Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors present a retrospective single-center study evaluating days alive and out of hospital within 30 days (DAOH30) after curative gastrectomy for gastric adenocarcinoma. The manuscript addresses an interesting and increasingly relevant concept in perioperative outcomes research, as DAOH30 integrates index hospitalization, early readmission, and mortality into a single measure of postoperative healthcare burden. Overall, the study is carefully conducted, the endpoint is clearly defined, and the statistical approach is generally appropriate. I particularly appreciated the authors' effort to avoid overinterpreting DAOH30 as being independent of length of stay and their transparent acknowledgment that, in this cohort, most of the variability in DAOH30 was driven by the index hospitalization because no 30-day deaths occurred and only eight patients were readmitted.

Nevertheless, I believe that several aspects of the manuscript could be further improved before publication.

First, the clinical interpretation of postoperative complications deserves some expansion. The manuscript convincingly demonstrates a progressive reduction in DAOH30 with increasing Clavien–Dindo severity, with the largest difference observed in patients experiencing grade III or higher complications. However, complication severity alone does not fully characterize the subsequent clinical trajectory. The type of complication and, importantly, the way in which it is managed may substantially influence length of stay, need for reoperation, subsequent readmission, and ultimately DAOH30. This point is particularly relevant for anastomotic leakage, which is one of the most clinically relevant complications following gastrectomy. The authors report five anastomotic leaks in their cohort and appropriately acknowledge that the small number of events precludes robust inference regarding their specific impact on DAOH30. Nevertheless, the management of postoperative upper gastrointestinal leaks has changed considerably in recent years, with therapeutic endoscopy increasingly providing alternatives to surgical reintervention through stenting, endoscopic vacuum therapy, internal drainage, closure techniques, and other tailored approaches. Since these strategies may themselves affect hospital exposure and recovery, they are conceptually highly relevant to an outcome such as DAOH30. I therefore suggest that the authors briefly discuss this evolving management paradigm when addressing the impact of major postoperative complications and anastomotic leakage. Recent comprehensive reviews specifically addressing endoscopic treatment of gastrointestinal leaks and the evolving role of stents and alternative endoscopic therapies in upper gastrointestinal defects would provide an appropriate contemporary framework for this discussion (PMID: 41479933 and PMID: 41008715). This addition would also emphasize that DAOH30 may potentially capture not only the occurrence and severity of complications but also the consequences of different strategies used to manage them.

A second issue concerns the interpretation of DAOH30 as a postoperative outcome measure. In the present cohort, 93.5% of patients survived without readmission, and no deaths occurred within 30 days. As the authors correctly state, DAOH30 was therefore mathematically equivalent to 30 minus index length of stay for most patients. This does not invalidate the endpoint, but it substantially limits the amount of additional information provided by DAOH30 over conventional LOS in this specific dataset. I believe this deserves even greater emphasis in the Discussion. The results primarily demonstrate the feasibility and known-groups discrimination of DAOH30 in a gastrectomy cohort rather than establishing its incremental clinical value over LOS. A larger multicenter population with greater variability in readmissions and early mortality would be required to demonstrate whether DAOH30 provides meaningful additional discrimination or prognostic information beyond conventional postoperative outcomes.

Related to this point, the manuscript frequently describes DAOH30 as a patient-centered endpoint. While the concept is intuitively patient relevant, no patient-reported outcome or direct measure of functional recovery, quality of life, or patient preference was collected in this study. I would therefore recommend slightly moderating this terminology throughout the manuscript. DAOH30 can reasonably be described as a recovery- or life-impact measure that is potentially patient centered, but its patient-centered validity in the gastrectomy setting has not yet been demonstrated. The authors acknowledge this issue in the limitations and appropriately call for future prospective patient-reported anchoring; this distinction should remain consistent throughout the manuscript.

The analysis of the Charlson Comorbidity Index is interesting but should also be interpreted cautiously. Most patients had a very limited comorbidity burden, with more than three quarters of the cohort having an index-cancer-excluded CCI of 0 or 1 and only 11 patients having a score of 3 or higher. The absence of an independent association with DAOH30 may therefore reflect, at least in part, the restricted distribution of the exposure rather than demonstrating that baseline comorbidity is unrelated to postoperative recovery. The authors already acknowledge this limitation, but I would avoid giving the negative finding too much prominence in the Abstract and Conclusions. Functional status, nutritional status, frailty, sarcopenia, and preoperative physiological reserve may plausibly be more informative for this particular endpoint than a disease-weighted comorbidity index, and this distinction could be further emphasized.

Another issue affecting generalizability is the surgical approach. Almost all procedures in this series were performed using an open approach, whereas minimally invasive gastrectomy has become increasingly established in contemporary gastric cancer surgery. Since surgical approach may affect postoperative pain, recovery, length of stay, complication management, and discharge pathways, the observed distribution of DAOH30 may not be directly transferable to centers with predominantly laparoscopic or robotic practice. This limitation is mentioned by the authors but, given that DAOH30 is highly dependent on index hospitalization in this cohort, it is particularly important and should perhaps receive greater emphasis when discussing external validity.

