Review Reports
- Anna C. S. Garrison 1,
- Evgeny Jenya Chumin 2,3 and
- Melissa A. Cyders 1,*,†
- et al.
Reviewer 1: Anonymous Reviewer 2: Anonymous
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsReview comments:
In this manuscript, the authors studied sex differences in the associations of anxiety and depression symptoms with stress-induced alcohol consumption and possible biological mechanisms of this association. The authors found that anxiety symptoms were associated with larger stress-induced alcohol consumptions for females but not for males. No statistically significant association was observed for depression symptoms and stress-induced alcohol consumption. Finally, network connectivity analyses revealed underlying mechanisms of potential explanatory value, although this requires more investigation.
While this project has some interesting aspects, I have several concerns relating to design and presentation.
General comments:
- The paper is very long and, at times, unfocused. I think it needs to be markedly shortened. This goes for all parts of the manuscript but especially the introduction where it needs to be much clearer how (1) this piece of work fills in an existing knowledge gap and (2) the specific series of conducted studies have a logical and sequential order to each other.
- In relation to the above point: I think aims and especially hypotheses could be stated clearer and more concise.
- Words as “subject anxiety intercept”, “subject anxiety slope” etc. makes it difficult to understand what the authors are assessing. What real world concepts does these expressions cover?
- Please refrain from repeating entire result sections in the discussion (e.g., line 596—604)
Methods, results, and interpretation:
- Figure 1: I appreciate the use of directed acyclic graphs (DAG).
- Please draw the DAGs according to standard DAG rules. E.g., does sex impact anxiety, subjective anxiety slope factor, stress-induced alcohol seeking, or all three? This is not clear from the current presentation.
- In relation to this: I think the DAGs need to reflect the hypotheses of the study more accurately. Hence, the hypothesis is that there exist sex differences and, in that light, it would be helpful to have separate DAGs for males and females. E.g., does the same relations exists for both males and females or how does sex modify the associations? I think this needs to be reflected in the DAGs.
- Please also include selection mechanisms of the study into these DAGs as they my impact the minimal set of adjustment variables.
- Table 3: Two parameters (Baseline Subjective Anxiety and Baseline Subjective Stimulation) differ statistically but the Cohen’s d is rather low (i.e., <0.5). What is the relation between p-values and Cohen’s d – are the statistically significant values clinically relevant? This needs to be addressed.
- Figure 2 and 2a: Please indicate on the figure what line that corresponds to what sex. It would also be beneficial to add 95%CIs to the figure to indicate at what points the two curves significantly depart, especially because they cross over. Where are the actual differences observed – I think this needs to be discussed further.
- Not sure what the +/-1 SD means in figure 3.
- Was any of the epidemiological analyses (not regarding the brain imagine part) corrected for multiple tests? I recommend this, as the authors conducted multiple comparisons.
- This study was conducted in a population of healthy adults, why was this population chosen? Is there an underlying hypothesis that alcohol consumption is a widespread coping mechanism in the general population? And should this be more frequent among one of the sexes according to the hypothesis?
- In relation to this: How does your sample reflect the male:female ratio for AUD (normally regarded 75:25)– there is an oversampling of females in this study. How does that impact interpretation of results – could it even explain the lack of associations among males?
- Finally, what implications does the study specific male:female ratio and the sample in general (healthy adults) have for generalizability?
- Line 555-567: The authors describe that the incongruent results are due to previous studies relying on self-reports and therefore covering stress exposure over extended time periods while the present study assessed acute stress. Is the problem rather that this section compares two distinct concepts (acute vs chronic stress) rather than previous studies relying on self-reports and hence the differing results? Does males and females react similarly to acute and chronic stress?
Other considerations:
- In this study, people were exposed to alcohol, was there conducted any follow-up assessments or checks on the participants?
Author Response
Reviewer 1:
Review comments:
In this manuscript, the authors studied sex differences in the associations of anxiety and depression symptoms with stress-induced alcohol consumption and possible biological mechanisms of this association. The authors found that anxiety symptoms were associated with larger stress-induced alcohol consumptions for females but not for males. No statistically significant association was observed for depression symptoms and stress-induced alcohol consumption. Finally, network connectivity analyses revealed underlying mechanisms of potential explanatory value, although this requires more investigation.
While this project has some interesting aspects, I have several concerns relating to design and presentation.
