Review Reports - NT-proBNP Discriminates Severe Systolic Dysfunction and Is Associated with Mortality in Advanced Duchenne Muscular Dystrophy: A Retrospective Cohort Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article submitted for review, "NT-proBNP Discriminates Severe Systolic Dysfunction and Is Associated with Mortality in Advanced Duchenne Cardiomyopathy: A Retrospective Cohort Study," is relevant to clinical medicine. However, several comments arose during the review process. In the Materials and Methods section, the method used to determine NT-proBNP is not specified. The catalog number of the assay kit is also not provided. Additionally, there is no description of the statistical methods—neither the software used nor the tests employed for the calculations are indicated. In the Results section, the tables have no titles. Furthermore, the results are discussed rather sparsely.

Author Response

Reply to the Reviewers

We sincerely thank the Reviewers for the thorough and constructive evaluation of our manuscript. We have carefully addressed each point and revised the manuscript accordingly. Our detailed responses are provided below.

Reviewer # 1

We thank the Reviewer for the helpful comments. We have now specified the NT-proBNP assay method in the Methods section. NT-proBNP was measured using an electrochemiluminescence immunoassay (Elecsys proBNP II, Roche Diagnostics, Mannheim, Germany) according to the manufacturer's instructions.

The statistical analysis section has been expanded to include the software used and the statistical tests applied. Specifically, continuous variables were expressed as mean ± standard deviation or median (interquartile range), as appropriate. Between-group comparisons were performed using Student’s t-test or Mann–Whitney U test, categorical variables using Fisher’s exact test, and correlations using Spearman’s coefficient. ROC analysis and logistic regression were used to evaluate diagnostic performance.

Titles have also been added to all tables, and the Results section has been expanded with additional descriptive interpretation of the findings and supplementary file. These modifications have improved the clarity of the manuscript.v