Finally, the manuscript is methodologically very detailed, sometimes to an extent that affects readability. Several concepts—particularly the mathematical relationship between DAOH30 and length of stay, the rationale for not modeling readmission as an independent predictor, and some aspects of the statistical procedures—are explained repeatedly across the Methods, Results, and Discussion. The rigor is appreciated, but the manuscript would benefit from some condensation. Reducing repetition would make the clinical message clearer without compromising methodological transparency.

Overall, this is a thoughtful and methodologically sound study addressing a clinically meaningful postoperative outcome after gastrectomy. The main findings are appropriately interpreted and the limitations are largely recognized by the authors. Addressing the points above, particularly the clinical implications of complication management, the incremental value of DAOH30 over length of stay in this specific cohort, and the external generalizability of the findings, would further strengthen the manuscript.

Author Response

Response to Reviewer 1

We sincerely thank Reviewer 1 for the thoughtful, clinically focused, and constructive assessment. We are particularly grateful for the reviewer’s recognition that DAOH30 is an increasingly relevant perioperative outcome, that the endpoint was clearly defined, and that the statistical approach was generally appropriate.

Comment 1. Clinical interpretation of postoperative complications and evolving management of anastomotic leakage

The Reviewer noted that complication severity alone does not fully characterize the subsequent clinical trajectory. The type of complication and its management—particularly contemporary endoscopic management of upper gastrointestinal anastomotic leaks—may influence length of stay, reoperation, readmission, and DAOH30. The Reviewer recommended adding a brief discussion supported by PMID 41479933 and PMID 41008715.

Response:
We fully agree. We have expanded the Discussion to clarify that the hospital-time consequences of a major complication depend not only on its Clavien–Dindo grade but also on its phenotype and management strategy. A new paragraph specifically addresses contemporary management of upper gastrointestinal anastomotic leaks, including percutaneous drainage, reoperation, covered stenting, endoscopic vacuum therapy, internal drainage, and endoscopic defect-closure techniques.

The following text has been added:

“The hospital-time consequences of a major complication may depend not only on its Clavien–Dindo grade but also on its phenotype and management. This is particularly relevant for upper gastrointestinal anastomotic leaks, for which contemporary treatment may include percutaneous drainage, reoperation, covered stenting, endoscopic vacuum therapy, internal drainage, or endoscopic defect-closure techniques within a multidisciplinary strategy. These approaches may alter the duration of hospitalization, the need for additional intervention, and the risk of readmission, all of which are reflected in DAOH30. Because only five anastomotic leaks occurred, the present study could not compare management strategies, and leak-specific findings remain descriptive.”

We also added the two suggested contemporary reviews as References 30 and 31:

1. Birda CL, et al. Endoscopic management for gastrointestinal leaks, perforations, and fistulae: technical tips and outcomes. World J Gastrointest Endosc. 2025;17(12):111782.
2. Bernardi F, et al. Stents and emerging alternatives in upper gastrointestinal endoscopy: a comprehensive review. Diagnostics (Basel). 2025;15(18):2344.

We deliberately did not perform management-specific inferential analyses because only five anastomotic leaks occurred, and such comparisons would have been statistically unstable and potentially misleading.

Location in revised manuscript: Discussion, paragraph addressing complication phenotype and management; References 30–31.

 

Comment 2. Limited incremental information of DAOH30 beyond index length of stay

The Reviewer emphasized that 93.5% of patients survived without readmission and that no 30-day deaths occurred. Therefore, DAOH30 was mathematically equivalent to 30 minus index length of stay for most patients. The Reviewer recommended making it even clearer that the study demonstrates feasibility and known-groups discrimination rather than incremental clinical value over length of stay.

Response:
We fully agree and have made this conceptual limitation central to the revised interpretation.

First, the Abstract conclusion now explicitly states:

“DAOH30 differentiated complication-severity groups and quantified readmission-related hospital exposure, but its incremental information beyond index length of stay was limited in this cohort.”

Second, the opening paragraph of the Discussion now directly acknowledges that DAOH30 was largely determined by index postoperative length of stay:

“However, because no 30-day deaths occurred and 93.5% of patients survived without readmission, DAOH30 was largely determined by index postoperative length of stay. The findings therefore support DAOH30 as a summary of realized early hospital burden and known-groups discrimination in this cohort, rather than establishing independent prognostic information or incremental clinical value beyond conventional postoperative measures.”

Third, a separate early Discussion paragraph now emphasizes the structural nature of this relationship:

“In 115 of 123 patients, DAOH30 was algebraically equivalent to 30 minus index postoperative length of stay, and its additional information arose only from the eight readmission episodes; the mortality component of the endpoint was not observed. Accordingly, the present study demonstrates feasibility and known-groups discrimination but does not establish superiority, incremental predictive value, or net clinical benefit over length of stay.”

Finally, the Conclusion now describes DAOH30 as a complementary descriptive measure rather than a validated replacement for length of stay.

Location in revised manuscript: Abstract, Conclusions; Discussion, opening two paragraphs; Limitations; Conclusions.

 

Comment 3. Moderation of “patient-centered” terminology

The Reviewer correctly noted that no patient-reported outcome, quality-of-life measure, functional recovery measure, or patient-preference assessment was collected. The Reviewer recommended moderating the description of DAOH30 as a patient-centered endpoint.