General comments:
- The paper is very long and, at times, unfocused. I think it needs to be markedly shortened. This goes for all parts of the manuscript but especially the introduction where it needs to be much clearer how (1) this piece of work fills in an existing knowledge gap and (2) the specific series of conducted studies have a logical and sequential order to each other.
Our response: We thoroughly edited the manuscript throughout, with particular attention to the introduction, to streamline writing, improve clarity, and more clearly present how this study fills an existing knowledge gap and fits within a logical and sequential order.
- In relation to the above point: I think aims and especially hypotheses could be stated clearer and more concise.
Our response: We streamlined our aims and hypotheses to increase clarity (Pages 2-3, Lines 93-108).
- Words as “subject anxiety intercept”, “subject anxiety slope” etc. makes it difficult to understand what the authors are assessing. What real world concepts does these expressions cover?
Our response: We edited language throughout to clarify we are referring to our measure of state anxiety (baseline state anxiety before alcohol infusion and change in state anxiety during alcohol infusion).
- Please refrain from repeating entire result sections in the discussion (e.g., line 596—604)
Our response: We edited the discussion to remove repeating results.
Methods, results, and interpretation:
- Figure 1: I appreciate the use of directed acyclic graphs (DAG). Please draw the DAGs according to standard DAG rules. E.g., does sex impact anxiety, subjective anxiety slope factor, stress-induced alcohol seeking, or all three? This is not clear from the current presentation.
Our response: Due to reviewer and editor comments regarding the utility of Figure 1, as well as requests to streamline and increase clarity, we deleted Figure 1.
- In relation to this: I think the DAGs need to reflect the hypotheses of the study more accurately. Hence, the hypothesis is that there exist sex differences and, in that light, it would be helpful to have separate DAGs for males and females. E.g., does the same relations exists for both males and females or how does sex modify the associations? I think this needs to be reflected in the DAGs.
Our response: Due to reviewer and editor comments regarding the utility of Figure 1, as well as requests to streamline and increase clarity, we deleted Figure 1.
- Please also include selection mechanisms of the study into these DAGs as they my impact the minimal set of adjustment variables.
Our response: Due to reviewer and editor comments regarding the utility of Figure 1, as well as requests to streamline and increase clarity, we deleted Figure 1.
- Table 3: Two parameters (Baseline Subjective Anxiety and Baseline Subjective Stimulation) differ statistically but the Cohen’s d is rather low (i.e., <0.5). What is the relation between p-values and Cohen’s d – are the statistically significant values clinically relevant? This needs to be addressed.
Our response: There is no direct relationship between p and Cohen’s d, as the p value is used to evaluate statistical significance and is sensitive to sample size, while Cohen’s d is a measure of practical significance (effect size) that should not be influenced by sample size. In other words, a smaller p-value is not necessarily associated with a bigger effect size. The significant p-values indicate that the mean baseline state anxiety and state stimulation likely differ between men and women in the population, and the corresponding Cohen’s d values show that the magnitude of these differences was likely in the small-to-moderate range. While one may interpret these baseline differences as having limited clinical impact, they may still play a role in shaping downstream outcomes, especially in processes that are sensitive to initial states. We added a statement concerning the small sizes of the effects to Section 3.1 (Page 5, Lines 245-246): “Women reported higher baseline state anxiety and stimulation (p’s= 0.03 – 0.05; Table 3), though effect sizes were small (Cohen’s d<0.5).” We considered adding more information, but we worried a broader discussion would distract from the more streamlined editing in the revised manuscript, especially given that this was a preliminary, and not a hypothesis-driven, analysis.
- Figure 2 and 2a: Please indicate on the figure what line that corresponds to what sex. It would also be beneficial to add 95%CIs to the figure to indicate at what points the two curves significantly depart, especially because they cross over. Where are the actual differences observed – I think this needs to be discussed further.
Our response: The figure legend defining which line corresponds to which sex was lost in formatting conversion in the submission portal; we added it back in, defined lines in the figure note, and added 95% confidence intervals. This figure is now Figure 1.
- Not sure what the +/-1 SD means in figure 3.
Our response: + 1 SD indicates one standard deviation above the mean GAD-7 score, Mean indicates mean GAD-7 score, and - 1 SD indicates one standard deviation below the mean GAD-7 score. We added this information to the figure caption. This figure is now Figure 2.
