HIV-Associated Microstructural Abnormalities in Default Mode, Executive Control, and Salience Networks: Insights from Tensor-Valued Diffusion Encoding

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an interesting neuroimaging study investigating white matter differences between people living with HIV and a control group, and their association with neurocognitive performance. The authors build on their previous published work of the same cohort utilizing a different approach; they utilize novel diffusion metrics to assess white matter integrity across functional brain networks. This approach is intriguing and potentially valuable in revealing patterns and characteristics of HIV-associated brain injury. However, there are some issues that if addressed could benefit the manuscript and make the result presentation more convincing. Below are my comments:

1. The authors acknowledge that none of the significant results in between-group dMRI comparisons survive correction for multiple comparisons. While transparent on the issue, the wording in the manuscript is very strong regarding the certainty of the results and their implications. The interpretation should remain proportionate to the strength of the evidence: the results support promising, network-specific associations and hypothesis generation, rather than definitive proof that tensor-valued dMRI outperforms conventional metrics.
2. The statistical plan and results include a two-way ANOVA analysis. This type of analysis involves two categorical variables; however, the model described in the manuscript include both a categorical and a continuous variable (dMRI metrics). So this is probably a linear regression or general linear model.
3. The manuscript should acknowledge that while cognitive metrics are adjusted for age and education, imaging metrics are not. Also, cognitive differences between PLWH and controls may not be HIV-driven. The cohort appears to consist of PLWH on stable cART with protocol-defined virologic suppression (HIV RNA ≤200 copies/mL), but the manuscript would benefit from clearer reporting of HIV disease characteristics in the analyzed sample, including current viral load, CD4 count, nadir CD4, duration of infection/treatment, ART regimen, AIDS history, and relevant comorbidity burden. Moreover, the pattern of cognitive impairment (memory/learning/fluency) is not the most typical for HIV-associated brain injury, where executive function, speed and visuospatial ability are most often implicated.
4. The limitations section of the manuscript needs expansion to include racial imbalance between groups, limited HIV-clinical characterization of the cohort, risk of residual confounding (e.g. substance use, comorbidities), multiple testing, the interpretive ambiguity of the diffusion metrics, and the limited external generalizability.
5. FDR correction is mentioned, but it is not obvious how it was applied, eg in which family of tests/analyses.

Author Response

Thank you for your thoughtful and constructive comments, which have significantly improved the clarity and rigor of our manuscript. In response, we have revised the manuscript to better reflect the exploratory nature of this pilot study, clarified the statistical methods, and updated the interpretation of the findings. We have also expanded the Methods and Limitations sections, added/modified new tables and figures, and included effect size estimates alongside unadjusted p-values to improve transparency and interpretability. We believe these revisions have strengthened the manuscript and addressed the reviewers’ concerns.

Comment 1: The authors acknowledge that none of the significant results in between-group dMRI comparisons survive correction for multiple comparisons. While transparent on the issue, the wording in the manuscript is very strong regarding the certainty of the results and their implications. The interpretation should remain proportionate to the strength of the evidence: the results support promising, network-specific associations and hypothesis generation, rather than definitive proof that tensor-valued dMRI outperforms conventional metrics.

Response 1: We thank the reviewer for this important comment and agree that the interpretation should reflect the strength of the evidence. As noted, the between-group diffusion MRI comparisons did not survive correction for multiple comparisons. Our aim was to highlight potentially meaningful network-specific trends for further investigation, rather than to claim definitive superiority of tensor-valued diffusion MRI metrics.

To address this, we have revised the manuscript (including new figures and tables and updates to the main text) to frame the results as preliminary, hypothesis-generating observations, emphasizing that larger studies are needed to assess whether tensor-valued diffusion MRI provides greater sensitivity than conventional metrics. These clarifications have been incorporated into the Abstract, Results, and Discussion sections to moderate the language and avoid overstating conclusions. As this is an exploratory pilot study, we now report effect sizes alongside unadjusted p-values and do not apply multiple-comparison correction.

Comment 2: The statistical plan and results include a two-way ANOVA analysis. This type of analysis involves two categorical variables; however, the model described in the manuscript include both a categorical and a continuous variable (dMRI metrics). So this is probably a linear regression or general linear model.

Response 2: Thank you for this helpful comment. We agree that the terminology used in the manuscript was not precise. The model we implemented includes a categorical variable (HIV status) and a continuous variable (dMRI metric) along with their interaction term; therefore, it is more appropriately described as a linear regression within the general linear model framework, rather than a two-way ANOVA.

We have revised the manuscript to clarify this point. Specifically, the statistical analysis is now described as an ordinary least squares (OLS) linear regression model of the form: Cognitive score ~ MRI metric + HIV status + HIV status × MRI metric.

This model tests the main effects of MRI metrics and HIV status, as well as their interaction, to determine whether the association between the MRI metric and cognitive score differs by HIV status. The terminology in the Methods and Results sections has been updated accordingly. Importantly, this change clarifies the statistical description and does not affect the study’s results or conclusions.