Reviewer 2 Report

Comments and Suggestions for Authors

The study includes only 31 patients without a power calculation. With only 11 patients in the EF <40% group, the derived cut-off and logistic regression models are inherently unstable.
The primary outcome groups EF <40% versus ≥40%, placing EF 40–49% (mild/moderate dysfunction) in the same category as preserved EF (>49%). This grouping dilutes the analysis and limits interpretability.
Bootstrap validation is an internal method that cannot compensate for small sample size or single‑center design. Describing it as confirming “stability” overstates the generalizability of the findings.
Patients have “established cardiomyopathy” but the duration of cardiac disease and prior treatment history are not specified. This may introduce lead‑time bias affecting both biomarker levels and outcomes.
Visual estimation of EF was used in some cases, which is subjective and operator‑dependent. Although a sensitivity analysis was performed, the lack of standardized quantitative protocols is a notable technical limitation.
SGLT2 inhibitors and neprilysin inhibitors were used exclusively in patients with EF <40% (Table 2). This creates confounding by indication that may influence NT‑proBNP levels independently of disease severity.
Logistic regression models include multiple predictors with only 11 events (EF <40%) and 5 deaths. This overfitting produces unstable odds ratios and excessively wide confidence intervals.
The mortality analysis is described as “exploratory” but conclusions about association are drawn from only five events. The reported odds ratio (2.56) is statistically underpowered and highly unstable.
The optimal NT‑proBNP cut‑off (>200 pg/mL) was derived and evaluated in the same cohort without a validation set. This “circular” analysis typically inflates diagnostic performance estimates.
Inclusion criteria require NT‑proBNP and EF measurements within a “short interval,” but the methods specify “within one month” for the biomarker and “within two weeks” for echocardiography. The lack of a fixed, narrow window introduces temporal mismatch.
The term “established cardiomyopathy” is not explicitly defined in the methods, making it difficult to understand the baseline patient population and the generalizability of the results.
While no correlation was found with respiratory parameters, the study does not explore potential interactions between NIV settings or respiratory failure severity and cardiac function in a more nuanced, multivariable fashion.
Cystatin‑C is reported but not included as a covariate in primary analyses. Renal function can influence NT‑proBNP clearance and should be adjusted for when evaluating diagnostic performance.
The manuscript does not contextualize its proposed cut‑off (>200 pg/mL) against established NT‑proBNP thresholds used in general heart failure populations, limiting clinical translation.
Requiring both a recent NT‑proBNP and echocardiogram within a narrow window may have excluded the sickest patients (unable to undergo imaging) or those with milder disease (no tests ordered), introducing selection bias.
The text reports a positive correlation between EF and LVEDD (r=0.45, p=0.01), which is physiologically paradoxical. As EF declines, LV dimensions typically increase, suggesting an error in analysis or reporting.
NIV use is presented inconsistently across tables, with denominators that do not clearly align. Table 1 and Table 2 show different numbers, raising concerns about data accuracy and clarity.
The retrospective design does not state whether echocardiogram readers were blinded to NT‑proBNP results. Lack of blinding can bias subjective assessments such as visually estimated EF.
The title and text refer to “advanced” Duchenne cardiomyopathy, yet the cohort includes patients with preserved EF (>49%). The term should be precisely defined or reserved for the EF <40% subgroup.
All patients were non‑ambulant with severe respiratory impairment, and many were on contemporary heart failure therapies. Findings may not generalize to younger, ambulatory patients or those in earlier disease stages.
Table 2 presents numerous comparisons without adjustment for multiple testing, increasing the risk of type I error for some reported associations.
Left ventricular dimensions were recorded “when available” without specifying the number of missing values or how missing data were handled, potentially biasing the reported correlations.
Table 2 shows a significant difference in ACE‑I/ARB use between EF groups (72.7% vs. 100%, p=0.01). This important confounder is not included in multivariable models.
Figure 2A shows an AUC of 0.96, but the curve appears to cross the diagonal—an unusual pattern that warrants verification of the underlying data and statistical calculations.
The odds ratio for log(NT‑proBNP) in Table 4 is <1 (0.04), suggesting that higher NT‑proBNP is associated with lower odds of EF <40%, which contradicts all other analyses and indicates a probable outcome coding reversal.
The use of EF <40% as the threshold for “severe” systolic dysfunction is common but not specifically justified for Duchenne cardiomyopathy, where the relationship between EF and clinical status may differ.
The conclusion that NT‑proBNP “discriminates severe systolic dysfunction” is overstated given the methodological limitations and the lack of external validation; “may complement imaging” is a more appropriate claim.
The statement that data are available “upon request” is vague. A stronger practice would specify whether anonymized data will be shared or if restrictions apply.

Author Response

Reply to the Reviewers

Reviewer # 2:

The study includes only 31 patients without a power calculation. With only 11 patients in the EF <40% group, the derived cut-off and logistic regression models are inherently unstable.

We agree that the sample size is limited. Duchenne muscular dystrophy with advanced cardiomyopathy represents a rare and clinically fragile population, making large cohorts difficult to assemble. We have now explicitly acknowledged this limitation in the Discussion and clarified that all regression and ROC analyses should be considered exploratory. No formal power calculation was performed due to the study's retrospective design. We have also reduced the emphasis on multivariable modelling and interpreted results cautiously.

The primary outcome groups EF <40% versus ≥40%, placing EF 40–49% (mild/moderate dysfunction) in the same category as preserved EF (>49%). This grouping dilutes the analysis and limits interpretability.

We agree that grouping patients with EF 40–49% together with those with preserved EF may reduce granularity. We selected EF <40% as the threshold because it is widely used to define severe systolic dysfunction and represents a clinically meaningful stage in Duchenne cardiomyopathy, often associated with treatment escalation and worse prognosis. In our cohort, 11 patients had EF <40%, 12 had EF 40–49%, and 8 had EF ≥50%. NT-proBNP levels differed significantly across EF categories (EF <40%, 40–49%, ≥50%; Kruskal–Wallis p = 0.00017). Post-hoc analysis showed higher NT-proBNP in EF <40% compared with EF 40–49% (p = 0.0002) and EF ≥50% (p = 0.001), whereas no difference was observed between EF 40–49% and EF ≥50% (p = 1.00). Age did not differ significantly across EF categories (EF <40%, 40–49%, and ≥50%: 31.0 ± 7.1, 35.6 ± 8.7, and 27.3 ± 7.9 years, respectively; p = 0.11).

