Integrated Analysis of Parenchymal and Vascular HRCT Patterns with Circulating Biomarkers in Severe COVID-19 Pneumonia

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Summary: Authors of this work investigated integrated radiologic, immuno-inflammatory and endothelial injury biomarkers to determine whether radiological patterns of COVID-19 pneumonia, with a particular focus on vascular abnormalities, are useful in categorizing the patients in clinical-oriented subgroups and in establishing the prognosis.

Major Comments: 1. I would like authors to mention the power analysis of their study considering relatively small number of studied cases.

2. Authors did not mention if elevated cytokines could be related to another health-related issues of the patients.

 

Author Response

1. I would like authors to mention the power analysis of their study considering relatively small number of studied cases.

 

We thank the Reviewer for this important methodological comment.

Given the monocentric and exploratory nature of the study, a formal a priori power analysis was not performed. The sample size was determined by the availability of patients meeting strict inclusion criteria during the pre-vaccination phase of the COVID-19 pandemic.

We acknowledge that the relatively limited number of cases may affect statistical power, and this has now been explicitly stated as a limitation of the study. Nevertheless, the sample size is comparable to that of previously published imaging-based studies focusing on detailed HRCT pattern stratification and biohumoral correlations in COVID-19 patients.

Importantly, despite the limited sample size, several associations reached statistical significance, supporting the biological plausibility and internal consistency of the observed findings. These results should therefore be interpreted as hypothesis-generating and warrant confirmation in larger, prospective cohorts, which we now explicit in the discussion session.

 

2. Authors did not mention if elevated cytokines could be related to another health-related issues of the patients.

 

We thank the Reviewer for highlighting this relevant point. In response to the comments from this and the other Reviewers, we have expanded the statistical analysis of the study by implementing multivariable models to better account for potential confounding factors and to strengthen the robustness of the main findings.

However, the specific issue raised here—namely the independent effect of individual comorbidities or pre-existing health-related conditions on cytokine levels or CT features—was not the primary objective of the present study and could not be reliably addressed with the available dataset. The study cohort included a wide and heterogeneous spectrum of comorbidities, each represented by a small number of patients, which precluded meaningful stratified analyses or condition-specific multivariable modeling without a substantial risk of model instability and overfitting.

For this reason, comorbidities were accounted for in the multivariable analyses as a global burden (i.e., number of comorbid conditions at admission), rather than being examined individually. This approach is consistent with the study’s primary aim, which was to investigate the relationship between radiological severity, inflammatory burden, and clinical outcomes, rather than to disentangle the contribution of specific pre-existing conditions to cytokine expression or CT phenotypes.

Importantly, existing literature consistently reports elevated cytokine levels in both COVID-related and non-COVID ARDS largely irrespective of baseline comorbidities or the use of anti-inflammatory or immunomodulatory therapies, suggesting that the acute inflammatory response is predominantly driven by disease severity rather than by pre-existing conditions.

Reviewer 2 Report

Comments and Suggestions for Authors

In this document, an analysis has been made of the relationship between circulating biomarkers and HRCT patterns in hospitalized patients diagnosed with COVID-19. Although it is addressing an important and timely subject, there are significant defects in the methodology used throughout the research, which considerably affect the accuracy of the findings and their relevance to the larger population. An extensive, specialized critique can be found in the following section. 1. According to the authors, the most significant methodological deficiency in this analysis is the lack of a matching control group. The authors’ claims that the reported radiographic patterns (particularly TIB / VEP) and biomarker profiles are specific to COVID-19 pathophysiology can only be substantiated by comparison to patients with: a. Non-COVID-19 ARDS (such as severe influenza, bacterial pneumonia, or sepsis). b. Hypoxemia from non-infectious causes (for example cardiogenic pulmonary edema-specific). In the absence of these comparisons, the authors’ claim that the results "may inform... other severe respiratory infections" would remain purely conjectural. Therefore, lacking these comparisons also significantly weakens the study's ability to specifically attribute the observed results to SARS-COV-2. 2. Because this study only takes place in one location, there could be some bias introduced within this single center as it relates to local protocols, expertise with imaging studies as well as demographics of their patients. This may severely limit external generalizability. Additionally, the chosen entry criteria (and their associated definitions of respiratory failure; i.e., PaO₂ < 60 mmHg) would lead to a sampling bias that would limit entry to moderate and severe cases only; therefore, any identified relationships may not apply to the general population of patients with COVID-19 or patients with mild disease thus inhibiting utility when developing strategies for early intervention. 3. The vascular score represents a crude semi-qualitative scoring system on a yes/no (present or absent) basis per lobe. The following are some of the limitations of this scoring system. a. There is no way to assess how much vascularity a lobe has or how severely vascularized it is. The same number of TIBs is assigned to a single area of TIB vascularization and a large area of VEP vascularization. b. There has been no independent validation of the COVID-19 vasculopathy scale used here. Results can be made more objective and reproducible with a quantitative scoring system based on: 1) vessel-to-bronchus ratio; 2) computed tomography angiography parameters; or 3) a textural analysis of the vascular tree. 4. Even though they employed two separate radiologists, the authors did not present inter-rater reliability data (e.g., Intraclass Correlation Coefficient [ICC] or Cohen's Kappa) for identifying and scoring subtle patterns such as TIB and VEP. The lack of these metrics makes it impossible to determine if the primary imaging results will be reliable and reproducible. 5. While this is a significant limitation, a major technical issue is separating primary vascular pathology from secondary damage caused by compression or pruning in the surrounding lung due to consolidation or atelectasis. Advanced imaging modalities, such as dual-energy CT for perfusion mapping, could help define microvascular function independently of parenchymal opacities. 6. 6. For the purposes of creating a high and low group of PS and VS values using the median value of PS and VS is an arbitrary cut-point determined by the data collected, which will reduce statistical power and increase the chance of finding false-positive relationships. Using PS and VS values in their continuous form or establishing clinical thresholds for their interpretation would be a more valid method of analysis. 7. The authors have only completed a univariate analysis, so one critical issue remains unanswered: Is having a high PS score at the time of ICU admission an independent predictor of ICU admission after controlling for age, comorbidities at the time of admission, and total inflammatory burden before admission? A multivariable logistic regression analysis that controls for these critical confounding variables will demonstrate that PS score can be considered an independent predictor of ICU admission and is not simply an association with overall severity of disease at the time of admission. 8. The study has an overall sample size of 84 and tests for multiple biomarkers (over 10). The lack of using corrections for multiple comparisons (e.g., Bonferroni corrections) may lead to a very high risk of Type I error (false positives) through the increased number of univariate tests run. Additionally, the subgroup analyses performed based on VS may involve lower sample sizes (numbers of patients), and with such smaller sample sizes, the study is at risk of being underpowered to find statistically significant differences in very infrequent outcomes (hard outcomes) such as mortality. 9. The discussion cites the inability to find an association between VS and clinical outcomes mainly because imaging was performed at only one time point; however, while this explanation is reasonable, the study should also have included a more critical, equally valid explanation for the lack of association. It is plausible the vascular signs seen on HRCT imaging represent an epiphenomenon, that is, they are secondary to the reaction of the peri-vascular inflammatory response to VS without any direct causal relationship to the deterioration seen clinically due to the VS. This hypothesis should also have been discussed and presented as a possible explanation for the study's findings. 10. The data for this study were collected at two time points - T1 and T2 - but due to the nature of the data collection methods used, it is not possible to perform a true longitudinal analysis of these data. Therefore, the current analyses only provide a limited understanding of CT score changes and their relationship to biomarker level changes and final outcome measures, as well as how the CT scores change over time. 11. As stated regarding the design flaws of this study, the assertion that the results of this study can be generalized to other respiratory infections are not supported by the research findings and represent an unwarranted extrapolation. This section should have concentrated on developing hypotheses to be tested in future comparative studies using other datasets rather than making generalizations based on this dataset's results. The purpose of this study was to explore an important topic and create new hypotheses through the study of participants at a single location. Due to the single-center nature of the study and other limitations in the design, such as lack of a control group, lack of operationalizing key vascular variables, and major limitations in conducting statistical analyses, the study results should be considered preliminary until the results are confirmed by others. The study provides meaningful associations between the potential for PS and various outcomes; however, there is also significant confounding in the results. Although the biological mechanisms being studied are fascinating, they do not establish a direct connection between the findings related to vascularity and clinical significance. The main contribution of this manuscript is the development of a conceptual and methodological framework for integrating various approaches to research. In order to consider future potential clinical applications of the study results, the findings must be confirmed through prospective, multicenter studies with appropriate control groups, quantitative imaging biomarkers, and adequate multivariable statistical models.

