Association Between Oral Antihyperglycemic Medications and Erectile Function in Men with Type 2 Diabetes Mellitus

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper is well written and outlines a good number of participants across almost all oral Anti-DM oral medications with outlining the limitations.

Overall glucose control intuitively seems to make logical sense in association with degree of erectile dysfunction. Did the authors find that a particular OHA is better than another in glucose control or were there any baseline parameters difference between the included patient group ?

• What is the main question addressed by the research? 

 

The paper has looked at the incidence of ED associated with DM management with different classes of OHA. The authors found that there is not any significant difference between the incidence of ED in patients treated with different OHA but rather the overall control of DM.

• 
  Do you consider the topic original or relevant to the field? Does it
  address a specific gap in the field? Please also explain why this is/ is not
  the case.

 

This paper helps us understand that it does not matter with which OHA we treat DM, the incidence of ED is related to overall BSL control and this should be the primary aim rather than the class of OHA used in the management.

• 
  What does it add to the subject area compared with other published
  material?

 

The class of OHA used to treat DM does not seem to influence the occurrence and severity of ED.

• 
  What specific improvements should the authors consider regarding the
  methodology?

 

The authors have identified the limitations in the paper which can be addressed in future research in the subject.

• 
  Are the conclusions consistent with the evidence and arguments presented
  and do they address the main question posed? Please also explain why this
  is/is not the case.

 

Yes, the conclusions are consistent with the evidence.

• 
  Are the references appropriate?

 

Yes 

Author Response

Comment 1:

"Overall glucose control intuitively seems to make logical sense in association with degree of erectile dysfunction. Did the authors find that a particular OHA is better than another in glucose control or were there any baseline parameters difference between the included patient group?"

Response 1:

Thank you for this insightful question. We address both aspects separately below.

(1) Baseline parameter differences between patient groups

Baseline characteristics stratified across the three glycemic control groups are presented in Supplementary Table 1. As described in Section 3.1, lines 191–200, beginning " Baseline characteristics across the three glycemic control groups are presented in Supplementary Table 1. The three groups were well-balanced across most clinical and laboratory parameters. A significant difference was observed only for diabetes dura-tion (F = 3.595, p = 0.030) and total OAD types (F = 30.693, p < 0.001), with the poorly controlled group having longer diabetes duration and taking more medications. Medi-cation distribution also differed significantly across groups for several OHA classes. Further supporting the presence of confounding by indication, HbA1c levels were sig-nificantly higher among users of sulfonylureas, thiazolidinediones, and GLP-1 receptor agonists compared with non-users, consistent with preferential prescribing to patients with poorer glycemic control (Supplementary Table 2)", the three groups were largely balanced across most clinical and laboratory parameters. Statistically significant differences were observed in diabetes duration (F = 3.595, p = 0.030) and total number of OAD types (F = 30.693, p < 0.001), with the poorly controlled group having a longer diabetes duration and a higher number of concurrent medications, consistent with clinical expectations.

(2) Whether a particular OHA demonstrated superior glycemic control

Our cross-sectional design does not allow for direct comparison of glycemic efficacy across medication classes. Patients were not randomized to treatment, and prescribing patterns reflect clinical decision-making rather than controlled allocation. Therefore, any comparison of HbA1c levels across OHA classes in our dataset would reflect patient selection rather than drug efficacy.

As shown in Supplementary Table 2, and discussed in Section 3.1, lines 198–202, beginning "Further supporting the presence of confounding by indication, HbA1c levels were significantly higher among users of sulfonylureas, thiazolidinediones, and GLP-1 receptor agonists compared with non-users, consistent with preferential prescribing to patients with poorer glycemic control (Supplementary Table 2)", these agents were preferentially prescribed to patients with poorer glycemic control. This underscores the importance of our analytical approach, which controlled for baseline HbA1c in all medication regression models to isolate medication-specific effects from concurrent glycemic status.

Head-to-head comparisons of glycemic efficacy across OHA classes are beyond the scope of this study and would require a prospective randomized design.

 

We sincerely thank Reviewer 1 for the careful reading of our manuscript and the constructive and encouraging comments.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript entitled “Association Between Oral Antihyperglycemic Medications and Erectile Function in Men with Type 2 Diabetes Mellitus” investigates the relationship between different classes of oral antihyperglycemic medications and erectile function in men with type 2 diabetes.

The manuscript is well structured and follows a conventional scientific format with clear sections for introduction, methods, results, and discussion. The use of the validated IIEF-5 for assessing erectile function and the attempt to adjust for several metabolic confounders strengthen the methodology. The classification of patients according to glycemic control patterns represents an interesting approach that attempts to distinguish medication effects from the influence of metabolic control.

The background section could be strengthened by including references to major clinical guidelines or consensus statements regarding erectile dysfunction in patients with diabetes. The rationale for the study could be more clearly emphasized by highlighting the methodological difficulty of separating medication effects from glycemic control.

The cross-sectional nature of the study limits causal inference. Because medication exposure and erectile function were assessed at the same time, it is not possible to determine whether medication use influenced erectile function or whether underlying disease characteristics influenced treatment selection. 

Several methodological aspects require clarification. The manuscript does not clearly describe how medication exposure was defined, including treatment duration, dosage, or combination therapy patterns. Many patients with type 2 diabetes receive multiple drugs simultaneously, these factors may influence the interpretation of the results. Testosterone levels and the frequency of HbA1c measurements used for glycemic classification would also improve transparency.