These descriptive data have now been added to the Results. This limitation has also been acknowledged in the Discussion, and findings should be interpreted with this grouping strategy in mind.

Bootstrap validation is an internal method that cannot compensate for small sample size or single‑center design. Describing it as confirming “stability” overstates the generalizability of the findings.

We agree. The text has been revised to clarify that bootstrap validation provides internal consistency only and does not establish generalizability. We have removed statements suggesting “stability” and replaced them with “internal validation”.

Patients have “established cardiomyopathy”, but the duration of cardiac disease and prior treatment history are not specified. This may introduce lead time bias affecting both biomarker levels and outcomes.

We agree that the duration of cardiomyopathy and treatment history may influence NT-proBNP levels. However, due to the retrospective design, the exact onset of cardiac dysfunction could not be reliably determined. We have therefore acknowledged the potential for lead-time bias in the Discussion.

Visual estimation of EF was used in some cases, which is subjective and operator-dependent. Although a sensitivity analysis was performed, the lack of standardized quantitative protocols is a notable technical limitation

We agree that visual estimation of EF represents a technical limitation due to its operator dependence. In our cohort, EF was visually estimated in 5 of 31 patients (16.1%), while the remaining assessments were based on quantitative echocardiographic measurements. To address this concern, we performed a sensitivity analysis excluding these five patients and repeated the main analyses. (SUPPLEMENTARY FILE). Results were materially unchanged. NT-proBNP retained excellent discriminatory performance for EF <40% (AUC 0.96), with the same optimal cut-off (>200 pg/mL). The inverse correlation between NT-proBNP and EF also remained significant (Spearman r = −0.62, p=0.0007). These findings support the robustness of the association between NT-proBNP and severe systolic dysfunction despite the use of visual EF estimation in a minority of patients. This limitation and the results of the sensitivity analysis have now been clarified in the Methods, Results, and Discussion sections.

SGLT2 inhibitors and neprilysin inhibitors were used exclusively in patients with EF <40% (Table 2). This creates confounding by indication that may influence NT proBNP levels independently of disease severity.

We agree that the exclusive use of SGLT2 inhibitors and neprilysin inhibitors in patients with EF <40% reflects confounding by indication. These therapies were initiated after the identification of severe systolic dysfunction and therefore represent markers of disease severity. Importantly, both SGLT2 inhibitors and neprilysin inhibitors are known to reduce NT-proBNP levels. In our cohort, NT-proBNP remained markedly higher in the EF <40% group despite treatment, suggesting that the observed association is unlikely to be explained by therapy and may, if anything, be underestimated.

Additionally, we performed an exploratory multivariable logistic regression analysis evaluating the association between log-transformed NT-proBNP and severe systolic dysfunction, adjusted for SGLT2 inhibitor use. In this model, NT-proBNP remained significantly associated with EF <40%, whereas SGLT2 inhibitor use was not statistically significant, further supporting that the relationship between NT-proBNP and systolic dysfunction was not driven by treatment allocation.

This limitation has now been clarified in the Discussion.

Logistic regression models include multiple predictors with only 11 events (EF <40%) and 5 deaths. This overfitting produces unstable odds ratios and excessively wide confidence intervals

We agree that the limited number of events increases the risk of overfitting in regression models. The mortality analysis (Table 3) is univariable and presented as exploratory only. For the EF <40% outcome, we simplified the multivariable model and limited covariates to NT-proBNP and SGLT2 inhibitor use, selected a priori for clinical relevance. All regression analyses are now explicitly described as exploratory and interpreted cautiously. This limitation has been acknowledged in the Discussion

The mortality analysis is described as “exploratory” but conclusions about association are drawn from only five events. The reported odds ratio (2.56) is statistically underpowered and highly unstable.