Author Response

1. According to the authors, the most significant methodological deficiency in this analysis is the lack of a matching control group. The authors’ claims that the reported radiographic patterns (particularly TIB / VEP) and biomarker profiles are specific to COVID-19 pathophysiology can only be substantiated by comparison to patients with: a. Non-COVID-19 ARDS (such as severe influenza, bacterial pneumonia, or sepsis). b. Hypoxemia from non-infectious causes (for example cardiogenic pulmonary edema-specific). In the absence of these comparisons, the authors’ claim that the results "may inform... other severe respiratory infections" would remain purely conjectural. Therefore, lacking these comparisons also significantly weakens the study's ability to specifically attribute the observed results to SARS-COV-2.

 

We thank the Reviewer for this important observation. We agree that the inclusion of matched control groups (e.g., non-COVID-19 ARDS or non-infectious hypoxemia) would be ideal to formally assess disease specificity.

However, due to the study design and the clinical setting of the pre-vaccination pandemic phase, inclusion of control cohorts with systematic HRCT acquisition and extended biohumoral profiling was not feasible.

Accordingly, we now have avoided claims of absolute specificity and clarified that our findings describe imaging–biomarker associations within a well-defined COVID-19 population. Any extension of these results to other forms of severe respiratory failure is acknowledged as speculative and is now more explicitly discussed in the Limitations section.

 

2. Because this study only takes place in one location, there could be some bias introduced within this single center as it relates to local protocols, expertise with imaging studies as well as demographics of their patients. This may severely limit external generalizability. Additionally, the chosen entry criteria (and their associated definitions of respiratory failure; i.e., PaO₂ < 60 mmHg) would lead to a sampling bias that would limit entry to moderate and severe cases only; therefore, any identified relationships may not apply to the general population of patients with COVID-19 or patients with mild disease thus inhibiting utility when developing strategies for early intervention.

 

As stated in the new version of the Limitations section, the single-centre nature of the study is recognised a potential source of bias which may limit the generalizability of the results. The inclusion of severe cases only reflects the clinical context of the pre-vaccination phase during the pandemic phase, in which HRCT was systematically acquired in hospitalized patients admitted to intensive care units or pulmonology wards, while it was not routinely indicated in milder cases without risk of progression. These aspects are now clearly acknowledged as limit to the applicability of our findings to patients with milder disease or to early-intervention strategies. More specifically, in this study we anayzed a population of patients affected by acute respiratory failure associated with COVID-19. Per definition, acute hypoxemic respiratory failure is characterized by a PaO₂ < 60 mmHg. Our study population clearly includes patients with moderate/severe COVID as per WHO definition, given the inclusion criteria (PaO₂ levels, hospitalization, respiratory care and ICU involvement in recruitment). For this reason, we chose the PaO₂ criteria as an inclusion factor. We totally agree with the reviewer that since our population is made of patients with acute respiratory failure, our findings are not applicable to all patients with COVID-19 (in the absence of respiratory failure).

 

3. The vascular score represents a crude semi-qualitative scoring system on a yes/no (present or absent) basis per lobe. The following are some of the limitations of this scoring system. a. There is no way to assess how much vascularity a lobe has or how severely vascularized it is. The same number of TIBs is assigned to a single area of TIB vascularization and a large area of VEP vascularization. b. There has been no independent validation of the COVID-19 vasculopathy scale used here. Results can be made more objective and reproducible with a quantitative scoring system based on: 1) vessel-to-bronchus ratio; 2) computed tomography angiography parameters; or 3) a textural analysis of the vascular tree.

 

We thank the Reviewer for this detailed methodological comment and acknowledge that the proposed vascular score represents a semi-quantitative, lobe-based approach. We agree that fully quantitative methods could, in principle, provide a more granular assessment of vascular involvement. However, segment-based or fully quantitative scoring of peripheral vascular abnormalities such as TIB and VEP would be extremely challenging in routine HRCT, particularly in the absence of dedicated software or artificial intelligence tools. Such an approach would likely reduce reproducibility and inter-reader agreement, thereby limiting the generalizability of the results.

Importantly, a similar semi-quantitative approach to CT microvascular signs in COVID-19 has already been applied and validated by our group in a high–impact peer-reviewed journal (Dalpiaz et al., Radiol Med, 2022), supporting the clinical feasibility and relevance of this methodology. In the present study, TIB and VEP were assessed separately and subsequently integrated into a composite vascular score in order to better capture the spectrum of vascular involvement.

With respect to vessel-to-bronchus ratio measurements, the abnormalities described in the manuscript predominantly involve peripheral branches, where strict quantitative assessment is inherently unreliable. Moreover, the definition of VEP itself relies on a comparative visual assessment rather than on absolute measurements. Automated quantitative tools were not employed, as their performance is known to be suboptimal in the presence of surrounding parenchymal consolidation or ground glass opacities, which are typical in COVID-19 pneumonia.

Regarding CT angiography–based metrics, HRCT was an inclusion criterion of the study, whereas CT angiography is not routinely performed in moderate-to-severe COVID-19 and was therefore not available. Finally, we agree that vascular textural analysis represents an interesting future direction; however, its applicability in patients with extensive parenchymal abnormalities remains to be demonstrated, as vascular changes in COVID-19 are rarely isolated and are closely intertwined with parenchymal disease.

 

4. Even though they employed two separate radiologists, the authors did not present inter-rater reliability data (e.g., Intraclass Correlation Coefficient [ICC] or Cohen's Kappa) for identifying and scoring subtle patterns such as TIB and VEP. The lack of these metrics makes it impossible to determine if the primary imaging results will be reliable and reproducible.

 

We thank the Reviewer for this important comment. We have enriched the manuscript with additional analyses to strengthen the methodological robustness, including assessment of inter-rater reliability for key imaging features. These additions help ensure that the scoring of subtle patterns such as TIB and VEP is reliable and reproducible. Please consider that vascular signs (vascular score = 0) were absent in a relatively large number of cases, which likely contributed to the higher inter-reader agreement.

 

5. While this is a significant limitation, a major technical issue is separating primary vascular pathology from secondary damage caused by compression or pruning in the surrounding lung due to consolidation or atelectasis. Advanced imaging modalities, such as dual-energy CT for perfusion mapping, could help define microvascular function independently of parenchymal opacities.

 

We fully agree with the Reviewer that distinguishing primary vascular pathology from secondary vascular alterations due to parenchymal involvement (e.g., consolidation or atelectasis-related compression or pruning) represents a technical limitation of CT-based visual assessment. This issue has already been highlighted in our previous work (Dalpiaz et al.), where extensive parenchymal involvement was supposed to potentially mask or confound the evaluation of vascular signs.

Nevertheless, as discussed in the manuscript and in line with the seminal work by Patel et al.  (Patel BV et al. Pulmonary Angiopathy in Severe COVID-19: Physiologic, Imaging, and Hematologic Observations. Am J Respir Crit Care Med. 2020) and subsequent studies, emerging evidence supports the existence of at least two partially distinct phenotypic patterns in affected patients: one characterized by predominant parenchymal involvement (with extensive ground-glass opacities and consolidations), and another with a more prominent vascular component, potentially reflecting different pathophysiological mechanisms and diagnostic–therapeutic implications. Within this framework, visual assessment of vascular alterations on standard chest CT remains clinically informative, albeit with acknowledged limitations.

We agree that advanced imaging techniques, such as dual-energy CT with perfusion mapping, would represent a valuable tool to better characterize microvascular dysfunction independently of parenchymal opacities. However, such data were not available in our cohort, and dual-energy acquisitions or CT pulmonary angiography were not routinely performed, similar to the study by Patel et al.

 

6. For the purposes of creating a high and low group of PS and VS values using the median value of PS and VS is an arbitrary cut-point determined by the data collected, which will reduce statistical power and increase the chance of finding false-positive relationships. Using PS and VS values in their continuous form or establishing clinical thresholds for their interpretation would be a more valid method of analysis.