Comment 3:  The manuscript should acknowledge that while cognitive metrics are adjusted for age and education, imaging metrics are not. Also, cognitive differences between PLWH and controls may not be HIV-driven. The cohort appears to consist of PLWH on stable cART with protocol-defined virologic suppression (HIV RNA ≤200 copies/mL), but the manuscript would benefit from clearer reporting of HIV disease characteristics in the analyzed sample, including current viral load, CD4 count, nadir CD4, duration of infection/treatment, ART regimen, AIDS history, and relevant comorbidity burden. Moreover, the pattern of cognitive impairment (memory/learning/fluency) is not the most typical for HIV-associated brain injury, where executive function, speed and visuospatial ability are most often implicated.

Response 3: We thank the reviewer for these suggestions. We have clarified that cognitive metrics were adjusted for age and education, whereas imaging metrics were not, and we acknowledge that the observed cognitive differences between PWH and controls may not be solely HIV-driven. We have added Table 1 to present cohort demographics and clinical characteristics, including current viral load, CD4 count, ART regimen, and relevant comorbidities.

We have also noted in the Discussion (limitations) that the pattern of cognitive differences observed (memory, learning, and fluency) does not reflect the classical profile of HIV-associated neurocognitive impairment, which more commonly involves executive function, processing speed, and visuospatial abilities, and we interpret these findings with appropriate caution.

Comment 4: The limitations section of the manuscript needs expansion to include racial imbalance between groups, limited HIV-clinical characterization of the cohort, risk of residual confounding (e.g. substance use, comorbidities), multiple testing, the interpretive ambiguity of the diffusion metrics, and the limited external generalizability.

Response 4: We thank the reviewer for this suggestion. We have expanded the Limitations section to address these points, including racial imbalance between groups, limited HIV clinical characterization, potential residual confounding (e.g., substance use, comorbidities), multiple testing, interpretive ambiguity of diffusion MRI metrics, and limited external generalizability of the findings.

 Comment 5:  FDR correction is mentioned, but it is not obvious how it was applied, eg in which family of tests/analyses.

Response 5: Thank you for pointing this out. In the original manuscript, Benjamini–Hochberg (BH) FDR correction was applied; however, in the revised version, we have removed the FDR correction because the study is intended as an exploratory pilot analysis. Instead, we now report uncorrected p-values along with effect sizes to allow readers to interpret the magnitude and direction of the observed effects. This clarification has been added to the Statistical Analysis section of the manuscript.

 

Reviewer 2 Report

Comments and Suggestions for Authors This study applied tensor-valued diffusion MRI (dMRI) to investigate the white matter metrics alternation and their association with cognition in people with HIV (PWH) across functionally defined brain regions. Compared with healthy controls, PWH showed altered dMRI metrics in default mode, executive control, and salience networks, along with poorer performance in learning, memory, and language domains. Importantly, tensor-valued diffusion MRI metrics demonstrated greater sensitivity than conventional diffusion measures in detecting HIV-related brain changes and their associations with domain-specific cognitive outcomes, suggesting their potential as biomarkers of HIV-associated neurocognitive vulnerability. The experiment was conducted and described clearly, the data processing and statistical analysis were robust and sound. However, there are a couple of comments and questions that need to be addressed before further consideration. Major

• In the Spearman correlation (line 211), what's the reason that only age is included as a covariate? It's understandable that the gender is not equally distributed in the sample, but it would be valuable to conduct a sensitivity analysis with gender as a covariate.
• Similarly, education is mentioned when generating cognitive measurements; it would be great to show the sensitivity analysis of the correlation analysis with education as additional covariance. Alternatively, some justification would be appreciated.
• It's not clear whether BH correction is also conducted in the ANOVA analysis. it should be clarified around lines 218-219.
• It's also not clear how many independent ANOVA tests were performed. I'd recommend that authors to clarify this information in session 2.6.2. For example, in line 215, "...ANOVA was performed with cognitive performance modeled as a function of XX (How many) dMRI metrics, HIV status, and .... over xx(how many) cognitive measurements across 6 functional networks."
•  

Minor

• On line 164, are the slices acquired in the axial direction?
• For Table 1, please highlight the significant results better, probably with an asterisk.
• Figure 4, I'd recommend removing the background colour to increase the contrast. Optionally, I'd recommend changing from colour code (red and green) to shape (solid vs circle, for example) to make it friendly for a black/white printer.
• There's a discrepancy regarding the FA DMN_D differences between HC and PWH, i.e., it's p=0.032 in line 235, but p=0.031 in figure 2.
• Please try to avoid split table in two pages, as these tables are quite informative. This probably needs to work with the editors in the later process.