Response:
We agree. We have revised our terminology throughout the manuscript to distinguish potential patient relevance from established patient-centered validity.

In the Introduction, the description has been changed from “flexible patient-centered outcomes” to:

“flexible perioperative outcomes with potential patient relevance.”

In the Discussion, DAOH30 is now described as:

“a complementary cohort-level measure for audits and perioperative trials,”

rather than as a validated patient-centered instrument.

We also explicitly state in the Limitations that DAOH30 does not directly measure:

• functional recovery;
• quality of life;
• patient preference;
• outpatient treatment burden; or
• readiness for adjuvant therapy.

The revised manuscript further states that convergent validity with patient-reported recovery, responsiveness, and a minimal clinically important difference have not yet been established.

The term “patient-centered” has been retained only where it forms part of the original title or terminology of cited publications.

Location in revised manuscript: Introduction, DAOH30 background paragraph; Discussion, clinical implications; Limitations; Conclusions.

 

Comment 4. Interpretation and prominence of the Charlson Comorbidity Index analysis

The Reviewer noted that the distribution of the index-cancer-excluded CCI was narrow, that most patients had scores of 0 or 1, and that the null finding should not be interpreted as evidence that baseline comorbidity is unimportant. The Reviewer also recommended avoiding excessive prominence of this negative result in the Abstract and Conclusions.

Response:
We fully agree.

The CCI analysis is now consistently labeled as exploratory in the Introduction, Methods, Results, and Discussion. In the Abstract Results, it is explicitly introduced as:

“In an exploratory analysis…”

The CCI finding has been removed from the Abstract Conclusions and from the main Conclusion paragraph so that it no longer competes with the principal findings regarding complication severity, readmission burden, and the relationship with length of stay.

We also revised the Discussion to emphasize that:

• 3% of patients had an index-cancer-excluded CCI of 0 or 1;
• only 8.9% had a score ≥3;
• no early deaths occurred;
• restricted exposure variation may have limited discrimination; and
• the null estimate should not be interpreted as evidence that baseline comorbidity is clinically unimportant.

We further clarify that the index-cancer-excluded CCI is not interchangeable with the original or age-adjusted CCI and that the descriptive categories used in the study are not externally validated risk strata.

Finally, the revised Discussion highlights frailty, nutritional status, sarcopenia, body composition, mobility, functional status, social support, and patient-reported recovery as potentially more informative constructs for future prospective research.

Location in revised manuscript: Abstract; Introduction, final background paragraph; Methods, Section 2.10.3; Results, Section 3.6; Discussion, exploratory CCI paragraph; Limitations.

 

Comment 5. Predominantly open surgical approach and external generalizability

The Reviewer emphasized that almost all operations were performed using an open approach and that the observed DAOH30 distribution may not be transferable to centers with predominantly laparoscopic or robotic practice.

Response:
We agree and have strengthened this limitation substantially.

The revised Limitations now state:

“118 of 123 procedures were open, and DAOH30 is sensitive to local intensive-care, nutritional, enhanced-recovery, discharge, and readmission pathways; the observed median of 21 days should therefore not be interpreted as a benchmark for gastrectomy or generalized to centers with predominantly laparoscopic or robotic practice.”

We also expanded the future-research paragraph to recommend validation across centers with:

• open, laparoscopic, and robotic surgical practice;
• different enhanced-recovery protocols;
• different intensive-care pathways;
• varying discharge criteria; and
• different readmission thresholds.

Location in revised manuscript: Discussion, clinical implications and future research; Limitations; Conclusions.

 

Comment 6. Repetition and readability

The Reviewer observed that the manuscript was methodologically detailed and that certain concepts—particularly the relationship between DAOH30 and length of stay, the reason for not modeling readmission as a predictor, and aspects of the statistical procedures—were repeated.

Response:
We appreciate this important recommendation and have substantially condensed the manuscript.

Specifically:

• the two general Discussion paragraphs describing DAOH and the broader literature were consolidated;
• the structural DAOH30–length-of-stay relationship was moved to the beginning of the Discussion and explained once as a central interpretive limitation;
• the readmission discussion was shortened and focused on the duration of hospital exposure;
• the explanation that readmission was not modeled as an explanatory predictor was reduced to a single concise statement;
• the clinical-implications and future-research paragraphs were merged;
• the Strengths paragraph was shortened; and
• repetitive language concerning superiority over length of stay was removed or consolidated.

The revised Discussion is now organized around three central findings:

1. the complication-severity gradient;
2. the duration of hospital exposure generated by the eight readmissions; and
3. the limited incremental information beyond index length of stay in this cohort.

Location in revised manuscript: Entire Discussion, particularly opening paragraphs, readmission paragraph, clinical-implications paragraph, and Strengths.