The main result that glycemic control shows a significant association with erectile function while individual medication classes do not. However, some sections of the results contain extensive numerical detail that largely repeats information already shown in the tables. Streamlining these descriptions would improve readability.

Minor improvements of tables could include clearer labeling of axes and reference groups, as well as inclusion of sample sizes within figure panels.

A few typographical errors are present in the abbreviation list and some sentences in the results and discussion sections could be reviwed.

Previous studies have already explored the relationship between antidiabetic medications and erectile function. The present work contributes additional observational data, particularly by examining multiple medication classes while considering longitudinal glycemic control patterns. The findings largely reinforce existing knowledge that metabolic control plays a key role in erectile dysfunction among diabetic patients.

The conclusions should be expressed more cautiously. Given the cross-sectional design, limited statistical power for some medication groups, and potential residual confounding, the findings should be interpreted as absence of detected association in this cohort rather than definitive evidence that medications have no direct effect on erectile function.

 

Author Response

Comment 1:

"The background section could be strengthened by including references to major clinical guidelines or consensus statements regarding erectile dysfunction in patients with diabetes. The rationale for the study could be more clearly emphasized by highlighting the methodological difficulty of separating medication effects from glycemic control."

Response 1:

We thank the reviewer for this valuable suggestion. In response, we have added citations to the American Diabetes Association (ADA) Standards of Care (now References 20 and 21) to provide guideline-level context for erectile dysfunction assessment in patients with diabetes. We have also expanded the discussion of the methodological challenge of separating individual medication effects from background glycemic control, noting that many prior studies relied on single cross-sectional HbA1c measurements or failed to account for longitudinal glycemic patterns. These revisions are reflected in the Introduction, lines 74–89, beginning “Current clinical guidelines, including the American Diabetes Association (ADA) Standards of Care, emphasize the comprehensive assessment of diabetes-related complications and comorbidities, including those affecting quality of life and sexual health, particularly in patients with cardiovascular risk factors or long-standing disease [20,21]. However, no major guideline has issued specific recommendations regarding the differential effects of individual oral antihyperglycemic agent classes on erectile function, reflecting a notable gap in the current evidence base. A major limitation of existing studies is the inability to disentangle the independent effects of antihyperglycemic medications from glycemic control itself. Because these medications directly influence glycemic status, and glycemic control is a key determinant of erectile function, observed associations may reflect indirect effects mediated through glycemic improvement rather than direct pharmacologic effects. Many prior studies have not adequately addressed this issue, often relying on single cross-sectional HbA1c measurements or failing to account for longitudinal glycemic patterns. Additionally, most existing studies have been conducted in Western populations, with limited data from Asian cohorts, who may exhibit distinct metabolic characteristics and treatment response patterns”. 

 

Comment 2:

"The cross-sectional nature of the study limits causal inference. Because medication exposure and erectile function were assessed at the same time, it is not possible to determine whether medication use influenced erectile function or whether underlying disease characteristics influenced treatment selection."

Response 2:

We agree with this important limitation. The cross-sectional nature of the study, in which medication exposure and erectile function were assessed simultaneously, precludes causal inference and cannot exclude the possibility that underlying disease characteristics influenced treatment selection. This has been explicitly acknowledged in the Limitations section (Section 4.7, line 379-381), beginning "Several limitations warrant consideration. First, the cross-sectional design precludes causal inference, as medication exposure and erectile function were assessed simultaneously. Prospective longitudinal studies are needed to establish directionality ", where we also note that prospective longitudinal studies are needed to establish directionality. The Conclusions section further emphasizes that findings should be interpreted as the absence of a detected association rather than definitive evidence of no effect.

 

Comment 3:

"The manuscript does not clearly describe how medication exposure was defined, including treatment duration, dosage, or combination therapy patterns. Many patients with type 2 diabetes receive multiple drugs simultaneously, these factors may influence the interpretation of the results. Testosterone levels and the frequency of HbA1c measurements used for glycemic classification would also improve transparency."

Response 3:

Thank you for raising these important methodological points. We address each in turn:

(1) Medication exposure definition:

We thank the reviewer for raising this important point. We have revised the Methods section to clarify that medication exposure was defined as current use at the time of IIEF-5 questionnaire completion, recorded as a binary variable, as updated in the Methods, lines 124–129, beginning "Medication exposure was defined as current use at the time of IIEF-5 questionnaire completion, recorded as a binary variable without consideration of dose, duration, or combination therapy patterns. Participants were classified as users of a given medication class if the medication was prescribed and actively being taken at the time of questionnaire administration, as documented in the electronic medical records". We acknowledge this represents a limitation, as it does not capture treatment intensity, duration of exposure, or polypharmacy interactions. This has been explicitly stated in the revised Limitations section (Section 4.7, lines 388–396), beginning " Sixth, medication exposure was defined as current use at the time of questionnaire completion, without information on treatment duration, dosage, or combination ther-apy patterns, which may not fully capture cumulative pharmacological exposure. Furthermore, as patients may concurrently receive multiple drug classes, overlapping medication use may introduce residual confounding between treatment groups, de-spite the use of separate regression models for each medication class. Taken together, these limitations preclude definitive conclusions regarding direct pharmacological ef-fects of individual OHA classes on erectile function". Regarding combination therapy, each medication class was examined in a separate fully adjusted regression model to address potential multicollinearity, as described in the Methods, lines 159-166, beginning "Multiple linear regression models evaluated associations between individual oral antihyperglycemic medications and IIEF-5 scores. To address potential multicollinearity arising from combination therapy, each medication was examined in a separate fully adjusted regression model rather than entering all medications simultaneously into a single model, controlling for age, diabetes duration, BMI, HbA1c, testosterone, insulin use, and comorbidities. Patients receiving multiple drug classes were included as users in each relevant medication model. Bonferroni correction (α=0.007) was ap-plied for seven medication comparisons."