We thank the Reviewer for this observation. We agree that dichotomizing the groups may reduce statistical power; however, the use of continuous variables would require the assumption of a linear association with the outcome. Moreover, at present, validated clinical thresholds for PS and VS are not available. Given the exploratory and hypothesis-generating nature of the present study, nature of our analysis and the absence of prior evidence supporting such linearity in this specific population, we chose a dichotomization approach. This methodology has been previously applied in similar patient cohorts, allowing comparability with existing literature and providing a more robust and interpretable framework for our exploratory objectives. (Colombo R, Wu MA, Ottolina D, Fossali T, Montomoli J, Lopez G, Catena E, Nebuloni M; Luigi Sacco Hospital COVID-19 ARDS Study Group. Failing categorization of severe COVID-19 ARDS into ventilatory subphenotypes studied via the clinical-histopathologic relationship. Respir Med. 2023 Aug-Sep;215:107283. doi: 10.1016/j.rmed.2023.107283).  Importantly, we now explicitly acknowledge that our findings should be interpreted as exploratory and require confirmation in larger, independent cohorts, ideally allowing for continuous modeling and/or the definition of clinically validated thresholds.

 

7. The authors have only completed a univariate analysis, so one critical issue remains unanswered: Is having a high PS score at the time of ICU admission an independent predictor of ICU admission after controlling for age, comorbidities at the time of admission, and total inflammatory burden before admission? A multivariable logistic regression analysis that controls for these critical confounding variables will demonstrate that PS score can be considered an independent predictor of ICU admission and is not simply an association with overall severity of disease at the time of admission.

 

We thank the reviewer for this important comment. As suggested, we performed a multivariable binary logistic regression analysis to assess whether the parenchymal score (PS), treated as a continuous variable, was independently associated with ICU admission after controlling for relevant confounders. The association between the PS and ICU admission was evaluated using a multivariable binary logistic regression model adjusting for age, C-reactive protein (CRP) as a marker of overall inflammatory burden, and number of comorbidities at admission. The overall model was statistically significant (χ² = 24.64, df = 4, p < 0.001), with a Nagelkerke R² of 0.587, indicating moderate explanatory power. After adjustment for demographic (age and number of comorbidities) and inflammation (CPR) variables, the parenchymal score (PS) remained independently associated with ICU admission (OR 2.57 per unit increase, p = 0.014), supporting its role as an independent predictor rather than a mere proxy of overall disease severity. Higher CRP levels were also independently associated with ICU admission (OR 1.21, p = 0.006), whereas age and number of comorbidities were not significantly associated with the outcome. The model demonstrated good discriminative performance, with an overall classification accuracy of 81.4%.  We have consequently modified the text.

 

8. The study has an overall sample size of 84 and tests for multiple biomarkers (over 10). The lack of using corrections for multiple comparisons (e.g., Bonferroni corrections) may lead to a very high risk of Type I error (false positives) through the increased number of univariate tests run. Additionally, the subgroup analyses performed based on VS may involve lower sample sizes (numbers of patients), and with such smaller sample sizes, the study is at risk of being underpowered to find statistically significant differences in very infrequent outcomes (hard outcomes) such as mortality.

We thank the reviewer for this important methodological comment. We acknowledge that the relatively small sample size (n = 84) and the number of biomarkers explored in univariate analyses may increase the risk of Type I error. However, the primary aim of the univariate analyses was exploratory, intended to screen potentially relevant variables and to generate hypotheses rather than to support definitive causal inferences. Univariate p-values were therefore interpreted cautiously and were not used in isolation to support the main conclusions. Importantly, the key findings were further evaluated in multivariable models, which mitigates the risk of spurious associations related to multiple testing. The main conclusions of the manuscript rely on these adjusted analyses rather than on uncorrected univariate comparisons.

Formal corrections for multiple comparisons (e.g., Bonferroni adjustment) were not applied, as these methods may be overly conservative in exploratory clinical studies involving correlated biomarkers, substantially increasing the risk of Type II error. This limitation has now been explicitly acknowledged in the revised manuscript.

In addition, we would like to emphasize that assembling a cohort with complete and contemporaneous radiological and inflammatory data is particularly challenging in real-world clinical settings, especially in acute care. The sample size reflects the stringent inclusion criteria required to ensure the availability of high-quality imaging data alongside a comprehensive inflammatory biomarker profile, which inevitably limits the number of eligible cases.

Regarding subgroup analyses based on ventilatory support, we agree that smaller subgroup sizes may limit statistical power, particularly for infrequent hard outcomes such as mortality. These analyses were therefore interpreted as descriptive and hypothesis-generating only, and no definitive conclusions regarding mortality were drawn. We have clarified this point in the revised Discussion. Overall, the study should be considered exploratory, aimed at identifying clinically meaningful signals that warrant confirmation in larger, adequately powered prospective cohorts.

The exploratory design, limited sample size, and multiple unadjusted comparisons increase the risk of both type I and type II errors, and the results—particularly from subgroup analyses—should therefore be interpreted with caution.

 

9. The discussion cites the inability to find an association between VS and clinical outcomes mainly because imaging was performed at only one time point; however, while this explanation is reasonable, the study should also have included a more critical, equally valid explanation for the lack of association. It is plausible the vascular signs seen on HRCT imaging represent an epiphenomenon, that is, they are secondary to the reaction of the peri-vascular inflammatory response to VS without any direct causal relationship to the deterioration seen clinically due to the VS. This hypothesis should also have been discussed and presented as a possible explanation for the study's findings.

 

We thank the reviewer for this comment, and we updated the manuscript with further discussion on possible explanation for the absence of identifiable association between VS and clinical outcomes.

 

10. The data for this study were collected at two time points - T1 and T2 - but due to the nature of the data collection methods used, it is not possible to perform a true longitudinal analysis of these data. Therefore, the current analyses only provide a limited understanding of CT score changes and their relationship to biomarker level changes and final outcome measures, as well as how the CT scores change over time.

 

We thank the Reviewer for this comment and agree with the observation. HRCT was performed only at admission; therefore, a true longitudinal imaging analysis is not possible, and no longitudinal correlations between CT score changes, biomarker dynamics, and clinical outcomes can be established. The manuscript has been revised to clearly state that all imaging–biomarker associations are cross-sectional and based on baseline HRCT only. We also note that the data analysed in the present work are part of a broader study, as already reported in the section addressing ethical approval.

 

11. As stated regarding the design flaws of this study, the assertion that the results of this study can be generalized to other respiratory infections are not supported by the research findings and represent an unwarranted extrapolation. This section should have concentrated on developing hypotheses to be tested in future comparative studies using other datasets rather than making generalizations based on this dataset's results. The purpose of this study was to explore an important topic and create new hypotheses through the study of participants at a single location. Due to the single-center nature of the study and other limitations in the design, such as lack of a control group, lack of operationalizing key vascular variables, and major limitations in conducting statistical analyses, the study results should be considered preliminary until the results are confirmed by others. The study provides meaningful associations between the potential for PS and various outcomes; however, there is also significant confounding in the results. Although the biological mechanisms being studied are fascinating, they do not establish a direct connection between the findings related to vascularity and clinical significance. The main contribution of this manuscript is the development of a conceptual and methodological framework for integrating various approaches to research. In order to consider future potential clinical applications of the study results, the findings must be confirmed through prospective, multicenter studies with appropriate control groups, quantitative imaging biomarkers, and adequate multivariable statistical models.

 

We thank the Reviewer for this detailed and thoughtful comment. We fully acknowledge the preliminary nature of our findings and the methodological limitations of the study, including its single-centre design, lack of control groups, and residual confounding. These aspects are now more explicitly and extensively addressed in the revised Limitations section. In addition, we have carefully revised the Discussion to attenuate any overly enthusiastic statements and to frame our results primarily as hypothesis-generating rather than confirmatory. We hope that these revisions clarify the exploratory intent of the study and render the manuscript more appropriate in tone and scope.