We agree that the mortality analysis is based on a very limited number of events and is therefore underpowered. The logistic regression model assessing the association between NT-proBNP and mortality has been explicitly labelled as exploratory. We have softened the wording in the Results and Discussion, avoided causal interpretation, and emphasized the wide confidence intervals and statistical uncertainty. These findings are now presented as hypothesis-generating only.

9.The optimal NT proBNP cut off (>200 pg/mL) was derived and evaluated in the same cohort without a validation set. This “circular” analysis typically inflates diagnostic performance estimates.

We agree that the cut-off was derived and evaluated within the same cohort, which may inflate diagnostic performance. The threshold should therefore be considered hypothesis-generating. To partially address optimism bias, we performed bootstrap internal validation and sensitivity analyses, which yielded similar AUC estimates and cut-off values. However, external validation in independent cohorts is required before clinical application. This limitation has now been acknowledged in the Discussion.

Inclusion criteria require NT proBNP and EF measurements within a “short interval,” but the methods specify “within one month” for the biomarker and “within two weeks” for echocardiography. The lack of a fixed, narrow window introduces temporal mismatch

We thank the Reviewer for this observation. We have clarified and harmonized the timing window in the Methods section. NT-proBNP measurements and echocardiographic EF assessments were required to be performed within a maximum interval of one month, with most evaluations occurring within a shorter timeframe. We acknowledge that this temporal window may introduce minor mismatch between biomarker levels and cardiac function assessment, and this has now been explicitly discussed as a limitation.

.11. The term “established cardiomyopathy” is not explicitly defined in the methods, making it difficult to understand the baseline patient population and the generalizability of the results.

We agree that the term “established cardiomyopathy” requires clarification. We have now added a formal definition in the Methods section. In Duchenne muscular dystrophy, early cardiac involvement is often defined as EF <55%, whereas clinically relevant systolic dysfunction is typically defined as EF <50%. In the present study, established cardiomyopathy was defined as EF <55% in the appropriate clinical context. This has now been clarified in the Methods section.

While no correlation was found with respiratory parameters, the study does not explore potential interactions between NIV settings or respiratory failure severity and cardiac function in a more nuanced, multivariable fashion

We agree that respiratory failure severity may influence cardiac function. In this advanced Duchenne cohort, nearly all patients were on long-term non-invasive ventilation with similar ventilatory settings, limiting variability in respiratory parameters. At this disease stage, the number of hours of NIV is the most meaningful marker of respiratory failure severity. We therefore explored the association between NIV hours and both NT-proBNP levels and EF using Spearman correlation. No significant associations were observed (NT-proBNP: r = −0.10, p = 0.58; EF: r = 0.06, p = 0.72). Given the limited sample size and minimal variability in ventilatory parameters, further multivariable modelling was not performed to avoid overfitting. This has now been clarified in the manuscript.

Cystatin C is reported but not included as a covariate in primary analyses. Renal function can influence NT proBNP clearance and should be adjusted for when evaluating diagnostic performance.

We agree that renal function may influence NT-proBNP levels and should be considered when interpreting biomarker performance. In our cohort, cystatin C values were within the normal range (0.8 ± 0.2 mg/L), indicating no evidence of clinically relevant renal impairment. Given the absence of abnormal values and the limited variability, adjustment for renal function was not performed in multivariable analyses to avoid overfitting. This point has now been clarified in the manuscript.

The manuscript does not contextualize its proposed cut off (>200 pg/mL) against established NT proBNP thresholds used in general heart failure populations, limiting clinical translation

We compared our threshold with NT-proBNP values used in chronic heart failure, where levels >125 pg/mL are generally considered abnormal. The slightly higher threshold observed in our cohort (>200 pg/mL) may reflect disease-specific features of advanced Duchenne muscular dystrophy, including severe respiratory involvement, chronic NIV use, and marked alterations in body composition, which may influence natriuretic peptide levels and limit direct comparability with general heart failure populations. We therefore interpret this threshold as cohort-specific and hypothesis-generating.