(2) Testosterone levels:

Testosterone was measured as morning fasting serum total testosterone, as described in the Methods section, lines 130-131, beginning "Laboratory parameters included: fasting plasma glucose, HbA1c, morning fasting serum total testosterone...". We also wish to note that a unit error was identified during the review process: a subset of testosterone measurements had been recorded in ng/mL in the source records and were inadvertently entered without unit conversion, resulting in artificially low values in the original dataset. All testosterone records were rechecked against the original laboratory reports and corrected accordingly. The corrected median testosterone is 356.0 ng/dL (IQR 278.0–459.8), as now reported in Table 1 and Section 3.1. All related statistical analyses were rerun using the corrected values; the overall direction and statistical significance of all findings were unchanged after correction, confirming the robustness of the results. 

(3) Frequency of HbA1c measurements:

The Methods section, lines 134-135, beginning " All participants had a minimum of two HbA1c measurements available during this period...", states that all participants had a minimum of two HbA1c measurements available during the 12-month period. We have added a clarifying sentence in the revised manuscript to make this more explicit.

 

Comment 4:

"Some sections of the results contain extensive numerical detail that largely repeats information already shown in the tables. Streamlining these descriptions would improve readability."

Response 4:

We appreciate this suggestion. We have reviewed and streamlined Section 3.2, lines 218-223, beginning "Table 2 and Figure 1 present associations between individual medications and IIEF-5 scores. After covariate adjustment, no medication class demonstrated a statistically significant association with IIEF-5 scores (all p > 0.05). Even after Bonferroni correction (α = 0.007), significance was not achieved for any medication. The wide confidence interval observed for GLP-1 receptor agonists reflects the limited sample size for this subgroup (n = 13). Detailed regression coefficients and confidence intervals are presented in Table 2", and Section 3.3, line 235-242, beginning "ANCOVA revealed a significant association between glycemic control and IIEF-5 scores (F(2,192) = 3.390, p = 0.036). A graded relationship was observed, with well-controlled patients having higher scores than poorly controlled patients.

Post-hoc analysis showed a significant difference between well-controlled and poorly controlled groups (mean difference = 2.488, p = 0.032), while differences between well-controlled and variably controlled groups and between variably controlled and poorly controlled groups did not reach significance. These results are illustrated in Table 3 and Figure 2", to reduce redundancy with table contents, retaining only key numerical findings in the text while directing readers to the tables for complete details.

 

Comment 5:

"Minor improvements of tables could include clearer labeling of axes and reference groups, as well as inclusion of sample sizes within figure panels."

Response 5:

Thank you for this suggestion. We have revised Figures 1, 2, and 3 to include sample sizes (n) within each panel. The figure legend for Figure 3 (lines 283-289), beginning "Forest plots showing adjusted regression coefficients (β) and 95% confidence intervals for as-sociations between oral antihyperglycemic medications and IIEF-5 scores within glycemic control subgroups: (a) well-controlled group (n = 85); (b) variably controlled group (n = 93); (c) poorly controlled group (n = 64). β represents the difference in IIEF-5 score between medication users and non-users; positive values indicate better erectile function in users. The vertical reference line at β = 0 indicates no association. Models were adjusted for age, diabetes duration, BMI, HbA1c, testosterone, insulin use, and comorbidities", has also been updated to clarify subgroup sample sizes and reference groups.

 

Comment 6:

"A few typographical errors are present in the abbreviation list and some sentences in the results and discussion sections could be reviewed."

Response 6:

Thank you for pointing this out. We have carefully reviewed and corrected the following: (1) the Abbreviation list, which contained formatting inconsistencies and (2) Section 4.3, lines 328-336, beginning "Our study addressed this through a dual approach: (1) controlling for baseline HbA1c in medication regression models to isolate medication-specific effects from concurrent glycemic status; and (2) examining longitudinal glycemic control patterns (sustained good control, fluctuating control, or sustained poor control) to assess the impact of long-term metabolic status beyond what single time-point HbA1c measurements can capture. Our subgroup analysis further supports this conclusion. When medication associations were examined separately within each glycemic control group (well-controlled, variably controlled, and poorly controlled), the null findings persisted consistently across all subgroups", where the term "variably controlled" had incorrectly appeared as "moderately controlled" in one instance. The Results and Discussion sections has also been fully proofread.

 

Comment 7:

"The conclusions should be expressed more cautiously. Given the cross-sectional design, limited statistical power for some medication groups, and potential residual confounding, the findings should be interpreted as absence of detected association in this cohort rather than definitive evidence that medications have no direct effect on erectile function."

