Diabetes-Related Differences in the Predictive Value of the Physiological Assessment of Myocardial Ischemia for Long-Term Clinical Outcomes

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the invitation to review this article by Zasada et al on the role of physiological indices in predicting mortality in DM vs non DM patients.

Here is my structured review.

Introduction

Overall provides adequate background to the topic.

The added benefit of other evaluation methods- such as intravascular imaging in evaluating borderline coronary lesions, especially in DM patients who often have complex and diffuse disease, should be emphasized here.

Methods

Overall provides good reproductibility for future studies.

Please state the exclusion criteria (ex ACS). Was left main disease included?

The 50-90% interval in too wide, and is not the usual cutoff for intermediate coronary lesions used in other studies, as it also includes significant coronary lesions.

Results

Results are overall clearly presented.

Can you add medication on baseline (ex statin, ACEI, antiplatelets, nitrates, BB).

Discussion

Please provide a limitations section- retroscpetive design, evaluation of all- cause mortality and not cardiac mortality etc.

Conclusion

Nothing to add here.

 

English language overall good, needs minor adjustments.

Author Response

Thank you for the invitation to review this article by Zasada et al on the role of physiological indices in predicting mortality in DM vs non DM patients.

Here is my structured review.

Introduction

Overall provides adequate background to the topic.

Thank you very much for this kind comment.

The added benefit of other evaluation methods- such as intravascular imaging in evaluating borderline coronary lesions, especially in DM patients who often have complex and diffuse disease, should be emphasized here.

I would like to thank you for this comment - I have added information in the introduction about the possibility of using intravascular imaging to extend the diagnosis of coronary circulation, as suggested by the reviewer.

Methods

Overall provides good reproductibility for future studies.

Thank you very much for this positive comment.

Please state the exclusion criteria (ex ACS). Was left main disease included?

Patients diagnosed with ACS were excluded from the presented analysis, but we did not define any exclusion criteria related to the location of the lesion being assessed. Therefore, the analyzed group of patients included individuals with borderline stenosis of the LMCA. This information was added to the text at the reviewer's suggestion.

The 50-90% interval in too wide, and is not the usual cutoff for intermediate coronary lesions used in other studies, as it also includes significant coronary lesions.

Of course, I fully agree with this comment. I presented such a broad scope because the collected data are "real-world data based on daily practice." The operators working in our laboratory have considerable discretion in qualifying patients for physiological assessment. Having this option, we have recently based this qualification much more frequently on visual angiographic assessment than on quantitative assessment. Therefore, in the entire analyzed group, there may have been lesions that obstructed the vessel lumen by almost 90%, which were assessed visually as slightly less advanced and therefore qualified for physiological assessment. This applies particularly to relatively short lesions, for which the operator had doubts about their hemodynamic significance. Therefore, we decided to present a very general description of the group in this way – in subsequent analyses, the degree of stenosis assessed angiographically is not directly addressed. Moreover, we add this issue into the limitations section.

Results

Results are overall clearly presented.

Thank you very much for this kind comment.

 

Can you add medication on baseline (ex statin, ACEI, antiplatelets, nitrates, BB).

Thank you very much for your comment. Regarding the treatment data, we did not collect this data for two reasons. First, we relied on data from the angiography laboratory, where information on chronic treatment is very limited. We could have obtained such data by analyzing medical history and hospitalization records—and we have performed such analyses several times before. However, based on our experience with this type of analysis, they usually do not provide very significant information due to the fact that in this group of patients, virtually everyone is prescribed aspirin, ACE inhibitors, and a statin. We do not use nitrates in almost any patient. If a patient has undergone a PCI procedure, they take a second antiplatelet drug—usually clopidogrel, unless the PCI procedure was very complex, in which case a stronger drug. Regarding beta-blockers—almost 90% of our patients have indications for these medications, and we recommend them to approximately that number of patients. I have added information about the lack of data on pharmacotherapy in the 'limitations' section.

Discussion

Please provide a limitations section- retroscpetive design, evaluation of all- cause mortality and not cardiac mortality etc.

