends are discernible, even at the descriptive level, while being clear that formal subgroup analyses by agent class are underpowered.

In response to the reviewer, we performed an exploratory subgroup analysis separating checkpoint inhibitors (CPI) from non-checkpoint (non-CPI) immunomodulatory therapies. Kaplan Meier curves and log-rank tests were used to compare overall survival and progression-free survival between these two subgroups. No statistically significant differences were observed in OS or PFS between the CPI and non-CPI immunomodulatory therapy recipients. We note that this comparison included only 17 CPI and 25 non-CPI immunomodulatory therapy recipients and was underpowered. The non-significant finding should be interpreted as inconclusive rather than as evidence of equivalent efficacy between the mechanistically distinct agents.

Further stratification within individual KRAS subtypes were not feasible due to sample size constraints, as each KRAS and immunotherapy subtype combination only had less than 10 patients, which precludes reliable survival estimation. We acknowledged that this is a limitation of our study and have noted them in the discussion section of the manuscript.

Confounding by treatment line. Table 3 shows an imbalance: 70% of patients not receiving immunomodulatory therapy were enrolled first-line, versus only 5.9% of those who did receive it. This is the primary confounding structure of the study, and while the multivariable model addresses it partially, the attenuation of the immunotherapy HR from 1.78 (univariate) to 1.61 (multivariable, p=0.06) suggests residual confounding cannot be excluded. This should be discussed more prominently not to invalidate the findings, but to frame them correctly.

We thank the reviewer for highlighting this important source of confounding. We agree that patients receiving immunomodulatory therapies were substantially more heavily pretreated than those receiving non-immunomodulatory therapies. To address this issue, treatment line was included in the multivariable Cox model, where immunomodulatory therapy exposure was attenuated from HR 1.78 in univariate analysis to HR 1.61 and no longer reached statistical significance (p=0.06). We have expanded the Discussion to further emphasize that the observed association may reflect differences in disease burden and prior treatment exposure rather than a direct detrimental effect of immunomodulatory therapy.

Selection of Fleming-Harrington weights. The use of different FH weighting schemes across subgroups (FH(1,0) for G12D, FH(1,1) for G12V and other variants) requires justification. If weights were selected based on observed KM curve patterns, this should be stated explicitly and the potential for inflated type I error in exploratory subgroup analyses acknowledged. Pre-specification of a single weighting scheme or reporting all FH variants alongside the standard log-rank for each subgroup, would be more transparent.

We thank the reviewer for this suggestion. We have expanded the Statistical Analysis section to clarify the rationale for selecting specific Fleming-Harrington weighting schemes based on the observed timing of separation in Kaplan-Meier curves. We have also acknowledged the exploratory nature of these analyses and the potential risk of inflated type I error associated with multiple subgroup comparisons.

Minor comments

The missing data on MSI, TMB, SMAD4, and CDKN2A status (30–38% unknown across variables) limits co-mutation analysis. This is an unfortunate circumstance but an understandable consequence of the study's temporal span NGS was not uniformly available in 2014–2018 and does not require apology but warrants an explicit acknowledgment that these variables may be important effect modifiers in future analyses.

We have expanded the discussion to more explicitly acknowledge the substantial proportion of missing molecular data for MSI, TMB, SMAD4, and CDKN2A and the potential impact of this limitation on interpretation.

Figure legends for Figures 2–4 should state the number of patients per arm and the specific FH weighting used.

We thank the reviewer, fixed.

The abstract reports the multivariable HR as 1.61 with 95% CI 0.98–2.66 and p=0.06, but Table 5 shows HR 1.613 (0.977–2.663) with p=0.062. Minor discrepancy, please harmonize.

We thank the reviewer for identifying this discrepancy. The values reflect rounding differences from the same model output. We have harmonized reporting throughout the manuscript.

Overall recommendation

We thank the reviewer fort he comments.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript addresses an interesting and clinically relevant question regarding whether KRAS subtype influences outcomes with immunomodulatory therapy in advanced PDAC; however, some methodological, interpretive, and presentation issues should be addressed to improve the clarity, reproducibility, and strength of the conclusions.