- Was any of the epidemiological analyses (not regarding the brain imagine part) corrected for multiple tests? I recommend this, as the authors conducted multiple comparisons.
Our response: We conducted two simultaneous models each for Aims 1 and 2. Because only two tests were performed in each aim, the risk of Type I error inflation is low. Given this and our relatively small sample size, we chose not to apply multiple-comparisons corrections in order to avoid increasing Type II error risk and because even applying a more conservative threshold (e.g., p<0.025) would not alter our primary conclusions.
- This study was conducted in a population of healthy adults, why was this population chosen? Is there an underlying hypothesis that alcohol consumption is a widespread coping mechanism in the general population? And should this be more frequent among one of the sexes according to the hypothesis?
Our response: Participants were healthy, heavy-drinking adults. We selected this population to minimize potential adverse effects related to study procedures (e.g., alcohol exposure, imaging). Despite the relatively healthy nature of the sample, drinking as a coping mechanism is common across the broader drinking spectrum (Corbin et al., 2013; Gilson et al., 2017; Holahan et al., 2001), and shows well-documented sex differences (Foster et al., 2014; Park & Levenson, 2002; Rodriguez et al., 2020). Thus, this sample provides a safe and appropriate context in which to examine our hypotheses while still capturing a meaningful variation in coping-related drinking and sex differences. We added a brief statement to this effect to the discussion (Page 13, Lines 468-473): “Second, because the study recruited healthy, heavy-drinking adults for safety during alcohol infusion and imaging, results may not generalize to lighter drinkers or individuals across the full AUD spectrum, despite evidence that coping-related drinking and its sex differences occur broadly (Corbin et al., 2013; Foster et al., 2014; Gilson et al., 2017; Holahan et al., 2001; Park & Levenson, 2002; Rodriguez et al., 2020).”
- In relation to this: How does your sample reflect the male:female ratio for AUD (normally regarded 75:25)– there is an oversampling of females in this study. How does that impact interpretation of results – could it even explain the lack of associations among males?
Our response: Our goal was not to mirror population-level AUD prevalence, but rather to ensure sufficient power to detect sex effects. Accordingly, our sample includes a relatively balanced distribution of women and men (55% women and 45% men in the full sample/56% women and 44% men in the imaging sample). This balance improves our ability to test sex differences compared to a sample that reflects the population ratio (approximately 75:25), which would limit our statistical power for male-female comparisons. Thus, the higher relative proportion of women is intentional and does not, on its own, explain the lack of associations observed among men. We added a brief statement to this effect to the discussion (Page 13, Lines 473-476): “Similarly, we intentionally balanced the sex ratio to maximize power for detecting sex effects. Therefore, our sample does not reflect population AUD prevalence, although the gender gap is narrowing (Agabio et al., 2017; Keyes et al., 2011; Verplaetse et al., 2025).”
- Finally, what implications does the study specific male:female ratio and the sample in general (healthy adults) have for generalizability?
Our response: We added a brief statement to this effect to the discussion (Page 13, Lines 473-476): “Similarly, we intentionally balanced the sex ratio to maximize power for detecting sex effects. Therefore, our sample does not reflect population AUD prevalence, although the gender gap is narrowing (Agabio et al., 2017; Keyes et al., 2011; Verplaetse et al., 2025).”
- Line 555-567: The authors describe that the incongruent results are due to previous studies relying on self-reports and therefore covering stress exposure over extended time periods while the present study assessed acute stress. Is the problem rather that this section compares two distinct concepts (acute vs chronic stress) rather than previous studies relying on self-reports and hence the differing results? Does males and females react similarly to acute and chronic stress?
Our response: We discuss the possibility that the study’s focus on acute stress likely contributes to differences between our findings and prior work (Page 11, Lines 384-389): “Additionally, the current study examined drinking in response to acute stress, which may be distinct from drinking associated with chronic stress and thus contribute unique clarification to sex differences in the relationship between anxiety and drinking. Prior work suggests that men may be more physiologically affected by chronic stress (Matthews et al., 2001), whereas women demonstrate greater neural responses to acute stress (Grodin et al., 2023).”