Reviewer 2 Report

Comments and Suggestions for Authors

This is a well-conducted and carefully presented retrospective study evaluating days alive and out of hospital within 30 days after curative gastrectomy for gastric adenocarcinoma. The manuscript addresses a clinically relevant aspect of postoperative recovery by examining a composite measure that incorporates index hospitalization, subsequent inpatient readmission, and early mortality. I found the manuscript generally clear, methodologically thoughtful, and appropriately cautious in its interpretation. In particular, the explicit definition of the 30-day observation window, the verification of hospitalization dates, the avoidance of data-driven dichotomization of the primary outcome, and the use of methods appropriate for its asymmetric distribution are important strengths. The analysis according to increasing Clavien–Dindo complication severity is clinically intuitive, and the progressive decrease in days alive and out of hospital from patients without complications to those with minor and major complications provides useful evidence of known-groups discrimination. The authors also appropriately avoid treating readmission as an independent predictor of an outcome of which readmission days are themselves a mathematical component. The use of median quantile regression, bootstrap confidence intervals, distributional effect sizes, and sensitivity analyses further strengthens the statistical presentation.

The most important issue concerns the incremental clinical information provided by days alive and out of hospital within 30 days in this particular cohort. There were no deaths within 30 days, and only eight of 123 patients were readmitted. Consequently, for the large majority of patients, the outcome was mathematically determined by index postoperative length of stay. The authors recognize this issue in the Discussion and correctly state that their findings do not establish superiority or incremental predictive value over length of stay. This is an important limitation and, in my view, should be made even more prominent when framing the principal message of the study. The observed discrimination across complication-severity groups is clinically meaningful, particularly for major complications, but part of this association is inherently expected because complications prolong hospitalization, and hospitalization duration directly reduces the primary outcome. Therefore, the findings should be interpreted primarily as evidence that this measure summarizes realized early hospital burden across clinically distinct postoperative trajectories rather than as evidence that it provides independent prognostic information beyond conventional postoperative measures. This distinction is particularly important because the strong inverse relationship with index length of stay is largely structural rather than an independent empirical association. I would encourage the authors to reinforce this point in both the Discussion and Conclusion so that readers do not interpret the findings as validation of the measure as a superior postoperative outcome. The current conclusion is already appropriately cautious, but this conceptual limitation deserves to remain central to the interpretation.

The small number of readmissions also deserves careful consideration. Although the eight readmissions generated 73 additional inpatient days and therefore illustrate an advantage of measuring duration rather than simply reporting a binary readmission rate, conclusions regarding the readmission component necessarily remain descriptive. The causes of readmission are clinically plausible, but subgroup interpretation should remain restrained because the number of events is very small. Similarly, only five anastomotic leaks occurred, which makes the corresponding comparison highly imprecise. The absence of any 30-day mortality is also important because one of the conceptual advantages of days alive and out of hospital is its ability to incorporate mortality into the same scale as hospitalization. In the present study, that component of the construct was not actually tested. I suggest making this especially clear when discussing the potential generalizability of the findings to higher-risk populations, emergency surgery, older or frailer populations, or centers with different postoperative mortality and readmission profiles. The manuscript already acknowledges these limitations, but their implications for the degree of validation that can be claimed could be emphasized further. 

External validity is another relevant consideration. This is a single-center study, and 118 of the 123 procedures were performed using an open surgical approach. Contemporary gastric cancer surgery increasingly includes laparoscopic and robotic approaches in many centers, with potentially different lengths of stay, enhanced recovery pathways, discharge criteria, complication profiles, and thresholds for readmission. Because days alive and out of hospital is directly influenced by these health-system and perioperative practices, the absolute values reported here should not be interpreted as broadly applicable reference values. The authors appropriately call for multicenter validation, and I agree strongly with this recommendation. It would be useful to state even more explicitly that the median value observed in this cohort should not currently be considered a benchmark for gastrectomy. Future validation should ideally include centers with different surgical approaches and enhanced recovery protocols and should assess whether between-center differences in discharge practices materially influence the distribution of this outcome. The authors themselves recognize that discharge and readmission practices may vary across health care systems and that the predominantly open cohort may not reflect centers with extensive minimally invasive practice; this limitation is particularly important for the eventual use of the measure in quality comparisons.

 

I also suggest providing slightly more clinical context regarding perioperative pathways during the study period. The cohort spans January 2020 through April 2026, a relatively long interval during which surgical techniques, postoperative management, nutritional support, discharge practices, and enhanced recovery protocols may have changed. Calendar year was appropriately included in the preoperative Charlson association model, which is reassuring, but readers would benefit from knowing whether there were meaningful changes in the institutional perioperative pathway during these years. This is particularly relevant for an outcome that is strongly dependent on hospitalization duration. If enhanced recovery protocols, discharge criteria, routine intensive care practices, or minimally invasive surgery changed during the study period, these changes could affect the observed distribution independently of patient characteristics. If perioperative management remained essentially stable, a brief statement to that effect would improve interpretability.