Response 7:

We thank the reviewer for this important suggestion. We agree that the cross-sectional design, limited statistical power for certain medication subgroups, and potential residual confounding all warrant appropriately cautious language. The Conclusions section (line 410–413), beginning "However, given the cross-sectional design, these results should be interpreted as the absence of a detected association rather than definitive evidence of no effect. Further prospective studies are needed to clarify the independent effects of individual medication classes on erectile function", explicitly states that these results should be interpreted as the absence of a detected association rather than definitive evidence of no effect. We have reviewed the full manuscript to ensure this cautious framing is applied consistently, and have softened any remaining language that could be interpreted as making definitive claims.

Regarding residual confounding, this has been explicitly acknowledged in the revised Limitations section (Section 4.7, lines 391–394), beginning "Furthermore, as patients may concurrently receive multiple drug classes, overlapping medication use may introduce residual confounding between treatment groups, despite the use of separate regression models for each medication class."

 

We believe the revised manuscript addresses all of Reviewer 2's comments and has been substantially improved as a result. We thank the reviewer for the detailed and constructive feedback.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors, I have revised the manuscript entitled “Association Between Oral Antihyperglycemic Medications and Erectile Function in Men with Type 2 Diabetes Mellitus.” Overall, the study addresses a relevant topic. However, the following revisions are suggested to improve the + quality of the manuscript.

Comments:

• Please use italics for “p” values.
• In the Materials and Methods section, please clarify the study design (retrospective or cross-sectional?) and provide additional details on the recruitment process and inclusion period.
• Please specify whether medication exposure refers to current use, long-term use, or use during the entire 12-month observation period.
• I would suggest adding additional information regarding possible confounding variables, such as smoking status, alcohol consumption, physical activity, or psychological factors.
• Given the small number of patients receiving GLP-1 receptor agonists, please comment more explicitly on the limited statistical power for this subgroup.
• The readability of the figures needs to be improved. Please revise it.
• I would suggest discussing the limitations of the study more explicitly, particularly the potential influence of unmeasured confounders.

Author Response

Comment 1:

"Please use italics for "p" values."

Response 1:

Thank you for this comment. We have carefully reviewed the entire manuscript and ensured that all p values are italicized consistently throughout the text, tables, and figure legends.

 

Comment 2:

"In the Materials and Methods section, please clarify the study design (retrospective or cross-sectional?) and provide additional details on the recruitment process and inclusion period."

Response 2:

Thank you for raising this point. The study is a cross-sectional study. This is stated in the Methods section, lines 99-101, beginning "This cross-sectional study was conducted at the out-patient clinic of the Endocrinology and Metabolism Department...", which also specifies the inclusion period (July 2020 to January 2021).

Regarding the recruitment process, the Methods section, lines 104-113, beginning "We enrolled men aged 18–80 years diagnosed with T2DM...", describing the inclusion and exclusion criteria, provides a comprehensive account of participant recruitment. Of 296 patients screened, 54 were excluded due to not meeting one or more of the above criteria, and 242 were included in the final analysis. We have added a brief clarification in the revised Methods section, to make the cross-sectional design more prominent at the opening of the section.

 

Comment 3:

"Please specify whether medication exposure refers to current use, long-term use, or use during the entire 12-month observation period."

Response 3:

Thank you for this important clarification. Medication exposure in this study refers to current use at the time of IIEF-5 questionnaire completion, recorded as a binary variable, as defined in the revised Methods section, lines 124-129, beginning "Medication exposure was defined as current use at the time of IIEF-5 questionnaire completion, recorded as a binary variable without consideration of dose, duration, or combination therapy patterns. Participants were classified as users of a given medication class if the medication was prescribed and actively being taken at the time of questionnaire administration, as documented in the electronic medical records". We acknowledge that this definition does not capture treatment duration, cumulative exposure, or dose information; this has been explicitly added as a limitation in the revised manuscript (Section 4.7, lines 388-394), beginning "Sixth, medication exposure was defined as current use at the time of questionnaire completion, without information on treatment duration, dosage, or combination therapy patterns, which may not fully capture cumulative pharmacological exposure. Furthermore, as patients may concurrently receive multiple drug classes, overlapping medication use may introduce residual confounding between treatment groups, despite the use of separate regression models for each medication class."

 

Comment 4:

"I would suggest adding additional information regarding possible confounding variables, such as smoking status, alcohol consumption, physical activity, or psychological factors."

Response 4:

We appreciate this suggestion. We acknowledge that smoking status, alcohol consumption, physical activity, and psychological factors such as depression and anxiety were not collected in this study and may have contributed to residual confounding. This has been noted in the Limitations section (Section 4.7, lines 386–388), beginning "Fifth, several potential confounders were not captured, including smoking status, alcohol consumption, physical activity, and psychological factors such as depression and anxiety, which may have contributed to residual confounding."

 

Comment 5:

"Given the small number of patients receiving GLP-1 receptor agonists, please comment more explicitly on the limited statistical power for this subgroup."

Response 5:

We agree that this deserves more explicit discussion. The limited statistical power for the GLP-1 receptor agonist subgroup (n = 13, 5.4%) is currently noted in two locations in the manuscript:

(1) Section 3.2, lines 221-223, beginning " The wide confidence interval observed for GLP-1 receptor agonists reflects the limited sample size for this subgroup (n = 13)", where we note the wide confidence interval for this subgroup reflecting the limited sample size.