Thank you very much for this valid point - I added the 'limitations' section as suggested by the reviewer.

Conclusion

Nothing to add here.

Thank you very much for this kind comment.

English language overall good, needs minor adjustments.

Thank you very much for your positive evaluation.

Reviewer 2 Report

Comments and Suggestions for Authors

1. Include a paragraph mentioning the diagnosis of the multiple regression model.
2. Try to follow the TRIPOD-IA statement (transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis). 

Author Response

 

1. Include a paragraph mentioning the diagnosis of the multiple regression model.

Thank you very much for this comment - as suggested by the reviewer, I have added a paragraph presenting the diagnostic results of both multiple regression models used for the described analyses.

At the same time, I would like to thank you for this comment also because, returning to the discussion with statistics on multiple models, which turned out to be reproducible and correct, we also verified univariate models and here minor corrections of parameters were necessary, which we introduced in the current version of the work.

1. Try to follow the TRIPOD-IA statement (transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis). 

Thank you very much for this valid comment - we have modified the text of the work, trying to add the information recommended by the TRIPOD+AI statement to the greatest extent possible.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript presents a single-center retrospective analysis of 381 patients with chronic coronary syndrome who underwent invasive physiological assessment of borderline coronary stenoses, with outcomes analyzed according to the presence or absence of diabetes mellitus. The authors report that non-hyperemic indices (RFR/iFR) demonstrated prognostic value for long-term all-cause mortality in both diabetic and non-diabetic populations, with iFR appearing particularly predictive in patients with diabetes and RFR in those without diabetes. The topic is clinically interesting, as diabetes is known to affect coronary physiology, microvascular function and long-term cardiovascular outcomes. However, several important issues need to be addressed before the manuscript can be considered for publication.

First, the statistical methodology requires substantial revision. The authors used logistic regression to assess predictors of death over approximately 4.4 years of follow-up. Given that mortality is a time-dependent outcome and follow-up duration is not perfectly identical across patients, logistic regression is not the optimal approach. Time-to-event methods, specifically Kaplan–Meier survival analysis and Cox proportional hazards modeling, would be more suitable. Hazard ratios would be more appropriate than odds ratios in this context and would allow a more accurate representation of risk over time. This is a major methodological issue that should be corrected.

Second, treatment decisions were clearly influenced by physiological findings, particularly in patients with diabetes, who underwent PCI more frequently. This introduces potential confounding by indication and treatment mediation. The prognostic value of physiological indices cannot be interpreted independently of the revascularization decisions they triggered. PCI status should at minimum be included as a covariate in multivariable models. Ideally, stratified analyses or propensity-adjusted analyses should be considered.

Third, the distinction made between iFR being prognostic in diabetic patients and RFR being prognostic in non-diabetic patients appears overstated. Both indices are physiologically similar resting measurements and are highly correlated. Moreover, the choice of non-hyperemic index was operator-dependent, which introduces potential selection bias. The statistical models do not robustly demonstrate superiority of one index over the other within each subgroup.

Fourth, the exclusive use of all-cause mortality as the endpoint significantly limits the interpretability of a coronary physiology study. No data are provided regarding myocardial infarction, repeat revascularization or composite major adverse cardiovascular events. While the use of national registry data strengthens mortality ascertainment, the absence of cause-specific outcomes reduces mechanistic insight. This limitation should be acknowledged.

Fifth, important clinical variables that could influence long-term outcomes are not reported, including baseline and discharge medical therapy (statins, ACE inhibitors/ARBs, beta-blockers, antiplatelet therapy) and modern glucose-lowering therapies such as SGLT2 inhibitors or GLP-1 receptor agonists. These treatments are highly relevant in diabetic populations and may substantially influence mortality risk. If these data are unavailable, this should be clearly stated as a limitation.

From a methodological standpoint, further clarification is also required regarding FFR measurement. Intracoronary adenosine boluses ranged from 100–400 µg, which is a relatively broad range. The authors should specify dosing by vessel, whether repeat measurements were performed in borderline cases and whether intravenous adenosine was used.