 

The analysis of comorbidity is interesting but should remain clearly exploratory. The index-cancer-excluded Charlson Comorbidity Index was not independently associated with days alive and out of hospital, but the distribution of comorbidity burden was relatively narrow, with most patients having low scores, and the absence of mortality further limits discrimination among higher-risk patients. Therefore, the lack of statistical association should not be interpreted as evidence that preoperative comorbidity is unimportant for postoperative recovery. The authors handle this issue appropriately in the Discussion by noting that the Charlson framework does not capture several potentially important dimensions such as mobility, nutrition, sarcopenia, independence, social support, and resilience. This is an important point that could be developed slightly further. In future prospective studies, incorporation of frailty, nutritional status, body composition, functional status, and patient-reported recovery would probably provide a more comprehensive evaluation of factors associated with time spent alive outside the hospital. The authors should also continue to emphasize that the modified index used in this study is not interchangeable with the original or age-adjusted Charlson Comorbidity Index and that the descriptive categories used here are not externally validated risk strata.

 

From a statistical perspective, the analyses are generally appropriate and unusually transparent for a retrospective cohort of this size. The use of nonparametric comparisons, an ordered trend test, adjusted pairwise comparisons, bootstrap confidence intervals, effect-size estimates, and median quantile regression is well justified by the distribution of the outcome. Nevertheless, the relatively modest sample size and the number of covariates included in the adjusted models should be kept in mind when interpreting the precision and stability of the regression estimates. I appreciate that the authors explicitly describe the regression analysis as explanatory rather than predictive and that they avoided automated variable selection. I would suggest maintaining the emphasis on effect estimates and confidence intervals rather than statistical significance throughout the Discussion, particularly for exploratory comparisons with few events. The manuscript already generally follows this principle, but consistent emphasis on magnitude and precision rather than dichotomous statistical significance would further strengthen the presentation.

 

The figures and tables are informative, although some simplification could improve readability. The table presenting supportive clinical comparisons contains several statistical measures simultaneously, including medians, median differences, bootstrap confidence intervals, probability values, and Cliff’s delta. All of these measures are defensible, but the density of information may make the principal clinical message less immediately apparent. The authors could consider whether every statistical measure needs to remain in the main table or whether some details could be transferred to the Supplementary Materials. The figure showing the patient-level temporal decomposition of readmissions is conceptually valuable because it illustrates something that a simple readmission proportion cannot show: the timing and duration of additional hospital exposure. I would retain this figure, but the legend should remain sufficiently self-contained for readers to understand exactly how index hospitalization, time outside the hospital, and readmission contribute to the 30-day measure. The figure legends are otherwise detailed and informative.

 

The Discussion is comprehensive and appropriately acknowledges most of the limitations of the study. It could nevertheless be somewhat more focused. Several paragraphs revisit the conceptual advantages of days alive and out of hospital and its relationship with conventional postoperative outcomes. Some condensation would improve the balance between interpretation of the present data and broader discussion of the measure. The most clinically relevant findings are the clear gradient associated with complication severity, the substantial duration of hospital exposure generated by a small number of readmissions, and the limited incremental information beyond index length of stay in a cohort without early mortality and with few readmissions. Keeping these three points at the center of the Discussion would make the manuscript more concise and sharpen its contribution.

 

Finally, I agree with the authors that future work should evaluate this outcome prospectively and across multiple institutions. In addition to comparing 30-day and longer observation windows, future studies should examine its relationship with patient-centered measures of recovery, nutritional recovery, functional status, postoperative quality of life, return to usual activities, and readiness to initiate adjuvant treatment. Establishing convergent validity, responsiveness to perioperative interventions, and a clinically meaningful difference would be necessary before this measure could be confidently used for benchmarking or individual patient counseling. The present study provides useful preliminary evidence regarding known-groups discrimination and illustrates the ability of the measure to quantify readmission-related hospital exposure, but it should be viewed as an initial single-center evaluation rather than definitive validation. The authors already acknowledge that responsiveness, convergent validity with patient-reported recovery, and a minimal clinically important difference were not established, which is an appropriately cautious assessment.

 

 

Author Response

Response to Reviewer 2

We sincerely thank Reviewer 2 for the detailed, balanced, and highly constructive review. We are grateful for the reviewer’s recognition that the study was carefully presented, methodologically thoughtful, and appropriately cautious, and that the statistical analyses were unusually transparent for a retrospective cohort of this size.

Comment 1. DAOH30 should be framed as a summary of realized hospital burden rather than as an independently prognostic or superior outcome

The Reviewer considered the limited incremental clinical information beyond index length of stay to be the most important issue and recommended making this distinction central to the Discussion and Conclusion.

Response:
We fully agree and have substantially revised the framing of the manuscript.

The revised Discussion now opens by stating that DAOH30 was largely determined by index postoperative length of stay because no 30-day deaths occurred and only eight patients were readmitted. We explicitly state that the findings support DAOH30 as:

“a summary of realized early hospital burden and known-groups discrimination,”

rather than as evidence of independent prognostic information or incremental clinical value.

A new paragraph immediately following the opening Discussion paragraph clarifies that DAOH30 was algebraically equivalent to 30 minus index postoperative length of stay in 115 of 123 patients and that the endpoint’s additional information arose only from the eight readmission episodes.

The revised Conclusion similarly states:

“It should therefore be considered a complementary descriptive measure rather than a validated substitute for length of stay or a benchmark for gastrectomy outcomes.”

We believe these revisions place the structural relationship with length of stay at the center of the interpretation, as recommended.