Requiring both a recent NT proBNP and echocardiogram within a narrow window may have excluded the sickest patients (unable to undergo imaging) or those with milder disease (no tests ordered), introducing selection bias

We agree that requiring both NT-proBNP and echocardiographic assessment within a defined time window may introduce selection bias. Only patients undergoing both evaluations were included, potentially selecting individuals with clinical indications for cardiac assessment. However, this approach was necessary to ensure temporal comparability between biomarker levels and cardiac function. This limitation, inherent to the retrospective design, has now been acknowledged in the Discussion.

16 The text reports a positive correlation between EF and LVEDD (r=0.45, p=0.01), which is physiologically paradoxical. As EF declines, LV dimensions typically increase, suggesting an error in analysis or reporting.

We thank the Reviewer for this observation. The apparent positive correlation resulted from inverse coding of EF in the original analysis. After correcting the coding, the expected inverse relationship between EF and left ventricular dimensions was observed. The analysis has been corrected, and the Results section updated accordingly.

NIV use is presented inconsistently across tables, with denominators that do not clearly align. Table 1 and Table 2 show different numbers, raising concerns about data accuracy and clarity

We thank the Reviewer for this observation. The apparent inconsistency resulted from differences in NIV categorization and denominators between tables. We have now harmonized the definition of NIV use and revised all tables to ensure consistent reporting of absolute numbers and percentages across the manuscript.

The retrospective design does not state whether echocardiogram readers were blinded to NT proBNP results. Lack of blinding can bias subjective assessments such as visually estimated EF.

We agree that blinding of echocardiographic assessment to NT-proBNP levels was not ensured due to the retrospective design. Echocardiographic evaluations were performed as part of routine clinical care, and readers were not formally blinded to biomarker results. This may introduce potential bias, particularly for visually estimated EF values. However, visually estimated EF accounted for only a minority of cases, and sensitivity analysis excluding these patients yielded similar results. This limitation has now been acknowledged in the Discussion.

The title and text refer to “advanced” Duchenne cardiomyopathy, yet the cohort includes patients with preserved EF (>49%). The term should be precisely defined or reserved for the EF <40% subgroup.

We thank the Reviewer for this observation. To avoid potential ambiguity, we have modified the title to remove the term “advanced Duchenne cardiomyopathy” and replaced it with “advanced Duchenne muscular dystrophy” This better reflects the clinical characteristics of the cohort, which consisted of non-ambulant patients with severe respiratory impairment, while not all had severe systolic dysfunction. The title has been updated accordingly.

All patients were non ambulant with severe respiratory impairment, and many were on contemporary heart failure therapies. Findings may not generalize to younger, ambulatory patients or those in earlier disease stages.

We agree that the cohort represents patients with advanced Duchenne muscular dystrophy, characterized by non-ambulant status, severe respiratory impairment, and widespread use of contemporary heart failure therapies. Therefore, the findings may not be generalizable to younger, ambulatory patients or to earlier stages of cardiac involvement. This limitation has now been explicitly acknowledged in the Discussion.

Table 2 presents numerous comparisons without adjustment for multiple testing, increasing the risk of type I error for some reported associations.

We agree that multiple comparisons in Table 2 may increase the risk of type I error. Although comparisons were performed between two groups, several variables were tested. To address this concern, we applied Bonferroni correction for multiple testing. After correction, the main findings remained unchanged: NT-proBNP, LVEDD, and LVESD remained significantly different between EF groups, whereas treatment-related differences no longer met the corrected significance threshold. This has now been clarified in the Methods, Results, and Table 2 footnote.

Left ventricular dimensions were recorded “when available” without specifying the number of missing values or how missing data were handled, potentially biasing the reported correlations

We thank the Reviewer for this observation. Left ventricular dimensions were available for all included patients, and no missing data were present for these variables. This has now been clarified in the Methods and Results sections.

Table 2 shows a significant difference in ACE I/ARB use between EF groups (72.7% vs. 100%, p=0.01). This important confounder is not included in multivariable models.