(2) Section 4.7, lines 372-377, beginning "Second, the small number of GLP-1 receptor agonist users (n=13, 5.4%) substantially limited statistical power for this subgroup, and the study was underpowered to detect small-to-moderate effect sizes for this medication class. The null finding for GLP-1 receptor agonists should therefore be interpreted with caution, and larger prospective studies are needed to clarify its potential effects on erectile function", where we explicitly state the study was underpowered to detect small-to-moderate effect sizes for this subgroup and that the null finding should be interpreted with caution.

We note that the low prevalence of GLP-1 receptor agonist use in our cohort reflects real-world prescribing patterns in Taiwan, where access to this drug class is constrained by cost and National Health Insurance reimbursement criteria.

 

Comment 6:

"The readability of the figures needs to be improved. Please revise it."

Response 6:

Thank you for this feedback. We have revised all figures (Figures 1, 2, and 3) to improve readability. Specific improvements include (1) adding sample sizes (n) within each figure panel; (2) clarifying reference group annotations and confidence interval bars; and (3) enhancing text weight and color contrast for key elements. The revised figures are included in the updated manuscript.

 

Comment 7:

"I would suggest discussing the limitations of the study more explicitly, particularly the potential influence of unmeasured confounders."

Response 7:

We agree with this suggestion. As noted in our response to Comment 4 above, smoking status, alcohol consumption, physical activity, and psychological factors such as depression and anxiety were not collected in this study. These have been explicitly acknowledged as potential sources of residual confounding in the Limitations section (Section 4.7, lines 386–388), beginning "Fifth, several potential confounders were not captured, including smoking status, alcohol consumption, physical activity, and psychological factors such as depression and anxiety, which may have contributed to residual confounding." 

 

We believe the revised manuscript has been substantially improved in response to Reviewer 3's comments. We thank the reviewer for the detailed and constructive feedback.

Reviewer 4 Report

Comments and Suggestions for Authors

1. Study design clarification

The manuscript describes the study as a retrospective observational study, yet the IIEF-5 questionnaire appears to have been completed during clinical visits while laboratory data were extracted from medical records. Please clarify whether the study should be considered a retrospective cross-sectional analysis or a prospective observational study with retrospective data extraction, as this affects interpretation of potential bias.

2. Age distribution inconsistency

Although the inclusion criteria allow patients aged 18–80 years, the analyzed cohort includes individuals aged 29–60 years (mean 50.79 ± 6.99 years). Please clarify why older patients were not represented in the final cohort, as erectile dysfunction prevalence increases substantially after age 60 and this may affect external validity.

3. Limited sample size for GLP-1 receptor agonists

Only 13 patients (5.4%) were treated with GLP-1 receptor agonists. This very small subgroup limits statistical power and likely explains the wide confidence interval reported in the regression analysis. The conclusions regarding this medication class should therefore be interpreted cautiously.

4. Handling of multiple antidiabetic therapies

Patients with T2DM commonly receive combination therapy. The manuscript states that each medication was analyzed in a separate regression model, but it is not clearly described how overlapping treatments were handled. Please clarify whether patients receiving multiple drug classes were included in several models and whether multicollinearity between treatments was assessed.

5. Glycemic control categorization

Patients were categorized according to HbA1c patterns over 12 months, but the manuscript does not specify how many HbA1c measurements were required per patient or how missing values were handled. Providing this information would improve methodological clarity.

6. Testosterone interpretation

The reported median testosterone level of 218.8 ng/dL appears relatively low. Please clarify whether testosterone measurements were performed under standardized conditions (e.g., morning sampling) and whether hypogonadism thresholds were considered in the analysis.

7. Data presentation

For skewed variables such as testosterone and UACR, the table reports the median but does not provide interquartile ranges (IQR). Including dispersion measures would improve interpretation of the data.

8. Minor editorial issues

• In Table 1, the value reported for “Any insulin use” appears inconsistent with the numbers for basal and basal-bolus insulin.

• In the Abbreviations section, “Erectyle dysfunction” should be corrected to “Erectile dysfunction.”

9. Interpretation of conclusions

Given the cross-sectional design and limited sample size for some medications, the conclusion that oral antihyperglycemic drugs do not independently affect erectile function may be too strong. It may be more appropriate to state that no significant associations were observed in this cohort after adjustment for metabolic variables.

Comments on the Quality of English Language

The overall English language quality of the manuscript is generally understandable, but several sections require minor grammatical correction, improved phrasing, and consistency in scientific terminology. Some sentences are overly long or slightly awkward, and a few typographical errors are present. Careful language editing would improve readability and clarity.

Examples include:

1. In the Abbreviations section, the term “Erectyle dysfunction” should be corrected to “Erectile dysfunction.”

2. In the Results section, the sentence “A total 296 patients were enrolled” should be corrected to “A total of 296 patients were enrolled.”

3. In the Methods section, the phrase “statistical analysis were performed using Jamovi version 2.3” should be corrected to “statistical analyses were performed using Jamovi version 2.3.”

4. In several parts of the manuscript, sentences are overly long and could be improved by splitting them into shorter statements for clarity, particularly in the Discussion section.

5. Minor stylistic inconsistencies are present in the reporting of statistical results and abbreviations, which should be standardized throughout the manuscript.

Overall, the language is acceptable for peer review, but minor professional English editing is recommended before publication.