Author Response

Thank you for your thoughtful and constructive comments, which have significantly improved the clarity and rigor of our manuscript. In response, we have revised the manuscript to better reflect the exploratory nature of this pilot study, clarified the statistical methods, and moderated the interpretation of the findings. We have also expanded the Methods and Limitations sections, added/modified new tables and figures, and included effect size estimates alongside unadjusted p-values to improve transparency and interpretability. We believe these revisions have strengthened the manuscript and addressed the reviewers’ concerns.

Comment 1: In the Spearman correlation (line 211), what's the reason that only age is included as a covariate? It's understandable that the gender is not equally distributed in the sample, but it would be valuable to conduct a sensitivity analysis with gender as a covariate.

Response 1: Thank you for these questions regarding the correlation analyses. The cognitive z-scores were norm-adjusted for age and education, which accounts for their primary effects on cognitive performance. In the correlation analyses with dMRI metrics, age was included as a covariate to account for any residual age-related variance that may remain despite normalization.

Comment 2: Similarly, education is mentioned when generating cognitive measurements; it would be great to show the sensitivity analysis of the correlation analysis with education as additional covariance. Alternatively, some justification would be appreciated.

Response 2: Education was not additionally included as a covariate because its influence is already addressed through the normalization procedure, and including it again could introduce over-adjustment. Gender was not included due to sample imbalance, and prior work suggests that sex differences do not substantially alter dMRI–cognition associations in similar cohorts. Given these considerations, we did not perform additional sensitivity analyses for gender or education, as the cognitive scores are already demographically corrected and the primary analyses remain robust. This clarification has been added to the manuscript.

Comment 3: It's not clear whether BH correction is also conducted in the ANOVA analysis.

Response 3: Because the primary goal of this study is hypothesis generation in an exploratory pilot context, we report uncorrected p-values alongside effect sizes in the revised manuscript. This allows readers to interpret both the magnitude and direction of the observed effects while recognizing the exploratory nature of the findings.

Comment 4: It's also not clear how many independent ANOVA tests were performed. I'd recommend that authors to clarify this information in session 2.6.2. For example, in line 215, "...ANOVA was performed with cognitive performance modeled as a function of XX (How many) dMRI metrics, HIV status, and .... over xx(how many) cognitive measurements across 6 functional networks.

Response 4: Thank you for this helpful suggestion. We have revised Section 2.6.2 to clarify the number of analyses performed. Specifically, for each cognitive measure, an ordinary least squares (OLS) linear regression model was run, including the corresponding dMRI metric, HIV status, and their interaction. This analysis was repeated across 7 dMRI metrics and 6 cognitive measures, resulting in a total of 42 independent tests.

Because this study is designed as an exploratory pilot investigation, we did not apply multiple-comparison correction to the OLS regression results (the more appropriate analysis, replacing the previously described two-way ANOVA as suggested by Reviewer 1) in the revised version of the manuscript. Instead, we report uncorrected p-values together with effect sizes to provide an estimate of the magnitude and direction of the observed effects. This clarification has been added to the Statistical Analysis section of the manuscript.

Comment 5: On line 164, are the slices acquired in the axial direction?

Response 5: Yes, tensor-valued diffusion weighted images were collected in the axial direction. This clarification has been added to the main text.

Comment 6: For Table 1, please highlight the significant results better, probably with an asterisk.

Response 6: Thank you for the suggestion. To improve readability, we have highlighted the significant results in bold rather than using asterisks, as this approach makes the table less visually cluttered while still clearly indicating statistically significant values.

Comment 7: Figure 4, I'd recommend removing the background colour to increase the contrast. Optionally, I'd recommend changing from colour code (red and green) to shape (solid vs circle, for example) to make it friendly for a black/white printer.

Response 7: The figure has been updated according to your suggestion.

Comment 8: There's a discrepancy regarding the FA DMN_D differences between HC and PWH, i.e., it's p=0.032 in line 235, but p=0.031 in figure 2.

Response 8: Sorry for the typo. It has now been corrected.

Comment 9: Please try to avoid split table in two pages, as these tables are quite informative. This probably needs to work with the editors in the later process.

Response 9: We agree that keeping the tables on a single page would improve readability. This formatting issue will likely be addressed during the journal’s production and typesetting stage, and we will work with the editors to ensure the tables are presented clearly in the final version.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author have significantly improved their manuscript. I suggest accept in present form.

Author Response

Thanks so much for your time. The manuscript has been updated to correct minor typos and replace Table 2, which used age and sex as covariates. With the inclusion of sex, the overall pattern of results and main conclusions remained unchanged, with consistent network-specific associations between dMRI metrics and cognitive performance.

 

Reviewer 2 Report

Comments and Suggestions for Authors

The revised manuscript is much clearer and sound. I don't have any further concerns.

Author Response

Thank you. The manuscript has been updated to correct minor typos and replace Table 2, which used age and sex as covariates. With the inclusion of sex, the overall pattern of results and main conclusions remained unchanged, with consistent network-specific associations between dMRI metrics and cognitive performance.