Other considerations:
- In this study, people were exposed to alcohol, was there conducted any follow-up assessments or checks on the participants
Our response: Participants who completed imaging sessions were assessed following their alcohol administration sessions. Because prior research indicates that laboratory alcohol administration is not associated with increases in subsequent drinking (Pratt & Davidson, 2005; Sommer et al., 2016), we did not conduct additional follow-up assessments.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe current article summarizes additional analyses of additional data collected from a a community sample of adults who completed an IV alcohol administration and fMRI procedures. The authors examined if sex moderated the mediating effects of subjective response to alcohol administration between anxiety and depression and stress-induced alcohol seeking. They found that sex moderated the association between anxiety and subjective stimulation response to alcohol and the association between anxiety and stress-induced alcohol seeking. They also found sex differences in functional connectivity between the hippocampus and amygdala to other networks. I can speak mostly to the modeling used in the mediated moderation models and less so to the fMRI results, so most of my comments focus on that aspect of the paper. Here are some suggestions to improve the readability of the paper and aid in interpretation.
Here are my overall comments:
- Overall the article was well written and thorough. The methods have several important strengths, including the use of IV alcohol self-administration.
- I felt that the paper was somewhat extensive and could be two separate studies, one with the subjective effects investigation and another with the fMRI results, but on the whole the authors did a good job of describing both parts of the paper.
Introduction
- I did feel that the introduction was fairly long and probably could have been trimmed in length. I think this is also a reflection of the fact that this was a fairly extensive investigation.
Methods:
- How did the authors define hat someone currently used alcohol? Within the past month?
- To what extent did baseline measures of alcohol use differ between the full sample, imaged, non-imaged, as well as by sex differences? Based on the Garrison 2025 paper, it looks like this data was collected.
- I know that the participant sample characteristics were summarized in the previous paper, but this information is very relevant for this paper and should be explicated in greater detail.
- Was a screening tool used to determine presence of alcohol use disorder or not?
Results:
- Figure 1 should include the Bs and 95% CI. The figure itself was useful as a key for the tables but I think that it would have helped to know what the overall estimates were for each line.
- I think including the simple slopes where the moderation was significant in the tables would have helped with interpretation as well. I know this information was included in text but it would have been helpful to have this information more readily available.
- Were covariates considered in the models, such as drinking history? I could see this as accounting for some of the alcohol response effect.
Discussion:
- It could be that the momentary subjective experience of anxiety might reduce alcohol seeking because the individuals are enticed to directly manage the stressor but, after conclusion of the stressor, drinking might be increased to manage residual anxiety related effects. In essence, how well someone is able to manage anxiety might map onto the GAD symptoms rather than the subjective experience of anxiety itself.
Author Response
Reviewer 2:
The current article summarizes additional analyses of additional data collected from a a community sample of adults who completed an IV alcohol administration and fMRI procedures. The authors examined if sex moderated the mediating effects of subjective response to alcohol administration between anxiety and depression and stress-induced alcohol seeking. They found that sex moderated the association between anxiety and subjective stimulation response to alcohol and the association between anxiety and stress-induced alcohol seeking. They also found sex differences in functional connectivity between the hippocampus and amygdala to other networks. I can speak mostly to the modeling used in the mediated moderation models and less so to the fMRI results, so most of my comments focus on that aspect of the paper. Here are some suggestions to improve the readability of the paper and aid in interpretation.
Here are my overall comments:
- Overall the article was well written and thorough. The methods have several important strengths, including the use of IV alcohol self-administration.
- I felt that the paper was somewhat extensive and could be two separate studies, one with the subjective effects investigation and another with the fMRI results, but on the whole the authors did a good job of describing both parts of the paper.
Introduction
- I did feel that the introduction was fairly long and probably could have been trimmed in length. I think this is also a reflection of the fact that this was a fairly extensive investigation.
Our response: We cut a significant amount of the introduction to streamline writing and improve clarity.
Methods:
- How did the authors define hat someone currently used alcohol? Within the past month?
Our response: For this sample, current alcohol use was defined as consumption of alcohol at least one time per week based on the 35-day Timeline Followback. On average, our participants drank alcohol 3.7 days/week (now shown in Table 2, Page 6, Line 250).
- To what extent did baseline measures of alcohol use differ between the full sample, imaged, non-imaged, as well as by sex differences? Based on the Garrison 2025 paper, it looks like this data was collected.
Our response: There were no significant differences in study variables across the imaged/non-imaged sample; we added Table 2 (Page 6, Line 250) to show these comparisons. There was only one difference across sex – pertaining to differences in baseline state anxiety – which we now discuss in Section 3.1 (Page 5, Lines 245-246). We added Table 3 (Page 6, Line 253) to show these comparisons.