The term “immunomodulatory therapy” is very broad and includes biologically different agents, so the authors should provide a clearer rationale for grouping checkpoint inhibitors, innate immune activators, myeloid/TME modulators, and other agents together in the same analysis.
Because patients receiving immunomodulatory therapy had significantly more prior systemic therapy lines, the manuscript should more carefully distinguish whether inferior survival reflects treatment effect, later-line use, disease burden, or selection bias.
The KRAS G12D subgroup analysis is interesting, but the small number of patients limits confidence in the conclusion, and the authors should present it more explicitly as hypothesis-generating rather than definitive.
It is not fully clear why different Fleming–Harrington weighting schemes were selected for different comparisons; therefore, the authors should explain the criteria used to choose each test and whether these choices were pre-specified.
The manuscript would be strengthened by reporting response-related outcomes, such as objective response rate, disease control rate, or best RECIST response, to better support the survival findings.
Missing molecular data for SMAD4, CDKN2A, MSI, and TMB are substantial, making interpretation difficult; the authors should clarify how missingness was handled and discuss its potential impact more directly.
Several conclusions imply that KRAS subtype may determine immunotherapy response, whereas the multivariable analysis shows only a non-significant trend for immunomodulatory therapy, so the language should be moderated throughout the abstract, results, and discussion.
The figures and tables contain some formatting and readability issues, including small Kaplan–Meier plots, inconsistent terminology such as “immunotherapy” versus “immunomodulatory therapy,” and minor grammatical errors, which should be corrected before publication.

Author Response

The term “immunomodulatory therapy” is very broad and includes biologically different agents, so the authors should provide a clearer rationale for grouping checkpoint inhibitors, innate immune activators, myeloid/TME modulators, and other agents together in the same analysis.

We thank the reviewer for this important observation. We agree that immunomodulatory therapy encompasses biologically distinct therapeutic classes. To improve transparency, we have added data comparing checkpoint inhibitor and non-checkpoint immunomodulatory therapies according to KRAS status to results. We have also expanded the discussion to acknowledge that pooling these agents may obscure class-specific effects and that larger studies are needed to evaluate individual therapeutic classes separately.

Because patients receiving immunomodulatory therapy had significantly more prior systemic therapy lines, the manuscript should more carefully distinguish whether inferior survival reflects treatment effect, later-line use, disease burden, or selection bias.

We agree that patients receiving immunomodulatory therapies were substantially more heavily pretreated than those receiving non-immunomodulatory therapies. To address this potential confounding, treatment line was incorporated into the multivariable Cox model. After adjustment, the association between immunomodulatory therapy exposure and overall survival was attenuated and no longer statistically significant (HR 1.61, 95% CI 0.98–2.66, p=0.06). We have expanded the Discussion to further emphasize that residual confounding may contribute to the observed associations.

The KRAS G12D subgroup analysis is interesting, but the small number of patients limits confidence in the conclusion, and the authors should present it more explicitly as hypothesis-generating rather than definitive.

We appreciate this comment and have revised the Abstract, Introduction, Discussion, and Conclusion to consistently frame the study as an exploratory, hypothesis-generating analysis requiring prospective validation.

It is not fully clear why different Fleming–Harrington weighting schemes were selected for different comparisons; therefore, the authors should explain the criteria used to choose each test and whether these choices were pre-specified.

We have expanded the Statistical Analysis section to clarify that Fleming-Harrington weighting schemes were selected post hoc based on visual inspection of Kaplan-Meier curves to explore potential non-proportional hazards.

The manuscript would be strengthened by reporting response-related outcomes, such as objective response rate, disease control rate, or best RECIST response, to better support the survival findings.

We agree that response-based endpoints would provide valuable complementary information. However, because patients were treated across multiple early-phase trials over a prolonged study period and response assessments were not uniformly captured in the retrospective database, robust response analyses were not feasible. We have added this limitation to the Discussion.

Missing molecular data for SMAD4, CDKN2A, MSI, and TMB are substantial, making interpretation difficult; the authors should clarify how missingness was handled and discuss its potential impact more directly.

We have expanded the Discussion to more explicitly acknowledge the extent of missing molecular data and the resulting limitations regarding interpretation of co-mutation and biomarker effects.

Several conclusions imply that KRAS subtype may determine immunotherapy response, whereas the multivariable analysis shows only a non-significant trend for immunomodulatory therapy, so the language should be moderated throughout the abstract, results, and discussion.

Thank you for suggestions, fixed.

The figures and tables contain some formatting and readability issues, including small Kaplan–Meier plots, inconsistent terminology such as “immunotherapy” versus “immunomodulatory therapy,” and minor grammatical errors, which should be corrected before publication.