 

Reviewer 3 Report

Comments and Suggestions for Authors

Review Report

Manuscript Title: Stratifying COVID‑19 via Parenchymal and Vascular HRCT Patterns: Insights from Biohumoral and Endothelial Signatures

This manuscript investigates parenchymal and vascular HRCT patterns in COVID‑19 patients and their association with biohumoral and endothelial biomarkers. While the integration of imaging and biomarkers is scientifically interesting, the study has significant limitations that reduce its clinical impact. The small, single-center cohort (n=84), lack of multivariate analyses, absence of reproducibility metrics for imaging scores, and limited correlation of vascular HRCT patterns with relevant clinical outcomes weaken the robustness and generalizability of the findings. Additionally, some interpretations imply causality despite predominantly cross-sectional data. The following report outlines major and minor concerns to improve clarity, methodological rigor, and translational relevance.

Overall Major Concerns

• Sample size & power: Small cohort (n=84) may be underpowered, particularly for vascular subgroup analyses and biomarker correlations. No power calculation provided.
• Generalizability: Single-center, unvaccinated population; findings may not extrapolate to vaccinated patients or other healthcare settings. Heterogeneous CT scanners may introduce measurement variability.
• Statistical robustness: No multivariate analyses to adjust for confounders; multiple comparisons performed without correction, increasing risk of false-positive findings.
• Outcome relevance: Vascular HRCT scores correlate only with Angiopoietin‑2 (Ang-2) and not with ICU admission, mechanical ventilation, or mortality, limiting clinical significance.
• Methodology gaps: Semi-quantitative parenchymal and vascular scores lack reported inter- and intra-rater reproducibility; vascular injury dynamics were not captured longitudinally.
• Novelty overstatement: Integration of imaging and biomarkers is interesting, but clinical impact is restricted by small sample size and weak outcome correlations.
• Interpretation bias: Causal language occasionally used despite primarily cross-sectional associations.

Section-by-Section Evaluation

1. Title

• Slightly long and dense; terms like “biohumoral” may limit accessibility.
• Implies causal insights; consider rephrasing to reflect exploratory associations.

 

2. Abstract

• Overstates causality (“reflect distinct biological processes”).
• Omits key limitations (sample size, single-center, unadjusted confounders).
• Technical terms like “vascular tree-in-bud” may not be immediately clear.
• No confidence intervals or effect sizes reported.
• Claims clinical stratification potential, but vascular score outcome prediction is weak.

3. Introduction

• Relies on selective literature highlighting positive associations.
• Insufficient discussion of conflicting evidence on vascular HRCT prognostic value.
• Overemphasizes potential clinical translation without sufficient supporting evidence.

4. Materials and Methods

Study Design & Population:

• Prospective, single-center, consecutive sampling reduces selection bias but limits generalizability.
• Small sample size (n=84) without power calculation; potential selection bias.

CT Acquisition & Scoring:

• Heterogeneous scanners; no harmonization described.
• Semi-quantitative parenchymal (PS) and vascular (VS) scores; no inter-/intra-rater reliability reported.
• Vascular scoring may be obscured by parenchymal opacities.

Biomarker Assessment:

• Single baseline measurements; longitudinal changes not captured.

Statistical Analysis:

• Mainly univariate comparisons; no multivariable regression.
• Dichotomization at median reduces analytic sensitivity.
• Multiple comparisons uncorrected; increases risk of Type I error.

5. Results

Parenchymal Score (PS):

• Small sample may overstate statistical significance; no effect sizes or confidence intervals reported.

Vascular Score (VS):

• Only correlated with Ang-2; no association with ICU admission, mechanical ventilation, or mortality.
• VS-PS correlation may reflect collinearity rather than distinct biological processes.

Tables/Figures:

• Data presented clearly, but reproducibility and effect sizes are missing.

6. Discussion

• Overstates novelty and clinical utility; causal language used despite cross-sectional data.
• Selective citation; conflicting evidence not critically discussed.
• Methodological limitations acknowledged but impact on clinical interpretation underemphasized.
• Lack of longitudinal outcome data or validation cohorts.

7. Conclusion

• Overstates clinical utility of vascular HRCT patterns.
• Parenchymal HRCT-biomarker correlations more robust but limited by small sample size.
• Claims about patient stratification not fully supported by data.

8. References

• Heavy reliance on small studies and overlapping cohorts; robustness of findings should be critically discussed.
• Lack of references addressing inter-observer variability for vascular HRCT signs.

 

Summary of Critical Gaps

• Small, single-center cohort limits statistical power and external validity.
• Semi-quantitative imaging scores lack reproducibility metrics.
• Vascular HRCT abnormalities correlated with Ang-2 but not clinically relevant outcomes.
• Cross-sectional associations interpreted with causal language.
• Statistical analyses limited to univariate comparisons; no adjustment for confounders or multiple testing.
• Novelty claims overstated given methodological and clinical limitations.

Recommendations for Improvement

1. Clarify and temper causal language in abstract, discussion, and conclusion.
2. Provide inter-rater reliability metrics for HRCT scoring.
3. Perform multivariable analyses to adjust for confounders.
4. Report effect sizes with confidence intervals.
5. Correct for multiple comparisons in biomarker analyses.
6. Discuss limitations transparently, particularly regarding clinical applicability of vascular scores.
7. Consider larger, multicenter cohorts with longitudinal imaging and outcome data to validate findings.

Author Response

1. Overall Major Concerns

Sample size & power: Small cohort (n=84) may be underpowered, particularly for vascular subgroup analyses and biomarker correlations. No power calculation provided.

 

We thank the Reviewer for this important methodological comment.

Given the mono-centric and exploratory nature of the study, a formal a priori power analysis was not performed. The sample size was determined by the availability of patients meeting strict inclusion criteria during the pre-vaccination phase of the COVID-19 pandemic.

We acknowledge that the relatively limited number of cases may affect statistical power, and this has now been explicitly stated as a limitation of the study. Nevertheless, the sample size is comparable to that of previously published imaging-based studies focusing on detailed HRCT pattern stratification and biohumoral correlations in COVID-19 patients.

Importantly, despite the limited sample size, several associations reached statistical significance, supporting the biological plausibility and internal consistency of the observed findings. These results should therefore be interpreted as hypothesis-generating and warrant confirmation in larger, prospective cohorts.

 

2. Generalizability:Single-center, unvaccinated population; findings may not extrapolate to vaccinated patients or other healthcare settings. Heterogeneous CT scanners may introduce measurement variability.

Regarding the concerns about generalizability, we have already addressed these points within the Limitations section of the manuscript. The study specifically focuses on an unvaccinated population to examine the host-virus interaction in an immunologically naïve state; we believe this is a methodological strength rather than a weakness, as it provides a baseline model that is crucial for understanding the pathophysiology of potential future respiratory pandemics. Furthermore, the use of different CT scanners does not affect the validity of our results, as our assessments are based on semi-quantitative morphological patterns assigned by expert radiologists rather than strictly automated, computer-based quantitative metrics that would require rigid standardization of acquisition parameters.

 

3. Statistical robustness:No multivariate analyses to adjust for confounders; multiple comparisons performed without correction, increasing risk of false-positive findings.

 

Thank you for the comment. As explained to Reviewer 2, as suggested, we performed a multivariable binary logistic regression analysis to assess whether the parenchymal score (PS), treated as a continuous variable, was independently associated with ICU admission after controlling for relevant confounders. The association between the PS and ICU admission was evaluated using a multivariable binary logistic regression model adjusting for age, C-reactive protein (CRP) as a marker of overall inflammatory burden, and number of comorbidities at admission. The overall model was statistically significant (χ² = 24.64, df = 4, p < 0.001), with a Nagelkerke R² of 0.587, indicating moderate explanatory power. After adjustment for demographic (age and number of comorbidities) and inflammation (CPR) variables, the parenchymal score (PS) remained independently associated with ICU admission (OR 2.57 per unit increase, p = 0.014), supporting its role as an independent predictor rather than a mere proxy of overall disease severity. Higher CRP levels were also independently associated with ICU admission (OR 1.21, p = 0.006), whereas age and number of comorbidities were not significantly associated with the outcome. The model demonstrated good discriminative performance, with an overall classification accuracy of 81.4%.  We have consequently modified the text.