- I know that the participant sample characteristics were summarized in the previous paper, but this information is very relevant for this paper and should be explicated in greater detail.
Our response: We added additional sample characteristics information to Section 3.1 (Page 5 Lines 241-249) and to the newly created Table 2 (Page 6, Line 250) and Table 3 (Page 6, Line 253).
- Was a screening tool used to determine presence of alcohol use disorder or not?
Our response: Participants completed the Semi-Structured Assessment of the Genetics of Alcoholism (Bucholz et al., 1994) at interview.
Results:
- Figure 1 should include the Bs and 95% CI. The figure itself was useful as a key for the tables but I think that it would have helped to know what the overall estimates were for each line.
Our response: Due to reviewer and editor comments regarding the utility of Figure 1, as well as requests to streamline and increase clarity, we deleted Figure 1.
- I think including the simple slopes where the moderation was significant in the tables would have helped with interpretation as well. I know this information was included in text but it would have been helpful to have this information more readily available.
Our response: We added simple slopes to Figure 1 (Page 7, Line 275) and Supplemental Figure 1 to aid interpretation.
- Were covariates considered in the models, such as drinking history? I could see this as accounting for some of the alcohol response effect.
Our response: We did not covary for drinking history because 1) IV alcohol administration procedures ensured equal alcohol exposure within and across participants, 2) the lack of significant sex differences in recent drinking history, and 3) due to our relatively small sample size. We added a statement concerning this to Page 13 (Lines 480-483): “Fifth, drinking history was not included as a covariate (due to limited sample size, the lack of significant sex differences in drinking history in this sample, and IV alcohol’s ability to ensure equal alcohol exposure within and across participants) and should be examined in future work.”
Discussion:
- It could be that the momentary subjective experience of anxiety might reduce alcohol seeking because the individuals are enticed to directly manage the stressor but, after conclusion of the stressor, drinking might be increased to manage residual anxiety related effects. In essence, how well someone is able to manage anxiety might map onto the GAD symptoms rather than the subjective experience of anxiety itself.
Our response: We’ve added this possible explanation to the discussion (Page 12, Lines 410-413): “Anxiety symptoms may reflect broader difficulties regulating anxiety that increase risk for drinking, whereas acute state anxiety in the laboratory may temporarily suppress alcohol-seeking behavior.”
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThanks for opportunity to review this manuscript. I appreciate the markedly improved clarity of the manuscript. I only have a few minor points, of which especially point 2 and 3 seems crucial to address:
1. In the abstract and the introduction (e.g., p. 1 line 14, or p. 2 line 82) the author states:
The overarching goal of this study is to examine how depression and anxiety symptoms relate to stress-induced alcohol-seeking across men and women (..)
The word across just implies that both males and females are included, I think it could be stated clearer the authors examines males and females separately.
- Regarding hypothesis (e.g., p.1, line 94) it is unclear what the sub-hypotheses means (H1a—H3b).
For example, H1a and H1b could have one of the following interpretations:
H1a: Females show strong positive associations between anxiety and depression with alcohol-seeking
H1b: Males show a non-significant association between anxiety and depression with alcohol-seeking
OR
H1a: Females show a stronger positive association between anxiety and alcohol-seeking than males
H1b: Females show a stronger positive association between depression and alcohol-seeking than males
This needs to be clearer for hypotheses H1a—H3b.
- On the background of the point estimate for figure 1 (b=13.90, p=0.04 for females, and b=8.11, p=0.22 for males) and the point estimate for the interaction term (b=22.01, p=0.02) the authors correctly conclude there is a sex difference. However, this sex difference only seems to exist for GAD-7 scores above 8—10 (according to figure 1). Hence, there is only an increased risk for alcohol-seeking among females with moderate to high GAD scores. Perhaps it could even be protective to be female with a lower score. This need to be mentioned in the result section and discussed more thoroughly in the discussion.
The authors conclude the following:
Women may benefit from interventions targeting anxiety symptoms to reduce stress-induced drinking (…). (p. 13, line 491—492)
Given Figure 1, I think this overstates the results. Maybe the authors should rather state that females with moderate to high anxiety symptoms constitutes a risk group.
Author Response
Please see the attachment.
Author Response File:
Author Response.docx