We thank the reviewer for these suggestions. Kaplan-Meier plots have been reformatted for improved readability, terminology has been standardized throughout the manuscript, and the manuscript has undergone additional language editing to correct grammatical and typographical errors.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,
This time, I have reviewed the revised manuscript alongside the authors' response to my prior comments. The revisions are generally well-executed, and the authors have addressed the principal concerns raised in the first round. However, several points require additional attention before the manuscript can be considered for acceptance:

Major comment 1: Exploratory framing (addressed, with residual concern)

The abstract, introduction, and conclusions have been revised to include explicit language characterizing this as an exploratory, hypothesis-generating analysis. This is a meaningful improvement. One residual issue remains: the opening sentence of the Discussion in v2 states that outcomes "differ significantly according to KRAS mutational subtype," whereas v1 used the more careful phrasing "may differ." The word "significantly" carries statistical implications and represents a step backward from the appropriately hedged language that now appears elsewhere in the revised manuscript. Given that the G12D subgroup result is based on n=13 immunotherapy recipients, a single exploratory Fleming-Harrington test without multiple comparison adjustment, and a p-value that would not survive Bonferroni correction across the three subtypes tested, this phrasing is not defensible. The authors should restore cautionary language consistent with the tone achieved throughout the rest of the revised discussion.

Additionally, the conclusion in v1 closed with "These findings should be considered hypothesis-generating and require validation in prospective KRAS subtype-informed studies." The revised conclusion in v2 reads: "These results warrant prospective validation and suggest that future clinical trials should incorporate KRAS mutational subtype stratification when evaluating immunomodulatory strategies in pancreatic cancer." While reasonable, my original comment specifically requested that the conclusion include a concrete proposal for the design of prospective validation namely, pre-specified KRAS subtype stratification at the design stage, and immunomodulatory agent class as a stratification or randomization factor. This has not been incorporated. A single sentence specifying these design elements would substantially strengthen the translational value of the conclusion.

Major comment 2: Heterogeneity of immunomodulatory agents (partially addressed)

The authors have added an exploratory post-hoc comparison of checkpoint inhibitor versus non-checkpoint immunomodulatory therapy recipients in the overall cohort, appropriately reporting non-significant differences in OS (6.7 vs. 3.25 months, p=0.08) and PFS (1.94 vs. 1.58 months, p=0.12), and acknowledging that this analysis was underpowered. The addition of this analysis and its transparent description in the limitations section is appreciated.

However, two issues remain. First, the sentence describing the checkpoint vs. non-checkpoint comparison ("checkpoint inhibitor recipients showered a longer median OS") contains a typographical error ("showered" should be "showed"). This should be corrected. Second, the authors have not provided, even at the descriptive level, the breakdown of which immunomodulatory agents were received within each KRAS subtype subgroup. Table 1 shows agent distribution by KRAS status (wild-type vs. mutant), but does not disaggregate by G12D, G12V, or other KRAS variants. Given that VT1021 (myeloid/TME modulator) accounts for 8 of 27 immunomodulatory therapy exposures in KRAS-mutant patients, and checkpoint inhibitors account for 11, the distribution of agent class within the G12D subgroup (n=13 immunotherapy recipients) is not transparent from the current tables. This information should be provided, even as a supplementary table or footnote, to allow readers to assess whether a single agent or agent class disproportionately drives the G12D signal.

Major comment 3: Confounding by treatment line (addressed)

The authors have expanded the discussion to address the treatment line imbalance more prominently. The revised text appropriately states that the observed associations "may reflect underlying tumor biology or confounding by treatment line and disease burden." The attenuation of the HR from 1.78 to 1.61 with loss of statistical significance upon multivariable adjustment is now foregrounded in the discussion rather than buried. This revision is satisfactory.

Major comment 4: Fleming-Harrington weight selection (addressed)

The statistical methods section now includes an explicit statement that weighting schemes were selected post hoc based on observed patterns of KM curve separation and that this approach should be considered exploratory. This is an appropriate and transparent clarification. No further revision is required on this point.

Minor comment 1 — Missing data on molecular biomarkers (addressed)

Satisfactory. The expanded discussion acknowledges the temporal limitations of NGS availability across the study period and frames missing data as a potential effect modifier in future analyses, without undue apology.

Minor comment 2: Figure legends (addressed)

The authors indicate that figure legends for Figures 2-4 have been updated to include the number of patients per arm and FH weighting used. However, in the revised manuscript as submitted, Figure 4's legend does not explicitly state the FH(1,0) weighting. The legends for Figures 2 and 3 reference standard log-rank tests (implicitly FH[0,0]), which is consistent with the text. Figure 4 uses FH(1,0), and this should be stated in the figure legend rather than only in the text body.