Author Response

Comment 1:

"The manuscript describes the study as a retrospective observational study, yet the IIEF-5 questionnaire appears to have been completed during clinical visits while laboratory data were extracted from medical records. Please clarify whether the study should be considered a retrospective cross-sectional analysis or a prospective observational study with retrospective data extraction, as this affects interpretation of potential bias."

Response 1:

Thank you for this important clarification. The study is a cross-sectional study. The IIEF-5 questionnaire was administered during routine outpatient clinic visits, while demographic, laboratory, and medication data were extracted retrospectively from electronic medical records for the 12-month period preceding each patient's IIEF-5 assessment. This design is best described as a cross-sectional study with retrospective extraction of clinical data. We have clarified this in the Methods section, lines 99, beginning "This cross-sectional study was conducted at the out-patient clinic of the Endocrinology and Metabolism Department..."

 

Comment 2:

"Although the inclusion criteria allow patients aged 18–80 years, the analyzed cohort includes individuals aged 29–60 years (mean 50.79 ± 6.99 years). Please clarify why older patients were not represented in the final cohort, as erectile dysfunction prevalence increases substantially after age 60 and this may affect external validity."

Response 2:

Thank you for raising this point. The narrower age distribution reflects real-world clinic demographics rather than intentional exclusion. Older patients may have been less represented for several reasons: (1) higher rates of incomplete medical records; (2) higher prevalence of current phosphodiesterase-5 inhibitor use; and (3) occurrence of acute diabetes complications within the preceding 3 months—both of the latter being exclusion criteria (Methods, lines 114-119, beginning "Exclusion criteria included... "). These factors are all more prevalent in older patients with long-standing T2DM and likely contributed to their underrepresentation in the final cohort. This is acknowledged in Section 4.7, lines 378-386, beginning " Fourth, the actual age distribution of our cohort (29–60 years) reflects the real-world demographic profile of men attending our diabetes outpatient clinic who met all eligibility criteria. Older patients may have been less represented due to several factors: higher rates of incomplete medical records; higher prevalence of current phos-phodiesterase-5 inhibitor use; and occurrence of acute diabetes complications within the preceding 3 months (with both of the latter being exclusion criteria). These factors are all more prevalent in older patients with long-standing T2DM. This may limit generalizability to older men with T2DM, in whom ED prevalence is substantially higher", where we note that this may limit generalizability to older men with T2DM, in whom ED prevalence is substantially higher.

 

Comment 3:

"Only 13 patients (5.4%) were treated with GLP-1 receptor agonists. This very small subgroup limits statistical power and likely explains the wide confidence interval reported in the regression analysis. The conclusions regarding this medication class should therefore be interpreted cautiously."

Response 3:

We fully agree. The limited statistical power for the GLP-1 receptor agonist subgroup (n = 13, 5.4%) is noted in two locations in the manuscript: (1) Section 3.2, lines 221–223, beginning "The wide confidence interval observed for GLP-1 receptor agonists reflects the limited sample size for this subgroup (n = 13)", where we note the wide confidence interval for this subgroup reflecting the limited sample size, and (2) Section 4.7, lines 372–377, beginning "Second, the small number of GLP-1 receptor agonist users (n=13, 5.4%) substantially limited statistical power for this subgroup, and the study was underpowered to detect small-to-moderate effect sizes for this medication class. The null finding for GLP-1 receptor agonists should therefore be interpreted with caution, and larger prospective studies are needed to clarify its potential effects on erectile function", where we explicitly state the study was underpowered to detect small-to-moderate effect sizes for this subgroup and that the null finding should be interpreted with caution. We note that the low prevalence of GLP-1 receptor agonist use in our cohort reflects real-world prescribing patterns in Taiwan, where access to this drug class is constrained by cost and National Health Insurance reimbursement criteria, rather than a sampling limitation of this study.

 

Comment 4:

"Patients with T2DM commonly receive combination therapy. The manuscript states that each medication was analyzed in a separate regression model, but it is not clearly described how overlapping treatments were handled. Please clarify whether patients receiving multiple drug classes were included in several models and whether multicollinearity between treatments was assessed."

Response 4:

We thank the reviewer for this insightful comment. To address potential multicollinearity arising from combination therapy, each medication class was examined in a separate, fully adjusted regression model rather than entering all medication classes simultaneously into a single model. Patients receiving multiple drug classes were categorized as "users" within each relevant medication-specific model.

We acknowledge that overlapping medication use may nonetheless introduce residual confounding between treatment groups. This inherent limitation of cross-sectional research has been explicitly addressed in the revised Limitations section (lines 391–394): "Furthermore, as patients may concurrently receive multiple drug classes, overlapping medication use may introduce residual confounding between treatment groups, despite the use of separate regression models for each medication class."

 

Comment 5:

"Patients were categorized according to HbA1c patterns over 12 months, but the manuscript does not specify how many HbA1c measurements were required per patient or how missing values were handled. Providing this information would improve methodological clarity."

Response 5:

Thank you for this suggestion. The Methods section, lines 130-135, beginning "Laboratory parameters included: fasting plasma glucose, HbA1c, morning fasting serum total testosterone, estimated glomerular filtration rate (eGFR) calculated using CKD-EPI equation, and urine albumin-to-creatinine ratio (UACR). HbA1c values were collected over a 12-month period preceding IIEF-5 assessment for glycemic control categorization", states that all participants had a minimum of two HbA1c measurements available during the 12-month period. Patients with incomplete laboratory or medication records were excluded at the screening stage, as specified in the exclusion criteria (Methods, lines 114-119, beginning "Exclusion criteria included: (1) age <18 or >80 years; (2) type 1 diabetes; (3) severe ED due to anatomical abnormalities or trauma; (4) current use of phosphodiesterase-5 inhibitors; (5) severe psychiatric disorders; (6) acute diabetes complications within 3 months; and (7) incomplete medical records.").