We acknowledge that the relatively small sample size (n = 84) and the number of biomarkers explored in univariate analyses may increase the risk of Type I error. However, the primary aim of the univariate analyses was exploratory, intended to screen potentially relevant variables and to generate hypotheses rather than to support definitive causal inferences. Univariate p-values were therefore interpreted cautiously and were not used in isolation to support the main conclusions. Importantly, the key findings were further evaluated in multivariable models, which mitigates the risk of spurious associations related to multiple testing. The main conclusions of the manuscript rely on these adjusted analyses rather than on uncorrected univariate comparisons.

Formal corrections for multiple comparisons (e.g., Bonferroni adjustment) were not applied, as these methods may be overly conservative in exploratory clinical studies involving correlated biomarkers, substantially increasing the risk of Type II error. This limitation has now been explicitly acknowledged in the revised manuscript.

In addition, we would like to emphasize that assembling a cohort with complete and contemporaneous radiological and inflammatory data is particularly challenging in real-world clinical settings, especially in acute care. The sample size reflects the stringent inclusion criteria required to ensure the availability of high-quality imaging data alongside a comprehensive inflammatory biomarker profile, which inevitably limits the number of eligible cases.

 

4. Outcome relevance:Vascular HRCT scores correlate only with Angiopoietin‑2 (Ang-2) and not with ICU admission, mechanical ventilation, or mortality, limiting clinical significance.

 

We share the Reviewer’s concern regarding the lack of correlation between the vascular score and primary clinical outcomes. However, we argue that this finding provides significant insight into the disease's pathophysiology that warrant further research, rather than limiting the study's relevance.

In our cohort, clinical outcomes like ICU admission and mechanical ventilation were primarily driven by the extent of parenchymal involvement (as shown by the strong correlation with the Parenchymal Score). The fact that the Vascular Score correlated specifically with Angiopoietin-2—a key marker of endothelial instability—but not with immediate respiratory failure, may suggest that HRCT can identify a distinct "endothelial phenotype."

This subclinical vascular involvement represents a specific layer of COVID-19 lung injury that is often masked by the more dominant parenchymal damage. Identifying this phenotype is crucial because, while it may not dictate immediate survival in all cases, it reflects an underlying microangiopathy that could require different therapeutic strategies or explain long-term vascular sequelae. We have further clarified this distinction in the Discussion to emphasize that these imaging findings serve as a surrogate for specific biological pathways (endothelial activation) rather than general clinical deterioration, and provide other possible explanations.

 

5. Methodology gaps:Semi-quantitative parenchymal and vascular scores lack reported inter- and intra-rater reproducibility; vascular injury dynamics were not captured longitudinally.

 

We thank for the observation. Regarding the methodology, we have assessed and reported the inter-rater variability as suggested by the Reviewer 2. Specifically, the inter-reader agreement was evaluated using the Intraclass Correlation Coefficient (ICC) for both parenchymal and vascular scoring. Furthermore, we acknowledge the Reviewer's point concerning the dynamics of vascular injury; as a cross-sectional study, our work was designed to capture a "snapshot" of the lung's status at admission. While we recognize that a longitudinal design would provide further insights into the evolution of these patterns, we believe our current findings offer a solid baseline for understanding the correlation between imaging and biomarkers in the acute phase.

 

6. Novelty overstatement:Integration of imaging and biomarkers is interesting, but clinical impact is restricted by small sample size and weak outcome correlations.
7. Interpretation bias:Causal language occasionally used despite primarily cross-sectional associations.

 

We acknowledge the Reviewer’s concerns regarding the study’s clinical reach. We have thoroughly revised the manuscript to moderate our claims, shifting from definitive conclusions to a more cautious, hypothesis-generating perspective. The results are now explicitly presented as preliminary, acknowledging that the small sample size and the specific nature of the vascular correlations limit immediate clinical extrapolation. We have lowered the tone throughout the text to reflect the exploratory nature of this research.

 

8. Title

• Slightly long and dense; terms like “biohumoral” may limit accessibility.
• Implies causal insights; consider rephrasing to reflect exploratory associations.

 

We now provide an updated version of the title, taking into account the reviewer’s observation.

 

9. Abstract

• Overstates causality (“reflect distinct biological processes”).
• Omits key limitations (sample size, single-center, unadjusted confounders).
• Technical terms like “vascular tree-in-bud” may not be immediately clear.
• No confidence intervals or effect sizes reported.
• Claims clinical stratification potential, but vascular score outcome prediction is weak.

We sincerely thank the Reviewer for the critical and constructive feedback. We have carefully revised the Abstract, also to ensure that the findings are presented as exploratory and preliminary.

While we have acknowledged the methodological limitations—including the single-center design and sample size—extensively in the Discussion to provide the necessary context for our results, we have chosen to keep the Abstract concise by focusing on the core findings. The preliminary nature of these findings is now clearly stated throughout the text to ensure a balanced interpretation.

Regarding the technical terminology, we believe that "vascular tree-in-bud" (TIB) is appropriate as it is widely recognized and established in the radiological literature, concerning COVID-19 pneumonia or other pathological conditions of the pulmonary vessels.

 

10. Introduction

• Relies on selective literature highlighting positive associations.
• Insufficient discussion of conflicting evidence on vascular HRCT prognostic value.
• Overemphasizes potential clinical translation without sufficient supporting evidence.

 

We thank the reviewer for the observation. We updated the manuscript accordingly.

 

11. Materials and Methods

CT Acquisition & Scoring:

• Heterogeneous scanners; no harmonization described.
• Semi-quantitative parenchymal (PS) and vascular (VS) scores; no inter-/intra-rater reliability reported.
• Vascular scoring may be obscured by parenchymal opacities.

We thank the Reviewer for these insightful comments and have addressed each point as follows:

1. Heterogeneous scanners / lack of harmonization
   Although two different CT scanners were used, both acquisitions were performed using a fully standardized high-resolution chest CT protocol, ensuring consistency in image quality and diagnostic performance. Importantly, this aspect does not represent a significant limitation for the present study, as we did not perform computer-based quantitative analyses, which are known to be highly sensitive to acquisition and reconstruction parameters. Our assessment was based on visual semi-quantitative scoring, which is less affected by minor technical variability.
2. Lack of inter- and intra-rater reliability for PS and VS
   In the revised version of the manuscript, we have now calculated inter- and intra-observer agreement for both parenchymal (PS) and vascular (VS) scores. These data have been added to the Materials and Methods section and reported in the Results, strengthening the methodological robustness of the scoring system.
3. Potential obscuration of vascular findings by parenchymal opacities
   We agree with this observation. This potential limitation has now been explicitly acknowledged and discussed in the Discussion section, and it has also been clearly listed among the study limitations.

 

12. Biomarker Assessment:

• Single baseline measurements; longitudinal changes not captured.

 

We acknowledge the reviewer’s observation, and updated the manuscript taking it into account.

 

13. Statistical Analysis:

• Mainly univariate comparisons; no multivariable regression.
• Dichotomization at median reduces analytic sensitivity.
• Multiple comparisons uncorrected; increases risk of Type I error.

 

Please, see the response to comment 3.

In addition, we agree that dichotomizing the groups may reduce statistical power; however, the use of continuous variables would require the assumption of a linear association with the outcome. Moreover, at present, validated clinical thresholds for PS and VS are not available. Given the exploratory and hypothesis-generating nature of the present study, nature of our analysis and the absence of prior evidence supporting such linearity in this specific population, we chose a dichotomization approach. This methodology has been previously applied in similar patient cohorts, allowing comparability with existing literature and providing a more robust and interpretable framework for our exploratory objectives. (Colombo R, Wu MA, Ottolina D, Fossali T, Montomoli J, Lopez G, Catena E, Nebuloni M; Luigi Sacco Hospital COVID-19 ARDS Study Group. Failing categorization of severe COVID-19 ARDS into ventilatory subphenotypes studied via the clinical-histopathologic relationship. Respir Med. 2023 Aug-Sep;215:107283. doi: 10.1016/j.rmed.2023.107283).  Importantly, we now explicitly acknowledge that our findings should be interpreted as exploratory and require confirmation in larger, independent cohorts, ideally allowing for continuous modeling and/or the definition of clinically validated thresholds.

 

14. Results

Parenchymal Score (PS):

• Small sample may overstate statistical significance; no effect sizes or confidence intervals reported.