Minor comment 3: HR discrepancy (addressed)

The harmonization of HR values (1.61, 95% CI 0.98-2.66, p=0.06) between the abstract and Table 5 is confirmed in the revised manuscript. This has been satisfactorily resolved.

Additional point not previously raised: Introduction, typographical error

The introduction contains a percentage formatting error that was present in v1 and persists in v2: "more than %90 of pancreatic cancer tissue samples" should read "more than 90% of pancreatic cancer tissue samples." This should be corrected.

Additional point: RASolute 302 trial reference

The v2 manuscript has removed the paragraph referencing the RASolute 302 trial and daraxonrasib (Ref. 43 / Wolpin et al., NEJM 2026), which appeared in v1's discussion. Given that this trial represents directly relevant context for the therapeutic landscape being discussed, its removal weakens the translational framing. The authors should consider reinstating this reference or providing a scientific rationale for its exclusion.

Regards

Author Response

We sincerely thank the reviewer for the careful re-evaluation of our revised manuscript and for the constructive feedback. We are pleased that the reviewer found the major concerns from the first review round to be adequately addressed. We have carefully considered the remaining comments and revised the manuscript accordingly.

Major comment 1: Exploratory framing (addressed, with residual concern)

We thank the reviewer for this important observation and agree that the wording should remain consistent with the exploratory nature of our analysis. We have revised the opening sentence of the Discussion to state that outcomes "may differ according to KRAS mutational subtype" rather than "differ significantly according to KRAS mutational subtype."

We have also expanded the Conclusion section to include a more specific recommendation for future study design. The revised text now states that prospective validation studies should incorporate pre-specified KRAS subtype stratification at study entry and consider immunomodulatory agent class as a stratification factor to better evaluate subtype-specific therapeutic effects.

Major comment 2: Heterogeneity of immunomodulatory agents (partially addressed)

We thank the reviewer for this insightful suggestion.

First, we have corrected the typographical error, replacing "showered" with "showed."

Second, to improve transparency regarding treatment heterogeneity, we have added a footnote to Table 1 describing the distribution of immunomodulatory agents across KRAS-mutant subtypes. Among patients receiving immunomodulatory therapy, KRAS G12D tumors (n=13) were treated with checkpoint inhibitors (n=3), BNT411 (n=4), IMSA101 (n=1), Q702 (n=1), and VT1021 (n=4). KRAS G12V tumors (n=7) received checkpoint inhibitors (n=4), IMSA101 (n=1), and VT1021 (n=2), while other KRAS variants (n=7) received checkpoint inhibitors (n=4), IMSA101 (n=1), and VT1021 (n=2).

Importantly, no single immunomodulatory agent or agent class accounted for the majority of exposures within the G12D subgroup, suggesting that the observed exploratory association was not driven by a single investigational therapy. Given the limited sample size within each KRAS subtype-agent combination, formal statistical comparisons were not feasible, and these data are therefore presented descriptively.

Major comment 3: Confounding by treatment line (addressed)

We thank the reviewer for acknowledging the revisions made in response to this concern and appreciate the positive assessment.

Major comment 4: Fleming-Harrington weight selection (addressed)

We thank the reviewer for acknowledging the clarification added to the Statistical Methods section.

Minor comment 1 — Missing data on molecular biomarkers (addressed)

We appreciate the reviewer’s positive assessment.

Minor comment 2: Figure legends (addressed)

We agree and have revised the Figure 4 legend to explicitly state that survival comparisons were performed using the Fleming-Harrington FH(1,0) weighted log-rank test.

Minor comment 3: HR discrepancy (addressed)

The harmonization of HR values (1.61, 95% CI 0.98-2.66, p=0.06) between the abstract and Table 5 is confirmed in the revised manuscript. This has been satisfactorily resolved.

We thank the reviewer for confirming that this issue has been resolved.

Additional point not previously raised: Introduction, typographical error

We thank the reviewer for identifying this typographical error. The text has been corrected.

Additional point: RASolute 302 trial reference

We appreciate this suggestion and agree that the RASolute 302 trial provides important context regarding emerging KRAS-targeted therapeutic strategies. We have reinstated a brief discussion of the trial in the Discussion section to provide contemporary context for KRAS-directed therapeutic development. We also emphasize that, unlike direct KRAS inhibition, our study focused on immunomodulatory approaches and exploratory differences among KRAS mutational subtypes, highlighting the need for future subtype-informed therapeutic strategies.

We thank the reviewer again for the thoughtful comments, which have strengthened the manuscript.

Regards

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript is now improved, and I believe it meets the standards for publication. I have no further concerns.

Author Response

We thank the reviewer,

Sincerely,