Location in revised manuscript: Abstract conclusion; Discussion, opening two paragraphs; Limitations; Conclusions.

 

Comment 2. Small numbers of readmissions and anastomotic leaks, and absence of 30-day mortality

The Reviewer noted that readmission and leak analyses should remain descriptive and that the mortality-sensitive component of DAOH30 was not tested. The Reviewer also recommended greater caution regarding generalizability to higher-risk, emergency, older, or frailer populations.

Response:
We agree.

The revised Discussion now states that:

• the eight readmissions generated 73 additional inpatient days;
• the duration-based readmission analysis is descriptive;
• causes and subgroups should be interpreted cautiously;
• only five anastomotic leaks occurred;
• management-specific leak comparisons were not attempted; and
• the mortality component of DAOH30 was not observed in this cohort.

The Limitations have also been expanded to state:

“Readmission and leak analyses are therefore descriptive, and the mortality-sensitive component of DAOH30 was not tested.”

We further added:

“Accordingly, these findings may not generalize to emergency surgery or to older, frailer, or higher-mortality populations in whom the mortality component of DAOH30 may contribute more substantially.”

These revisions clarify that the present study is an initial evaluation in a selected elective surgical cohort and not a definitive validation across all gastrectomy populations.

Location in revised manuscript: Discussion, readmission and leak-management paragraphs; Limitations.

 

Comment 3. Predominantly open surgical practice and the risk of interpreting the observed median as a benchmark

The Reviewer emphasized that 118 of 123 operations were open and that the median DAOH30 of 21 days should not be interpreted as a broadly applicable reference value or benchmark.

Response:
We fully agree.

The revised manuscript now explicitly states that:

“the observed median of 21 days should therefore not be interpreted as a benchmark for gastrectomy or generalized to centers with predominantly laparoscopic or robotic practice.”

The future-research paragraph also calls for multicenter validation across open, laparoscopic, and robotic practices and across centers with different enhanced-recovery, intensive-care, discharge, and readmission pathways.

The Conclusion now avoids any benchmark implication and describes DAOH30 as a complementary descriptive measure pending external validation.

Location in revised manuscript: Discussion, clinical implications and future research; Limitations; Conclusions.

 

Comment 4. Perioperative pathway changes during the 2020–2026 study period

The Reviewer recommended clarifying whether meaningful changes occurred in enhanced-recovery protocols, intensive-care practice, nutritional support, discharge criteria, or minimally invasive surgery during the study period.

Response:
We appreciate this important point. We reviewed the institutional perioperative pathway and added the following clarification to the Methods:

“Throughout the study period, the principal institutional gastrectomy pathway—including routine postoperative intensive care monitoring, nutritional advancement, and clinical discharge criteria—remained broadly stable, and no formal pathway change was implemented that was expected to systematically alter postoperative length of stay.”

Calendar year was also retained in the preoperative association model to account, as far as possible, for residual temporal variation.

Location in revised manuscript: Materials and Methods, Section 2.6, Surgical and Perioperative Care.

 

Comment 5. Exploratory interpretation of the Charlson analysis

The Reviewer recommended maintaining the CCI analysis as exploratory, emphasizing the narrow score distribution, and discussing frailty, nutritional status, body composition, functional status, and patient-reported recovery as potentially more informative constructs.

Response:
We fully agree.

The revised manuscript now:

• labels the CCI analysis as exploratory throughout;
• reports the restricted score distribution;
• explicitly states that the null estimate does not demonstrate that comorbidity is unimportant;
• clarifies that the modified score is not interchangeable with the original or age-adjusted CCI;
• states that the categorical score groups are not externally validated risk strata; and
• highlights frailty, nutritional status, sarcopenia, body composition, mobility, functional status, social support, and patient-reported recovery as priorities for future prospective studies.

The CCI result has also been removed from the Abstract Conclusions and the principal Conclusion paragraph.

Location in revised manuscript: Abstract; Introduction; Methods, Section 2.10.3; Results, Section 3.6; Discussion; Limitations.

 

Comment 6. Precision and stability of the adjusted regression estimates

The Reviewer noted that the modest cohort size and number of covariates should be considered when interpreting the precision and stability of the regression estimates and recommended emphasizing effect estimates and confidence intervals rather than dichotomous statistical significance.

Response:
We agree.

The revised manuscript continues to describe the regression analysis as explanatory rather than predictive and explicitly states that:

• no automated predictor selection was used;
• no stepwise procedure, nomogram, or machine-learning model was developed;
• bootstrap confidence intervals constituted the primary inferential measure;
• model-based p values were reported separately; and
• secondary and sensitivity analyses were interpreted principally according to effect estimates and confidence intervals.

We also added the following limitation:

“Given the modest cohort size and the number of covariates included, adjusted regression estimates should be interpreted as explanatory and may remain imprecise despite the consistency of the sensitivity analyses.”

In the Discussion, interpretations have been revised to emphasize the magnitude and precision of estimates rather than statistical significance alone.

Location in revised manuscript: Materials and Methods, Section 2.10.3; Discussion, complication-severity interpretation; Limitations.