Vascular Score (VS):

• Only correlated with Ang-2; no association with ICU admission, mechanical ventilation, or mortality.
• VS-PS correlation may reflect collinearity rather than distinct biological processes.

Tables/Figures:

• Data presented clearly, but reproducibility and effect sizes are missing.

 

Please, see the responses to the previous comments.

 

15. Discussion

• Overstates novelty and clinical utility; causal language used despite cross-sectional data.
• Selective citation; conflicting evidence not critically discussed.
• Methodological limitations acknowledged but impact on clinical interpretation underemphasized.
• Lack of longitudinal outcome data or validation cohorts.

 

We sincerely thank you for the critical and constructive feedback on our manuscript. We have carefully revised the Discussion and Conclusion sections to address your concerns, particularly regarding the tone and the interpretation of our findings.

Acknowledging that our study is based on cross-sectional data, we have replaced all causal terminology with more appropriate associative language. We have also significantly toned down the claims regarding the novelty and immediate clinical utility of the results, framing them as exploratory and hypothesis-generating.

To provide a more balanced perspective, we have expanded the discussion of conflicting evidence, specifically clarifying how our lack of association between vascular scores and clinical outcomes aligns with previous observations (e.g., Dalpiaz et al.). Finally, we have strengthened the "Limitations" section to more explicitly emphasize that the absence of a validation cohort and longitudinal data limits the generalizability and the prognostic interpretation of our work.

We believe these changes have significantly improved the balance and scientific rigor of the paper.

 

16. Conclusion

• Overstates clinical utility of vascular HRCT patterns.
• Parenchymal HRCT-biomarker correlations more robust but limited by small sample size.
• Claims about patient stratification not fully supported by data.

We have moderated the overly emphatic statements throughout the revised manuscript. We acknowledge that larger prospective studies are needed to confirm our findings and to achieve a deeper and more comprehensive understanding of this topic.

 

17. References

• Heavy reliance on small studies and overlapping cohorts; robustness of findings should be critically discussed.
• Lack of references addressing inter-observer variability for vascular HRCT signs.

We appreciate your observations. However, these are indeed the studies that have explored this topic to date. We acknowledge that the scientific evidence on the correlation of CT signs, laboratory biomarkers, and prognosis in COVID-19 is not yet complete, which is precisely why we believe our work can be valuable by providing an additional, albeit small, piece to the overall understanding.

Regarding your second point on inter-observer variability for vascular HRCT signs, to the best of our knowledge, no prior assessment of inter-observer agreement has been reported in the literature. We have now included this evaluation in the revised version of the manuscript.

 

18. Summary of Critical Gaps

• Small, single-center cohort limits statistical power and external validity.
• Semi-quantitative imaging scores lack reproducibility metrics.
• Vascular HRCT abnormalities correlated with Ang-2 but not clinically relevant outcomes.
• Cross-sectional associations interpreted with causal language.
• Statistical analyses limited to univariate comparisons; no adjustment for confounders or multiple testing.
• Novelty claims overstated given methodological and clinical limitations.

Recommendations for Improvement

1. Clarify and temper causal language in abstract, discussion, and conclusion.
2. Provide inter-rater reliability metrics for HRCT scoring.
3. Perform multivariable analyses to adjust for confounders.
4. Report effect sizes with confidence intervals.
5. Correct for multiple comparisons in biomarker analyses.
6. Discuss limitations transparently, particularly regarding clinical applicability of vascular scores.
7. Consider larger, multicenter cohorts with longitudinal imaging and outcome data to validate findings.

We sincerely thank you for the time and effort dedicated to reviewing our manuscript. We greatly appreciate your detailed comments, suggestions, and observations, which have allowed us to reflect more thoroughly on each aspect of the work.

As addressed point by point in our previous responses, we have endeavoured to tackle all criticisms and suggestions, making the necessary revisions and providing detailed explanations where appropriate. We hope that these efforts fully address your comments and contribute to significantly improving the overall quality of the manuscript.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

Reviewer Report (Second Round)

I have carefully re-evaluated the revised version of the manuscript following the authors’ response to the initial review. The study addresses an important and timely topic by integrating high-resolution CT (HRCT) parenchymal and vascular patterns with circulating endothelial and inflammatory biomarkers in patients with COVID-19 pneumonia. The prospective design and the attempt to link imaging phenotypes with biological signatures are commendable. However, despite some improvements, the revised manuscript still presents substantial methodological, statistical, and interpretative limitations that restrict its scientific robustness and clinical impact. Several key concerns raised in the first review remain only partially addressed, and additional issues related to literature coverage and clarity have emerged.

 

Major Comments

1. Literature Coverage Remains Incomplete

The reference list has not been sufficiently updated to reflect recent (2023–2025) evidence in critical areas directly relevant to the manuscript, particularly:

• CT diagnostic accuracy in COVID-19.
• Validation and reproducibility of CT severity scoring systems, including inter-observer agreement and multicenter studies.
• Automated and quantitative CT analysis, which has become central to COVID-19 imaging research and offers greater reproducibility than semi-quantitative visual scoring.
• Radiation dose optimization and diagnostic reference levels (DRLs) for chest CT in COVID-19, especially important in settings with heterogeneous scanners and limited resources.

The weakness of recent studies in these domains weakens the scientific context and limits the manuscript’s relevance to current radiology practice especially in limited resources region.

 

2. Methodological Limitations Persist

• The study remains single-center with a small sample size (n = 84) and no formal power calculation, limiting statistical robustness and external validity.
• Although inter-reader agreement is now mentioned, the vascular score (VS) remains insufficiently standardized, with limited justification for equal weighting of different vascular signs.
• The inclusion of two CT time points (T1 and T2) is not consistently reflected in the analysis, creating ambiguity regarding longitudinal assessment.

 

3. Statistical Analysis and Confounding

• Multivariable analyses remain limited and do not adequately adjust for key confounders such as timing from symptom onset, respiratory support level, or treatment effects.
• Dichotomization of imaging scores at the median reduces analytical sensitivity and clinical interpretability.
• Multiple comparisons across biomarkers are still performed without clear correction strategies, increasing the risk of false-positive findings.

 

4. Clinical Relevance of Vascular HRCT Findings

While parenchymal HRCT scores show more consistent associations with inflammatory biomarkers and ICU admission, vascular HRCT scores demonstrate limited correlation with clinically meaningful outcomes (ICU admission, mechanical ventilation, mortality). Consequently, claims regarding vascular patterns defining a distinct “vasculopathic phenotype” remain insufficiently supported by the data and should be substantially tempered.

 

5. Interpretation and Language

• Despite some revisions, causal or mechanistic language is still used to interpret predominantly cross-sectional associations.
• The discussion overstates novelty and clinical applicability, particularly regarding patient stratification based on vascular HRCT patterns.

Author Response

Response to Reviewer 3

We sincerely thank Reviewer 3 for the careful re-evaluation of our revised manuscript and for the detailed and constructive comments. We appreciate the recognition of the clinical relevance of the topic and of the prospective design integrating HRCT parenchymal and vascular patterns with circulating endothelial and inflammatory biomarkers in COVID-19 pneumonia. Below, we provide a point-by-point response to the major comments raised.

 

1. Literature Coverage

Reviewer comment:
Literature Coverage Remains Incomplete The reference list has not been sufficiently updated to reflect recent (2023–2025) evidence in critical areas directly relevant to the manuscript, particularly: CT diagnostic accuracy in COVID-19. Validation and reproducibility of CT severity scoring systems, including inter-observer agreement and multicenter studies. Automated and quantitative CT analysis, which has become central to COVID-19 imaging research and offers greater reproducibility than semi-quantitative visual scoring. Radiation dose optimization and diagnostic reference levels (DRLs) for chest CT in COVID-19, especially important in settings with heterogeneous scanners and limited resources. The weakness of recent studies in these domains weakens the scientific context and limits the manuscript’s relevance to current radiology practice especially in limited resources region.

Response:
We thank the Reviewer for highlighting the importance of an updated literature framework. In the revised manuscript, we have added several more recent references (2023–2025) addressing HRCT findings, disease severity assessment, and pathophysiological mechanisms in COVID-19. At the same time, we would like to note that, based on a comprehensive bibliographic search, a substantial proportion of the pivotal evidence on COVID-19 imaging and pathophysiology was generated during the earlier phases of the pandemic, when the disease burden was highest and large, well-characterized cohorts were available. Many of these studies remain highly relevant and are still widely cited in current literature.