The manuscript would significantly benefit from the inclusion of a central illustration.

Minor issues include several typographical errors, duplicated references in the bibliography and some imprecise language.

In summary, this manuscript addresses a clinically important question and contains potentially valuable data. However, major revisions are required.

Author Response

This manuscript presents a single-center retrospective analysis of 381 patients with chronic coronary syndrome who underwent invasive physiological assessment of borderline coronary stenoses, with outcomes analyzed according to the presence or absence of diabetes mellitus. The authors report that non-hyperemic indices (RFR/iFR) demonstrated prognostic value for long-term all-cause mortality in both diabetic and non-diabetic populations, with iFR appearing particularly predictive in patients with diabetes and RFR in those without diabetes. The topic is clinically interesting, as diabetes is known to affect coronary physiology, microvascular function and long-term cardiovascular outcomes. However, several important issues need to be addressed before the manuscript can be considered for publication.

Thank you very much for your positive evaluation.

First, the statistical methodology requires substantial revision. The authors used logistic regression to assess predictors of death over approximately 4.4 years of follow-up. Given that mortality is a time-dependent outcome and follow-up duration is not perfectly identical across patients, logistic regression is not the optimal approach. Time-to-event methods, specifically Kaplan–Meier survival analysis and Cox proportional hazards modeling, would be more suitable. Hazard ratios would be more appropriate than odds ratios in this context and would allow a more accurate representation of risk over time. This is a major methodological issue that should be corrected.

Thank you very much for this valid point, with which I fully agree. The issue of selecting statistical methodology was discussed within our team, and we are aware that the Cox model and Kaplan-Meier analyses would be very useful in our work. However, we decided that unfortunately, we would not be able to perform them. This is due to the fact that our data indicates the patient's status. We know exactly who is deceased at the time of verification of this status, but in most cases, we do not have the exact date of death. Therefore, we had a choice between using the missing data imputation method or a logistic regression model. We decided to use logistic regression. I hope this method will ultimately meet with the reviewer's approval, especially since, following another reviewer's suggestions, we expanded the description of the model used and added a limitations section describing the reasons for not using the Cox model.

Second, treatment decisions were clearly influenced by physiological findings, particularly in patients with diabetes, who underwent PCI more frequently. This introduces potential confounding by indication and treatment mediation. The prognostic value of physiological indices cannot be interpreted independently of the revascularization decisions they triggered. PCI status should at minimum be included as a covariate in multivariable models. Ideally, stratified analyses or propensity-adjusted analyses should be considered.

Thank you for your comment. The impact of PCI on prognosis, as well as coronary revascularization in general by any method (including surgical), was analyzed from various perspectives while preparing this publication. Initially, it seemed obvious that revascularization per se should significantly impact patient prognosis, meaning it should improve it. However, during the analysis, it turned out that this impact is likely somewhat smaller than we expected. In the previous version of the paper, we presented the variable 'LAD revascularization' in the univariate analysis, denoting any type of LAD revascularization. We assumed that this vessel should have the greatest impact on prognosis, but it turned out that its revascularization was not a significant factor influencing prognosis in either the diabetic or non-diabetic group. In the current version, I also added the variable 'PCI' to the univariate analysis, denoting any percutaneous revascularization. This variable also did not significantly impact prognosis in either group. After a deeper analysis, we concluded that, overall, these observations are consistent with the assumptions of physiological assessment: unless a hemodynamically significant stenosis is confirmed, revascularization of such a lesion does not improve the prognosis. It appears that in such cases, the prognosis does not differ significantly from that observed in the group of patients in whom a hemodynamically significant lesion was revascularized.

Third, the distinction made between iFR being prognostic in diabetic patients and RFR being prognostic in non-diabetic patients appears overstated. Both indices are physiologically similar resting measurements and are highly correlated. Moreover, the choice of non-hyperemic index was operator-dependent, which introduces potential selection bias. The statistical models do not robustly demonstrate superiority of one index over the other within each subgroup.