 

Comment 6:

"The reported median testosterone level of 218.8 ng/dL appears relatively low. Please clarify whether testosterone measurements were performed under standardized conditions (e.g., morning sampling) and whether hypogonadism thresholds were considered in the analysis."

Response 6:

We thank the reviewer for identifying this discrepancy. The value of 218.8 ng/dL cited by the reviewer corresponds to the original submitted manuscript, which contained a data error. During data preparation, a subset of testosterone measurements had been recorded in ng/mL in the source records due to a unit convention used at the time of collection, which was subsequently standardized to ng/dL. These values were inadvertently entered without unit conversion, resulting in artificially low testosterone values (divided by a factor of approximately 100) in the original dataset.

We sincerely apologize for the oversight and have corrected all testosterone records against the original laboratory reports accordingly. All testosterone-related data, statistical analyses, and figures were rerun using the corrected values. The corrected median testosterone is 356.0 ng/dL (IQR 278.0–459.8), as reported in Section 3.1, line 183, beginning "Laboratory parameters showed median testosterone 356.0 ng/dL...", and in Table 1. The overall direction and statistical significance of all findings were unchanged after correction, confirming the robustness of the results.

Regarding measurement conditions, testosterone was measured as morning fasting serum total testosterone, as specified in the Methods section, lines 130-131, beginning "Laboratory parameters included: fasting plasma glucose, HbA1c, morning fasting serum total testosterone...". Hypogonadism thresholds were not applied as a formal stratification variable, as the study focused on associations between glycemic control and erectile function. The non-significant association between testosterone and IIEF-5 scores (r = 0.038, p = 0.562) is reported in Section 3.4, lines 257-258, beginning " Pearson correlation analysis confirmed no significant association between serum testosterone levels and IIEF-5 scores (r = 0.038, 95% CI: −0.092 to 0.167, p = 0.562)."

 

Comment 7:

"For skewed variables such as testosterone and UACR, the table reports the median but does not provide interquartile ranges (IQR). Including dispersion measures would improve interpretation of the data."

Response 7:

Thank you for this suggestion. We have added IQR values for testosterone and UACR in the revised Table 1: testosterone is now reported as 356.0 (278.0–459.8) ng/dL and UACR as 9.0 (4.5–30.8) mg/g. The table footnote, lines 203-205, has also been updated to specify that non-normally distributed variables are presented as medians (interquartile range).

 

Comment 8:

"In Table 1, the value reported for "Any insulin use" appears inconsistent with the numbers for basal and basal-bolus insulin. In the Abbreviations section, "Erectyle dysfunction" should be corrected to "Erectile dysfunction.""

Response 8:

Thank you for identifying this inconsistency. We acknowledge that the original submitted manuscript contained an error in the percentage reported for "Any insulin use" and "Basal-bolus or mixed insulin". In the original version, the percentage for "Any insulin use" was incorrectly reported as 9.1% and "Basal-bolus or mixed insulin" as 8.6%, both of which were erroneous. The correct values, now reflected in the revised manuscript, are basal insulin: n = 29, 12.0%; basal-bolus or mixed insulin: n = 22, 9.1%; and any insulin use: n = 51, 21.1%, where 51/242 = 21.1%. These two subcategories are mutually exclusive and their sum correctly equals the total insulin users.

Regarding the typographical error, the Abbreviations section in the revised manuscript correctly reads "Erectile dysfunction".

 

Comment 9:

"Given the cross-sectional design and limited sample size for some medications, the conclusion that oral antihyperglycemic drugs do not independently affect erectile function may be too strong. It may be more appropriate to state that no significant associations were observed in this cohort after adjustment for metabolic variables."

Response 9:

We thank the reviewer for raising this important point. We fully agree that the cross-sectional design and limited sample size for certain medication subgroups preclude definitive conclusions. The Conclusions section (line 410–413), beginning "However, given the cross-sectional design, these results should be interpreted as the absence of a detected association rather than definitive evidence of no effect. Further prospective studies are needed to clarify the independent effects of individual medication classes on erectile function", explicitly states that "these results should be interpreted as the absence of a detected association rather than definitive evidence of no effect”. We have reviewed the full manuscript to ensure this cautious framing is applied consistently, and have softened any remaining language that could be interpreted as making definitive claims regarding the absence of pharmacological effects.

 

We believe the revised manuscript has been substantially improved in response to Reviewer 4's comments. We thank the reviewer for the detailed and constructive feedback.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The revised version of the manuscript shows clear improvement, particularly in the way the authors have tempered their interpretations and more appropriately framed the limitations of a cross-sectional design. The conclusions are now more cautious and better aligned with the presented data, which strengthens the overall credibility of the study. The manuscript addresses a clinically relevant question, and the analytical approach—especially the attempt to disentangle glycemic control from medication effects—is commendable. The discussion has also been expanded in a meaningful way, providing a more balanced interpretation of the findings within the context of existing literature. A few issues still require clarification before the manuscript can be considered for acceptance. There is some inconsistency in the description of covariates across the methods section, Table 2, and the figure legends, which should be harmonized to ensure transparency of the statistical model. The interpretation of null findings should more explicitly acknowledge the limited statistical power for certain medication classes, particularly GLP-1 receptor agonists. Minor clarification regarding the discrepancy between the stated age inclusion criteria and the actual age range of the cohort would also improve clarity.