Regarding radiation dose exposure, we would like to clarify that indications for chest CT imaging in COVID-19 patients are already explicitly discussed in the manuscript, in line with international recommendations. In our cohort, CT examinations were performed exclusively according to routine clinical practice in patients with moderate to severe disease, where imaging was considered clinically justified. Dose optimization was applied in all cases by means of automatic exposure control systems, with adjustment based on patient body habitus, as specified in the Methods section. We have further reinforced this point in the revised version of the manuscript to improve clarity.

Importantly, major European and international scientific societies and expert consensus statements have consistently emphasized the clinical utility of chest CT in patients with severe or deteriorating COVID-19, particularly for disease severity assessment, detection of complications, and support of clinical decision-making in acute settings. Our imaging protocol and clinical indications are fully consistent with these recommendations, and no additional radiation exposure beyond standard-of-care practice was introduced for research purposes.

With respect to automated and quantitative CT analysis, multicenter validation of CT severity scores, and detailed discussions on diagnostic reference levels, we acknowledge the relevance of these topics in contemporary COVID-19 imaging research. However, the primary aim of the present study was not to provide a comprehensive review of CT performance, scoring validation, or technical optimization strategies, but rather to explore the relationship between visually assessed HRCT parenchymal and vascular patterns and circulating endothelial and inflammatory biomarkers in a clinically characterized cohort. Expanding the literature review to comprehensively cover all these methodological and technical domains would have gone beyond the specific scope of the study and risked diluting its central pathophysiological focus. Nevertheless, where appropriate, we have cited recent representative studies to contextualize our findings within current radiological practice.

 

2. Methodological Limitations

Reviewer comment:
The study remains single-center with a small sample size (n = 84) and no formal power calculation, limiting statistical robustness and external validity. Although inter-reader agreement is now mentioned, the vascular score (VS) remains insufficiently standardized, with limited justification for equal weighting of different vascular signs. The inclusion of two CT time points (T1 and T2) is not consistently reflected in the analysis, creating ambiguity regarding longitudinal assessment.

Response:
We fully acknowledge the single-center nature and sample size of the study. The cohort size reflects the prospective enrollment during a defined time window and the availability of paired HRCT and biomarker data.

Regarding the absence of a formal power calculation, we would like to clarify that the study was designed as an exploratory, hypothesis-generating investigation. At the time of study conception, no reliable estimates of effect size were available for the association between vascular HRCT patterns and circulating endothelial biomarkers in COVID-19, making an a priori sample size calculation not feasible. Patient enrollment was therefore driven by the prospective inclusion of consecutive eligible subjects during a predefined pandemic time window, rather than by target sample size considerations. This aspect is explicitly acknowledged as a limitation, and the results should be interpreted accordingly.

Regarding the vascular score, we would like to emphasize that, to date, no validated or widely accepted vascular HRCT scoring systems are available for COVID-19 pneumonia. As already stated in the previous review round, the adoption of more complex visual-based scoring approaches may paradoxically increase inter-observer variability, particularly in the absence of standardized definitions and thresholds. For this reason, equal weighting of vascular signs was chosen as a pragmatic and transparent strategy to summarize the overall burden of vascular abnormalities, without implying equivalence in their underlying pathophysiological significance. We have clarified this rationale in the Methods section. Furthermore, we now explicitly acknowledge in the Limitations that advanced quantitative approaches and artificial intelligence–based analysis may help overcome some of the intrinsic limitations of visual-based scoring and improve reproducibility in future studies, and this point has been added accordingly.

Concerning CT timing, we would like to clarify that the study was not designed as a longitudinal imaging study and did not include two predefined CT time points. All analyses were based on a single HRCT examination performed at hospital admission, in accordance with routine clinical practice. Any reference to multiple CT time points has been removed to avoid ambiguity, and the manuscript has been revised to clearly reflect the cross-sectional imaging design of the study.

 

3. Statistical Analysis and Confounding

Reviewer comment:
Multivariable analyses remain limited and do not adequately adjust for key confounders such as timing from symptom onset, respiratory support level, or treatment effects.

Dichotomization of imaging scores at the median reduces analytical sensitivity and clinical interpretability.

Multiple comparisons across biomarkers are still performed without clear correction strategies, increasing the risk of false-positive findings.

Response:
We thank the Reviewer for this detailed methodological comment and for the opportunity to further clarify the rationale underlying our analytical and imaging choices.

Regarding the vascular score, we acknowledge that it represents a semi-quantitative, lobe-based approach. While fully quantitative or segment-based methods could theoretically provide greater granularity, the absence of dedicated software or validated artificial intelligence tools for this specific task would markedly reduce reproducibility and inter-reader agreement, ultimately limiting clinical feasibility and external validity. For this reason, a simplified semi-quantitative approach was deliberately adopted. We have added this point in the Limitations paragraph.

Importantly, a comparable methodology for the assessment of CT microvascular signs in COVID-19 has already been applied and validated by our group in a high–impact peer-reviewed journal (Dalpiaz et al., Radiologia Medica, 2022), supporting both the feasibility and the clinical relevance of this approach. In the present study, TIB and VEP were first evaluated separately and subsequently integrated into a composite vascular score in order to capture a broader spectrum of vascular involvement while minimizing observer variability.

Moreover, we agree that advanced approaches such as vascular textural analysis represent an interesting future direction; however, their applicability in patients with extensive parenchymal disease remains to be demonstrated, given that vascular abnormalities in COVID-19 are rarely isolated and are closely intertwined with parenchymal involvement.

Concerning score dichotomization, we acknowledge that this strategy may reduce statistical power. However, modeling parenchymal and vascular scores as continuous variables would require the assumption of a linear relationship with clinical outcomes, an assumption that is currently unsupported in this specific population. In addition, no clinically validated thresholds for either the parenchymal or vascular score are presently available. Given the exploratory and hypothesis-generating nature of the study, we therefore opted for a median-based dichotomization to provide a robust and clinically interpretable framework for group comparisons. This approach has been previously applied in similar COVID-19 ARDS cohorts, facilitating comparability with existing literature (Colombo et al., Respiratory Medicine, 2023). Importantly, to mitigate potential information loss, the parenchymal score was also analyzed as a continuous variable in multivariable regression, where it retained an independent association with ICU admission.

Finally, regarding multivariable adjustment and multiple comparisons, the statistical strategy was intentionally conservative. The multivariable model was kept parsimonious to avoid overfitting and inappropriate adjustment for downstream mediators of disease severity, in line with established methodological recommendations for exploratory cohorts of limited size. Biomarker analyses were explicitly framed as exploratory, biologically driven, and hypothesis-generating; therefore, formal multiplicity correction was not applied, as this could have disproportionately increased the risk of type II error. This limitation is now explicitly acknowledged in the revised manuscript, and the findings are interpreted accordingly.

Overall, we believe that the adopted methodological choices represent a balanced compromise between rigor, feasibility, and interpretability, and that the conclusions are appropriately cautious and proportionate to the exploratory aims of the study.

 

4. Clinical Relevance of Vascular HRCT Findings

Reviewer comment:
While parenchymal HRCT scores show more consistent associations with inflammatory biomarkers and ICU admission, vascular HRCT scores demonstrate limited correlation with clinically meaningful outcomes (ICU admission, mechanical ventilation, mortality). Consequently, claims regarding vascular patterns defining a distinct “vasculopathic phenotype” remain insufficiently supported by the data and should be substantially tempered.

Response:
We thank the Reviewer for this thoughtful comment and agree that, in our data, parenchymal HRCT scores show more consistent and clinically direct associations with inflammatory biomarkers and major outcomes such as ICU admission. Conversely, vascular HRCT scores were not independently associated with hard clinical endpoints, including mechanical ventilation or mortality.

In response to this observation, we have further tempered the wording of the manuscript to avoid any implication that vascular HRCT patterns alone define a prognostically validated phenotype. However, we respectfully maintain that the concept of a vasculopathic component of COVID-19 lung disease remains biologically and radiologically supported, even if not directly translatable into short-term clinical outcomes in our cohort.