Thank you very much for this valid comment. I completely agree; the presented analyses were not designed to identify differences between different non-hyperemic assessment methods. Furthermore, the researcher's choice of the physiological assessment system also determined the non-hyperemic assessment method, so we do not have data on the results of such an assessment using more than one method for a given patient. These are significant limitations – we mentioned them in the limitations section, and we also tried to soften the message by emphasizing that the observed differences between the different non-hyperemic assessment methods may be random in our case.

Fourth, the exclusive use of all-cause mortality as the endpoint significantly limits the interpretability of a coronary physiology study. No data are provided regarding myocardial infarction, repeat revascularization or composite major adverse cardiovascular events. While the use of national registry data strengthens mortality ascertainment, the absence of cause-specific outcomes reduces mechanistic insight. This limitation should be acknowledged.

Thank you for your comment, which I fully agree with. I've described the limitations of this chosen endpoint in the limitations section, as suggested by the reviewer.

Fifth, important clinical variables that could influence long-term outcomes are not reported, including baseline and discharge medical therapy (statins, ACE inhibitors/ARBs, beta-blockers, antiplatelet therapy) and modern glucose-lowering therapies such as SGLT2 inhibitors or GLP-1 receptor agonists. These treatments are highly relevant in diabetic populations and may substantially influence mortality risk. If these data are unavailable, this should be clearly stated as a limitation.

Thank you very much for your comment. Regarding pharmacotherapy, we did not collect such information in our database – we wrote about this directly in the limitations section. Regarding the mentioned treatment, virtually all patients diagnosed with chronic coronary syndrome receive statins and ACE-I/ARBs. The vast majority receive beta-blockers, especially those with reduced left ventricular systolic function or arrhythmias. For antiplatelet therapy, virtually every patient receives aspirin, and if we performed a PCI procedure, most are also prescribed clopidogrel, unless we recommended stronger treatment due to the complexity of the procedure. Regarding SGLT2 inhibitors, we currently use this treatment very frequently in our group of patients, while GLP-1 receptor agonists are used much less frequently, primarily for economic reasons.

From a methodological standpoint, further clarification is also required regarding FFR measurement. Intracoronary adenosine boluses ranged from 100–400 µg, which is a relatively broad range. The authors should specify dosing by vessel, whether repeat measurements were performed in borderline cases and whether intravenous adenosine was used.

Thank you very much for this observation – as per your comment, I have expanded on this topic in the 'Materials and Methods' section. Our standard procedure is to administer a 200 µg adenosine bolus intravenously, regardless of the vessel being assessed. If we have doubts about achieving complete hyperemia after such a bolus, we sometimes decide to administer a larger dose – up to 400 µg. This approach is quite rare, most often related to situations where the geometry of the catheter and the ostium of the vessel being examined raises suspicions that an incomplete dose of adenosine may be administered to the coronary vessel. Reducing the adenosine dose is most often used in two situations: if we have a clearly positive result from the non-hyperemic assessment, which we perform first, and if we are confident that even with incomplete hyperemia, we will obtain a positive FFR result. The second case is a multivessel assessment, when the patient responded with a prolonged complete block after the first 200 µg bolus. In such cases, we sometimes decide to reduce the dose, especially the first dose, in subsequent assessments. Continuous adenosine infusion is currently rarely used; we consider it especially when planning a more comprehensive assessment of the coronary circulation, including IMR.

The manuscript would significantly benefit from the inclusion of a central illustration.

Thank you very much for this valid comment - I have prepared such an illustration and attached it to the current version of the work.

Minor issues include several typographical errors, duplicated references in the bibliography and some imprecise language.

Thank you very much for your comments - we have modified the text to correct the issues mentioned.

In summary, this manuscript addresses a clinically important question and contains potentially valuable data. However, major revisions are required.

Thank you very much for this favorable comment.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors answered accordingly to the questions raised.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript has been adequately revised and it is suitable for publication.