We thank the Reviewer for this observation. Although ACE-I/ARB use differed between groups in the unadjusted analysis (p = 0.01), this difference did not remain significant after Bonferroni correction for multiple comparisons. Therefore, ACE-I/ARB use was not considered a significant confounder. In addition, treatment allocation reflects disease severity and clinical management rather than being an independent determinant of NT-proBNP levels. Given the limited number of events, additional covariates were not included in the multivariable model to avoid overfitting. This point has now been clarified in the manuscript.

Figure 2A shows an AUC of 0.96, but the curve appears to cross the diagonal—an unusual pattern that warrants verification of the underlying data and statistical calculations.

We thank the Reviewer for this observation. We carefully rechecked the ROC analyses and verified the coding of both the outcome and predictor variables. The ROC curve for identifying EF <40% remains consistently above the no-discrimination line, and the AUC (0.96) was confirmed. The stepwise appearance of the curve reflects the limited sample size and discrete threshold values rather than an error in the analysis. The figure has been regenerated for clarity.

In the mortality analysis, the ROC curve is more irregular due to the very small number of events (n = 5) and the resulting wide confidence intervals. This instability is expected in small samples and does not indicate an error in the calculations. The mortality analysis is explicitly described as exploratory, and the figure has been regenerated accordingly.

The odds ratio for log(NT proBNP) in Table 4 is <1 (0.04), suggesting that higher NT proBNP is associated with lower odds of EF <40%, which contradicts all other analyses and indicates a probable outcome coding reversal

We thank the Reviewer for this observation. In the logistic regression model, the outcome was coded as preserved systolic function (EF ≥40%). Therefore, odds ratios <1 indicate a lower probability of preserved EF and, consequently, a higher probability of EF <40%. Accordingly, higher NT-proBNP levels were associated with severe systolic dysfunction, consistent with the ROC and correlation analyses. This has now been clarified in the Methods and Table 4 legend.

The use of EF <40% as the threshold for “severe” systolic dysfunction is common but not specifically justified for Duchenne cardiomyopathy, where the relationship between EF and clinical status may differ

We agree that the use of EF <40% required justification. In Duchenne muscular dystrophy, earlier cardiac involvement may be recognised at higher EF thresholds, often <55%, whereas clinically relevant systolic dysfunction is commonly defined as EF <50%. We selected EF <40% to identify a subgroup with more advanced systolic dysfunction, consistent with the late-stage DMD literature, which uses LVEF ≤40% to define reduced systolic function associated with worse outcomes, and in keeping with conventional heart-failure frameworks for reduced EF. We have now clarified this rationale in the Methods and Discussion (see REF N° 17)

The conclusion that NT proBNP “discriminates severe systolic dysfunction” is overstated given the methodological limitations and the lack of external validation; “may complement imaging” is a more appropriate claim.

We agree that the original wording may have overstated the findings, given the retrospective design, small sample size, and lack of external validation. We have therefore softened the language throughout the manuscript. The conclusion has been revised to:
“NT-proBNP may complement imaging in identifying severe systolic dysfunction in advanced Duchenne cardiomyopathy.” We have also avoided causal or definitive statements and now present the findings as exploratory and hypothesis-generating.

The statement that data are available “upon request” is vague. A stronger practice would specify whether anonymized data will be shared or if restrictions apply.

We agree and have revised the data availability statement to clarify that anonymized individual-level data are available upon reasonable request, subject to institutional approval and applicable privacy regulations. We thank the Reviewer again for the insightful comments, which have significantly improved the clarity and rigour of the manuscript.

Reply to the Reviewers

Reviewer # 2:

The study includes only 31 patients without a power calculation. With only 11 patients in the EF <40% group, the derived cut-off and logistic regression models are inherently unstable.

The primary outcome groups EF <40% versus ≥40%, placing EF 40–49% (mild/moderate dysfunction) in the same category as preserved EF (>49%). This grouping dilutes the analysis and limits interpretability.

These descriptive data have now been added to the Results. This limitation has also been acknowledged in the Discussion, and findings should be interpreted with this grouping strategy in mind.