Importantly, the aim of the present study was not to demonstrate the prognostic superiority of vascular imaging patterns, but rather to explore their biological correlates and pathophysiological relevance. In this regard, the selective association between higher vascular scores and elevated angiopoietin-2 levels—a well-established marker of endothelial activation—supports the interpretation that vascular HRCT abnormalities reflect a distinct underlying biological process. This interpretation is further reinforced by the significant correlation observed between parenchymal and vascular scores, suggesting that vascular involvement represents a complementary dimension of lung injury rather than an independent driver of clinical deterioration.

As already discussed in the manuscript, the lack of association between vascular HRCT scores and major clinical outcomes may be explained by several factors, including the timing of imaging acquisition, the single–time point assessment, and the intrinsic limitations of visual-based evaluation of vascular abnormalities in the presence of extensive parenchymal disease. These aspects are explicitly acknowledged in the Discussion.

Accordingly, we have further mitigated statements in the manuscript related to this point, in order to avoid overstatement while preserving the pathophysiological insight provided by the study.

 

5. Interpretation and Language

Reviewer comment:
Despite some revisions, causal or mechanistic language is still used to interpret predominantly cross-sectional associations.

The discussion overstates novelty and clinical applicability, particularly regarding patient stratification based on vascular HRCT patterns.

Response:
We thank the Reviewer for this comment and respectfully clarify that no causal or mechanistic inferences are intended in the interpretation of our results.

Throughout the manuscript, associations derived from cross-sectional data are consistently framed as descriptive and exploratory. Where biological or pathophysiological interpretations are discussed, they are explicitly presented as contextual hypotheses grounded in previously established literature, rather than as causal conclusions derived from the present dataset. To further avoid any potential ambiguity, we have carefully reviewed the manuscript and revised wording where necessary to ensure that associative language is consistently used.

Regarding novelty and clinical applicability, we would like to emphasize that the manuscript does not claim immediate clinical implementation or validated patient stratification based on vascular HRCT patterns. Instead, the novelty of the study lies in the integrated evaluation of parenchymal and vascular HRCT features alongside circulating biomarkers in a prospective cohort, a combination that has been only marginally explored in prior studies. Importantly, we now more explicitly state that the proposed imaging–biomarker associations are hypothesis-generating and intended to contribute to pathophysiological understanding rather than to define actionable clinical algorithms.

In response to the Reviewer’s concern, we have further tempered statements in the Discussion and Conclusions to clearly distinguish biological insight from clinical applicability. Patient stratification based on vascular HRCT patterns is now framed as a conceptual and exploratory perspective that requires confirmation in larger, longitudinal, and methodologically dedicated studies before any clinical translation can be considered.

We believe these revisions ensure that the manuscript remains scientifically rigorous, appropriately cautious, and aligned with the intrinsic limitations of a cross-sectional design.

 

Final remark

We thank Reviewer 3 again for the thoughtful critique, which has helped us improve the clarity, balance, and transparency of the manuscript. We believe that the revised version more accurately reflects the scope, strengths, and limitations of the study while preserving its original scientific intent.

Round 3

Reviewer 3 Report

Comments and Suggestions for Authors

Review Report – Third Round

This manuscript investigates HRCT parenchymal and vascular patterns in COVID-19 patients and their associations with circulating endothelial and inflammatory biomarkers. The authors have addressed several prior concerns, including inter-rater reliability and multivariable analysis for parenchymal scores. While these revisions improve clarity and methodological transparency, key limitations remain, particularly regarding vascular HRCT analysis, sample size, and literature coverage. Importantly, the manuscript does not adequately address studies from limited-resource settings, especially in terms of scoring validation, automated/quantitative CT analysis, and radiation dose optimization/reference levels, which limits the global relevance of the findings.

 

Overall Major Concerns

• Sample size & power: Cohort remains small (n = 84), limiting statistical power for vascular analyses. Single-center study reduces external validity.
• Clinical relevance of vascular HRCT scores: VS correlates with Angiopoietin-2 but not ICU admission, mechanical ventilation, or mortality. The claim of a “vasculopathic phenotype” remains hypothesis-generating, not clinically validated.
• Statistical robustness: Multivariable analysis limited to parenchymal scores and ICU admission. Dichotomization at median reduces sensitivity. Multiple comparisons remain uncorrected.
• Cross-sectional design: Only one HRCT per patient; vascular injury dynamics and longitudinal progression are not captured. Interpretations remain exploratory.
• Literature coverage (critical gap):
• Recent references (2023–2025) are included, but the manuscript fails to cite studies from low- and middle-income or resource-limited settings, where imaging protocols, disease characteristics, and availability of advanced CT scoring may differ.
• No discussion of validation of visual/quantitative scoring systems in such settings.
• Absence of references addressing automated or quantitative CT analysis applicable in limited-resource environments.
• No coverage of radiation dose optimization or diagnostic reference levels (DRLs) in settings with heterogeneous scanner availability or limited infrastructure.
• Methodological constraints: Equal weighting of vascular signs is arbitrary; parenchymal opacities may obscure vascular abnormalities. Semi-quantitative scoring is subjective, though inter- and intra-rater reliability is now reported.

 

Section-by-Section Evaluation

1. Title

• Slightly long and dense; “biohumoral” may reduce accessibility.
• Correctly emphasizes exploratory associations rather than causal insights.

2. Abstract

• Revised to clarify preliminary nature.
• Effect sizes/confidence intervals remain limited; vascular findings still not independently linked to clinical outcomes.

3. Introduction

• Updated references included, but literature from resource-limited settings is still underrepresented.
• Vascular HRCT literature partially cited; conflicting evidence and reproducibility issues remain insufficiently discussed.

4. Materials and Methods

• Inter- and intra-rater agreement reported for both PS and VS.
• Vascular score methodology justified; equal weighting remains arbitrary.
• Single HRCT time point clarified; any prior mention of T1/T2 removed.
• Dichotomization explained but reduces analytic sensitivity.

5. Results

• Parenchymal scores: multivariable analysis confirms independent association with ICU admission.
• Vascular scores: only correlate with Angiopoietin-2; no independent association with clinical outcomes.
• Effect sizes and confidence intervals for vascular associations remain modest.

6. Discussion

• Tone revised to avoid causal language.
• Vascular HRCT interpreted as biologically plausible but not directly clinically actionable.
• Literature coverage improved but still lacks studies from low- and middle-income/resource-limited settings, including:
• Validation of visual scoring and reproducibility
• Automated/quantitative CT approaches
• Radiation dose optimization and DRLs

7. Conclusion

• Statements on vascular stratification appropriately tempered.
• Parenchymal score associations more robust, but sample size limits generalizability.

8. References

• Updated with recent literature (2023–2025).
• Absence of references on automated/quantitative CT, scoring validation, and radiation dose optimization in limited-resource settings limits the manuscript’s applicability for global COVID-19 imaging practice.

 

Summary of Critical Gaps

• Small, single-center cohort limits statistical power and external validity.
• Vascular HRCT correlations present but not linked to clinical outcomes.
• Semi-quantitative scoring is subjective; equal weighting of vascular signs remains arbitrary.
• Statistical analyses partially adjusted; multiple comparisons still uncorrected.
• Cross-sectional design prevents mechanistic conclusions or longitudinal insight.
• Literature coverage remains incomplete: lack of automated/quantitative scoring studies, validation studies, and radiation dose optimization references in resource-limited settings.

 

Recommendations for Improvement

1. Highlight exploratory nature of findings throughout the manuscript.
2. Expand literature review to include:
• Validation studies of visual/quantitative CT scoring in resource-limited settings
• Automated/quantitative CT analysis in low- and middle-income countries
• Radiation dose optimization strategies and diagnostic reference levels applicable in these environments
3. Include effect sizes with confidence intervals for key associations.
4. Clarify limitations regarding clinical applicability of vascular HRCT patterns.
5. Maintain tempered language for vasculopathic phenotype; emphasize hypothesis-generating scope.
6. Future studies should explore larger, multicenter cohorts with longitudinal imaging to validate vascular patterns.

Author Response

Dear Reviewer,

Following the Academic Editor’s evaluation and guidance, the manuscript has been revised accordingly, and all requested minor modifications have been implemented.