 

Comment 7. Simplification of tables and preservation of the readmission timeline figure

The Reviewer suggested simplifying the supportive clinical-comparisons table and retaining the patient-level readmission timeline figure with a sufficiently self-contained legend.

Response:
We agree and have revised the presentation accordingly.

In the main Table 2, Panel B has been simplified by retaining:

• group sizes;
• group medians and interquartile ranges;
• median differences;
• bootstrap confidence intervals; and
• unadjusted Mann–Whitney p values.

The Cliff’s δ column was removed from the main supportive-comparisons table to reduce visual density. Detailed effect-size estimates remain available in the Results and in Supplementary Table S2.

We also retained Figure 4, as recommended. Its legend now explicitly explains:

• the operation date as postoperative day 0;
• the temporal positions of index hospitalization, time alive outside hospital, and readmission;
• the meaning of the labeled readmission intervals;
• the individual DAOH30 values;
• the total 73 additional inpatient days; and
• the median readmission-related decrement of 9.5 days.

Location in revised manuscript: Table 2; Supplementary Table S2; Figure 4 and its legend.

 

Comment 8. Condensation and refocusing of the Discussion

The Reviewer recommended reducing repetition and keeping the Discussion centered on the complication-severity gradient, readmission-related hospital exposure, and limited incremental information beyond index length of stay.

Response:
We fully agree and reorganized the Discussion accordingly.

The revised sequence is now:

1. principal findings and limited incremental value beyond length of stay;
2. structural relationship between DAOH30 and length of stay;
3. concise context from the broader DAOH literature;
4. complication-severity gradient;
5. complication phenotype and contemporary leak management;
6. readmission duration and hospital exposure;
7. exploratory CCI findings;
8. clinical implications and future research;
9. strengths; and
10.  

Overlapping DAOH-literature paragraphs were consolidated, the readmission discussion was shortened, clinical implications and future research were merged, and the Strengths paragraph was condensed.

Location in revised manuscript: Entire Discussion.

 

Comment 9. Prospective multicenter validation and patient-reported anchoring

The Reviewer recommended future evaluation across multiple institutions and comparison of DAOH30 with longer windows, functional recovery, nutritional recovery, quality of life, return to usual activities, and readiness for adjuvant therapy.

Response:
We agree and expanded the future-research recommendations.

The revised Discussion now recommends that future studies:

• include open, laparoscopic, and robotic practice;
• include centers with different enhanced-recovery, intensive-care, discharge, and readmission pathways;
• evaluate DAOH30 and DAOH90 in parallel;
• examine quality of life;
• examine functional and nutritional recovery;
• assess return to usual activities;
• assess readiness for adjuvant treatment; and
• establish convergent validity, responsiveness, and a clinically meaningful difference.

The Conclusion now calls for:

“Multicenter prospective validation across diverse surgical and perioperative pathways, with patient-reported anchoring.”

Location in revised manuscript: Discussion, clinical implications and future research; Limitations; Conclusions.

Reviewer 3 Report

Comments and Suggestions for Authors

The study is well-constructed and scientifically sound. This is a retrospective study at a single center investigating patients with resectable gastric adenocarcinoma after gastrectomy. The proposed DAOH30 is useful for evaluation. However, I cannot find the figures in the manuscript. The missing figures in the manuscript hamper the reviewers' ability to evaluate it. Please attach the missing figures.

 

1. The submitted PDF contains only the figure legends, and the figure images (Figures 1–4) are missing. Please provide the complete set of high-resolution figures or an updated PDF that embeds them.
2. Thin spaces should be used before and after the following mathematical symbols: <, >, ≥, and =.
3. The authors indicate that among 123 patients, no 30-day deaths occurred; eight patients were readmitted; and five anastomotic leaks occurred. DAOH30 correlated very strongly and inversely with index postoperative length of stay. Consequently, DAOH30 was algebraically equivalent to 30 minus index length of stay for the large majority of patients. The incremental value of DAOH30 came from the eight readmissions. The authors should emphasize this relationship more clearly.
4. The authors compared DAOH30 across three groups of postoperative complications. Median DAOH30 decreased from 22 days without complications to 21 days after Clavien–Dindo grade I–II complications and 17 days after grade ≥ III complications. The authors indicated this as known-group discrimination. However, the severity of complications and DAOH30 are mechanistically related. All severe complications increase the hospital days and lower DAOH30. Thus, the observed gradient is useful evidence of the expected clinical association. The authors should not interpret it as evidence that DAOH30 provides incremental information. It would be better to clarify this distinction in the discussion and conclusions.
5. Could you explain who assigned the Clavien-Dindo grades and how?
6. If readmissions occur outside the study institution, how do the authors acknowledge that external readmissions may have been missed?

Author Response

Reviewer 3 — General Response

Response:
We sincerely thank the Reviewer for the careful and constructive evaluation of our manuscript and for recognizing that the study is well constructed and scientifically sound. We are grateful for the specific recommendations regarding figure presentation, mathematical typography, the structural relationship between DAOH30 and length of stay, interpretation of the complication-severity gradient, retrospective assignment of Clavien–Dindo grades, and ascertainment of readmissions outside the institutional network. We have addressed each point as detailed below.