Bootstrap validation is an internal method that cannot compensate for small sample size or single‑center design. Describing it as confirming “stability” overstates the generalizability of the findings.

Patients have “established cardiomyopathy”, but the duration of cardiac disease and prior treatment history are not specified. This may introduce lead time bias affecting both biomarker levels and outcomes.

Visual estimation of EF was used in some cases, which is subjective and operator-dependent. Although a sensitivity analysis was performed, the lack of standardized quantitative protocols is a notable technical limitation

SGLT2 inhibitors and neprilysin inhibitors were used exclusively in patients with EF <40% (Table 2). This creates confounding by indication that may influence NT proBNP levels independently of disease severity.

This limitation has now been clarified in the Discussion.

Logistic regression models include multiple predictors with only 11 events (EF <40%) and 5 deaths. This overfitting produces unstable odds ratios and excessively wide confidence intervals

The mortality analysis is described as “exploratory” but conclusions about association are drawn from only five events. The reported odds ratio (2.56) is statistically underpowered and highly unstable.

Inclusion criteria require NT proBNP and EF measurements within a “short interval,” but the methods specify “within one month” for the biomarker and “within two weeks” for echocardiography. The lack of a fixed, narrow window introduces temporal mismatch

.11. The term “established cardiomyopathy” is not explicitly defined in the methods, making it difficult to understand the baseline patient population and the generalizability of the results.

While no correlation was found with respiratory parameters, the study does not explore potential interactions between NIV settings or respiratory failure severity and cardiac function in a more nuanced, multivariable fashion

Cystatin C is reported but not included as a covariate in primary analyses. Renal function can influence NT proBNP clearance and should be adjusted for when evaluating diagnostic performance.

The manuscript does not contextualize its proposed cut off (>200 pg/mL) against established NT proBNP thresholds used in general heart failure populations, limiting clinical translation

Requiring both a recent NT proBNP and echocardiogram within a narrow window may have excluded the sickest patients (unable to undergo imaging) or those with milder disease (no tests ordered), introducing selection bias

NIV use is presented inconsistently across tables, with denominators that do not clearly align. Table 1 and Table 2 show different numbers, raising concerns about data accuracy and clarity

The retrospective design does not state whether echocardiogram readers were blinded to NT proBNP results. Lack of blinding can bias subjective assessments such as visually estimated EF.

The title and text refer to “advanced” Duchenne cardiomyopathy, yet the cohort includes patients with preserved EF (>49%). The term should be precisely defined or reserved for the EF <40% subgroup.

All patients were non ambulant with severe respiratory impairment, and many were on contemporary heart failure therapies. Findings may not generalize to younger, ambulatory patients or those in earlier disease stages.

Table 2 presents numerous comparisons without adjustment for multiple testing, increasing the risk of type I error for some reported associations.

Left ventricular dimensions were recorded “when available” without specifying the number of missing values or how missing data were handled, potentially biasing the reported correlations

Table 2 shows a significant difference in ACE I/ARB use between EF groups (72.7% vs. 100%, p=0.01). This important confounder is not included in multivariable models.

Figure 2A shows an AUC of 0.96, but the curve appears to cross the diagonal—an unusual pattern that warrants verification of the underlying data and statistical calculations.

The odds ratio for log(NT proBNP) in Table 4 is <1 (0.04), suggesting that higher NT proBNP is associated with lower odds of EF <40%, which contradicts all other analyses and indicates a probable outcome coding reversal

The use of EF <40% as the threshold for “severe” systolic dysfunction is common but not specifically justified for Duchenne cardiomyopathy, where the relationship between EF and clinical status may differ

The conclusion that NT proBNP “discriminates severe systolic dysfunction” is overstated given the methodological limitations and the lack of external validation; “may complement imaging” is a more appropriate claim.

The statement that data are available “upon request” is vague. A stronger practice would specify whether anonymized data will be shared or if restrictions apply.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I have no comments on the corrected manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

Authors answered comments very well, hence i accept this manuscript in the present form