Author Response

Comment 1:

“There is some inconsistency in the description of covariates across the methods section, Table 2, and the figure legends, which should be harmonized to ensure transparency of the statistical model.”

Response 1:

We thank the reviewer for this observation. Upon careful review, we identified the following inconsistencies and have corrected them accordingly:

(1) Methods section (line 164): The term “comorbidities” has been expanded to “comorbidities (hypertension, dyslipidemia, cardiovascular disease, and renal disease)” to explicitly specify the variables included in the model.

(2) Table 2 footnote (lines 229-230): The covariate list has been updated to consistently read: “Adjusted for age, diabetes duration, BMI, HbA1c, testosterone, ACR, eGFR, and comorbidities (hypertension, dyslipidemia, cardiovascular disease, and renal disease).”

(3) Figure 1 and Figure 3 legends (lines 241-243, lines 288-290): Both figure legends have been updated to match the Methods section and Table 2 footnote, replacing “insulin use” with “ACR, eGFR” and expanding “comorbidities” with the full parenthetical specification “comorbidities (hypertension, dyslipidemia, cardiovascular disease, and renal disease)”.

(4) Table 4 footnote (lines 260-262): The covariate list has similarly been updated to include HbA1c, ACR, and eGFR, and “comorbidities” has been expanded with the full parenthetical specification “comorbidities (hypertension, dyslipidemia, cardiovascular disease, and renal disease)”.

All four locations — the Methods section, Table 2 footnote, Figure 1 legend, and Figure 3 legend — now consistently read: “...adjusted for age, diabetes duration, BMI, HbA1c, testosterone, insulin use, and comorbidities (hypertension, dyslipidemia, cardiovascular disease, and renal disease).”

To further clarify this point, we have reviewed the relevant sections and confirmed that the current wording appropriately reflects this limitation.

 

Comment 2:

“The interpretation of null findings should more explicitly acknowledge the limited statistical power for certain medication classes, particularly GLP-1 receptor agonists.”

Response 2:

We thank the reviewer for this additional comment. We note that this concern was also raised by Reviewer 4 during Round 1 (Comment 3) and was addressed in the previous revision. The limited statistical power for the GLP-1 receptor agonist subgroup (n=13, 5.4%) is explicitly acknowledged in two locations in the current manuscript:

(1) Section 3.2 (lines 222–224), beginning “The wide confidence interval observed for GLP-1 receptor agonists reflects the limited sample size for this subgroup (n=13)”, where the wide confidence interval is noted as a direct consequence of the small sample size.

(2) Section 4.7 (lines 373–378), beginning “Second, the small number of GLP-1 receptor agonist users (n=13, 5.4%) substantially limited statistical power for this subgroup, and the study was underpowered to detect small-to-moderate effect sizes for this medication class. The null finding for GLP-1 receptor agonists should therefore be interpreted with caution, and larger prospective studies are needed to clarify its potential effects on erectile function”, where we explicitly state that the study was underpowered to detect small-to-moderate effect sizes for this medication class and that the null finding should therefore be interpreted with caution.

We have carefully reviewed the manuscript and believe that this limitation is sufficiently and clearly addressed.

 

Comment 3:

“Minor clarification regarding the discrepancy between the stated age inclusion criteria and the actual age range of the cohort would also improve clarity.”

Response 3:

We thank the reviewer for raising this point. We note that this issue was also identified by Reviewer 4 during Round 1 (Comment 2) and was addressed in the previous revision. The inclusion criteria specified age 18–80 years, while the actual cohort range was 29–60 years. This discrepancy reflects real-world clinic demographics rather than intentional exclusion. Older patients were likely underrepresented due to: (1) higher rates of incomplete medical records; (2) higher prevalence of current phosphodiesterase-5 inhibitor use; and (3) higher occurrence of acute diabetes complications within the preceding 3 months — the latter two being pre-specified exclusion criteria — all of which are more prevalent in older patients with long-standing T2DM.

This limitation is explicitly acknowledged in Section 4.7 (lines 379–387), beginning “Fourth, the actual age distribution of our cohort (29–60 years) reflects the real-world demographic profile of men attending our diabetes outpatient clinic who met all eligibility criteria. Older patients may have been less represented due to several factors: higher rates of incomplete medical records; higher prevalence of current phos-phodiesterase-5 inhibitor use; and occurrence of acute diabetes complications within the preceding 3 months (with both of the latter being exclusion criteria). These factors are all more prevalent in older patients with long-standing T2DM. This may limit generalizability to older men with T2DM, in whom ED prevalence is substantially higher”, where we note that this may limit generalizability to older men with T2DM, in whom ED prevalence is substantially higher.

We appreciate this important point and have confirmed that it is clearly addressed in the current manuscript.

 

We trust that the revisions have satisfactorily addressed the comments raised. We greatly appreciate the Editor’s and reviewer’s time and effort, and we look forward to your consideration of our revised manuscript.