Comment 1 — Missing figures

Response:
We apologize that the submitted PDF contained the figure legends without the corresponding figure images. Figures 1–4 have now been embedded in the revised manuscript at their first relevant citations, and a new complete PDF has been generated and checked. The full set of high-resolution figures has also been uploaded separately as supplementary submission files.

Comment 2 — Thin spaces around mathematical symbols

Response:
We thank the Reviewer for identifying this typographic inconsistency. Thin spaces have now been applied before and after the mathematical symbols <, >, ≥, and = throughout the Abstract, main text, tables, table notes, figure legends, and Supplementary Materials. Mathematical equations have also been formatted using the Word Equation Editor.

Comment 3 — DAOH30 and index length of stay

Response:
We fully agree. The revised manuscript now emphasizes more explicitly that DAOH30 was algebraically equivalent to 30 minus index postoperative length of stay in 115 of 123 patients, that no 30-day deaths occurred, and that the endpoint’s additional information arose from only eight readmission episodes. The Abstract Conclusion, opening Discussion paragraphs, Limitations, and Conclusion have been revised to clarify that the present study does not establish incremental predictive value, net clinical benefit, or superiority over length of stay.

Comment 4 — Mechanistic relationship with complication severity

Response:
We agree with the Reviewer that the observed complication-severity gradient is mechanistically expected because more severe complications generally prolong hospitalization, which directly reduces DAOH30. We have therefore clarified in the Discussion and Conclusion that this gradient represents the expected clinical association between postoperative morbidity and hospital exposure and should not be interpreted as evidence that DAOH30 provides incremental information beyond index length of stay or its constituent components.

Comment 5 — Assignment of Clavien–Dindo grades

Response:
We sincerely thank the Reviewer for requesting this important methodological clarification. Postoperative events occurring within 30 days of the index operation were identified retrospectively from source-verified progress notes, operative and interventional reports, intensive-care documentation, and discharge summaries. The Clavien–Dindo grades were assigned by two surgical investigators, A.O. and G.G. Any uncertain or discordant classifications were resolved by consensus with a third surgical investigator, A.B. For patients who experienced more than one postoperative event, the highest Clavien–Dindo grade occurring within the 30-day postoperative period was retained for analysis.

To improve methodological transparency, we expanded Section 2.7 of the revised manuscript as follows:

“Postoperative events occurring within 30 days of the index operation were identified through review of source-verified progress notes, operative and interventional reports, intensive-care documentation, and discharge summaries and were classified according to the Clavien–Dindo system [19,20]. Clavien–Dindo grades were assigned retrospectively by two surgical investigators, and classification disagreements were resolved by consensus with a third surgical investigator. For each patient, the highest Clavien–Dindo grade occurring within 30 days was retained for analysis. Grade III or higher events were defined as major postoperative complications.”

Location in the revised manuscript: Materials and Methods, Section 2.7, Postoperative Complications and Readmission.

Comment 6 — External readmissions

Response:
We agree that readmissions occurring outside the institutional network may not have been completely captured. We have clarified in the Methods that external admissions could be identified only when documented during Surgical Oncology follow-up and could not otherwise be independently verified. We have also strengthened the Limitations to state explicitly that a missed external readmission would have resulted in overestimation of DAOH30.

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

No further comments from me. The authors answered all questions and comments.

Author Response

Comment: No further comments from me. The authors answered all questions and comments.

Response: We sincerely thank the reviewer for the positive reassessment of our manuscript and for confirming that the previous comments have been satisfactorily addressed.

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for your responses. The authors address the reviewers' comments properly. The revised manuscript has improved. 

1. Please clarify whether the directional hypothesis underlying the one-sided Jonckheere–Terpstra test was prespecified. If it was not prespecified, a two-sided sensitivity analysis would be helpful.
2. Please consider consistently describing the CCI result as no statistically significant association was detected.

Author Response

Comment 1: Please clarify whether the directional hypothesis underlying the one-sided Jonckheere–Terpstra test was prespecified. If it was not prespecified, a two-sided sensitivity analysis would be helpful.

Response: We thank the reviewer for this important statistical comment. The directional hypothesis was not formally prespecified in the original study protocol. It was based on the clinical ordering of the Clavien–Dindo categories and the expectation that increasing complication severity would be accompanied by decreasing DAOH30. To avoid relying solely on one-sided inference, we performed an additional two-sided permutation sensitivity analysis based on the absolute deviation of the observed Jonckheere–Terpstra statistic from its permutation null expectation, using the same 50,000 label permutations. The result was unchanged (two-sided permutation sensitivity p < 0.001). We clarified the prespecification status and analytical procedure in the Statistical Analysis section and added the sensitivity result to the Abstract, Results, Table 2, and Figure 3 legend.

Comment 2: Please consider consistently describing the CCI result as no statistically significant association was detected.

Response: We thank the reviewer and agree that a statistically nonsignificant result should not be interpreted as evidence of no association. We therefore revised the Abstract, Results, sensitivity-analysis summary, and Discussion to state consistently that no statistically significant association was detected between the index-cancer-excluded CCI and DAOH30. The corresponding effect estimates, confidence intervals, and p values were retained to preserve information regarding magnitude and precision.