Review Reports - Barriers to Care Among LGBT Cancer Survivors: An Analysis of the All of Us Research Program

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for the opportunity to review this very interesting and necessary manuscript.

The large sample sizes for the analysis of barriers in healthcare utilisation and potentially related psychosocial factors within subgroups of the LGBTQ+ community offer insightful results. Indeed, the heterogeneity within the LGBTQ+ group is largely unaccounted for within this literature so this report is very welcome.

I will point out some areas of improvement which will improve this manuscript further.

Some very general and minor points. Please review manuscript to:

refer consistently to LGBTQ+ or LGBT - currently they are used interchangeably;
small typos (e.g. row 340 - which 'may' be due to...).

Introduction: I believe this is clear and comprehensive with points well made as to the need for this analysis and report.

Methods:

I would like to see an assessment and discussion around potential reporting bias (even if there is none), namely:

How many people within the sample did not respond to the questions related to sexual orientation and & gender identity and hence were not included in the analyses?
How many people responded without a history of cancer and hence were not included in the analyses?

A more generic methodological question here is also, if you could comment please upon it:

How did you ensure respondents actually had a history of cancer when this was reported?

Related to the questionnaires used:

To what extent did you have to adapt any of the questions to the group? I refer here, first, specifically to the healthcare utilisation barriers questions and the one referring to taking less medicines. To what extent is it that perhaps people took less medicines because of interference with gender affirming interventions? In results for instance - row 237 - you state that TGE participants had higher rates of skipping doses, delaying refills and asking for lower cost options - but it is unclear how we could differentiate between those trying to avoid negative effects of specific medicines (hormonal therapies for instance) on their transition versus lack of finances vs discriminatory behaviours of some type vs perhaps lack of discussions on multi-medication interactions due to indeed biased healthcare behaviour?

Response: We did not adapt any of the healthcare utilization barrier measures for this study. All healthcare utilization items, including those related to medication use (e.g., skipping doses, delaying refills, requesting lower-cost options), were administered as binary variables in the original All of Us survey. The available data do not allow for distinguishing underlying motivations for medication-related behaviors. As such, we cannot differentiate whether reported behaviors reflect financial barriers, discriminatory experiences, concerns about medication interactions (including with gender-affirming therapies), or other clinical or contextual factors. We have revised the Limitations sections to explicitly acknowledge this measurement constraint and to caution against attributing medication nonadherence to any single mechanism, particularly among transgender and gender-expansive participants.
- Page 26, “Additionally, medication-related healthcare utilization barriers were assessed using binary, self-reported items, which do not capture the underlying reasons for nonadherence. As a result, we were unable to distinguish whether reported behaviors reflected financial barriers, experiences of discrimination, concerns about medication interactions (including with gender-affirming therapies), or other clinical or contextual factors, particularly among TGE survivors.”

Everyday discrimination scale - you state that items were adapted - can you provide clarification as to what was adapted and why?

Response: Thank you for this thoughtful question. We did not adapt the Everyday Discrimination Scale for this study; rather, we used the items implemented by the All of Us Research Program. Relative to the original validated scale, the All of Us survey excluded one item (“In your day-to-day life, how often do any of the following things happen to you? You are followed around in stores”). In addition, whereas the original scale asked respondents to select from predefined categories when identifying the perceived reason for discriminatory experiences, the All of Us survey collected this information using open-text responses. These modifications reflect survey design decisions made by the AoU Research Program and were not introduced by the study team.

Results:

What are the differences between these 4 LGBTQ+ groups and other cancer patients in terms of the demographics, clinical characteristics and results on the questionnaires?

Response: We appreciate this thoughtful comment and agree that contextualizing findings relative to the broader cancer survivor population is important. However, the primary objective of this study was to explicitly examine heterogeneity of barriers to healthcare utilization within LGBT cancer survivors, rather than to compare LGBT survivors with cisgender and heterosexual cancer survivors, or LGBT people without a personal cancer history. Substantial prior literature has already established disparities between LGBTQ+ and cisgender/heterosexual populations in healthcare access and outcomes. Our goal was to address a different and less studied gap – namely, whether healthcare utilization barriers and psychosocial experiences differ meaningfully within LGBT subgroups, which are often treated as homogeneous in survivorship research.
That said, we have taken several steps to strengthen contextual interpretation. We expanded descriptive reporting within Tables 1 and 2 to describe age distributions, cancer types, and survivorship characteristics across LGBT subgroups. We also added discussion noting that observed differences in cancer type prevalence (e.g., breast cancer among lesbian survivors) are consistent with known sex-linked cancer patterns and may reflect underlying sex assigned at birth distributions within subgroups, including among transgender and gender-expansive participants. Importantly, we caution against attributing such differences to sexual orientation or gender identity alone.
- Results Page 11, “These subgroup differences highlight important variation in age, education, and cancer type distributions within LGBT cancer survivors, factors that may shape healthcare needs and utilization independent of sexual orientation or gender identity.”
- Discussion Page 21, “These findings should be interpreted as reflecting differential burden within existing cancer survivorship care systems rather than as evidence that LGBTQ+ identity itself causes healthcare utilization barriers.”

Table 2 results - are the p's adjusted for the multiple outcomes tested?

Response: Table 2 was meant to describe bivariate associations between sexual orientation, gender identity, and individual healthcare utilization barriers. Table 2 was not intended to provide multivariable results with adjustments.
- Because many of the healthcare barriers are conceptually and empirically correlated, strict family-wise error corrections (e.g., Bonferroni) may be overly conservative and inappropriate. Therefore, we present unadjusted p-values and emphasize effect patterns rather than individual hypothesis tests. We revised the Results text accompanying Table 2 to clarify its descriptive and exploratory purpose. Specifically, we removed inferential language (e.g., “significant,” “revealed”) and references that could be interpreted as confirmatory hypothesis testing. The revised text now emphasizes comparative prevalence patterns across sexual orientation and gender identity groups, with interpretation focused on overall trends rather than individual p-values. We also consolidated redundant statements and removed bullet-point summaries to avoid overemphasizing statistical testing. Finally, we added clarifying language indicating that findings from Table 2 are interpreted descriptively and that items with sparse cell counts are interpreted cautiously.
  - Page 11, “Table 2 presents descriptive comparisons of healthcare utilization barriers and cost-related behaviors across lesbian, gay, bisexual, and transgender/gender expansive (TGE) cancer survivors. Overall, bisexual and TGE respondents consistently reported higher prevalence of access barriers and unmet healthcare needs in the past 12 months compared with lesbian and gay respondents.
  - Delays in care related to transportation were more commonly reported by bisexual (15.2%) and TGE (21.4%) survivors than by lesbian (5.6%) and gay (8.0%) survivors. Similarly, inability to take time off work was more frequently reported among TGE (20.9%) and bisexual (11.6%) respondents relative to lesbian and gay respondents.
  - Cost-related barriers were prevalent across all groups but were notably more common among bisexual and TGE survivors. Nearly 30% of TGE survivors and 21.5% of bisexual survivors reported delaying care due to out-of-pocket costs, compared with approximately 13% among lesbian and gay survivors. Inability to afford prescription medications was also more frequently reported among TGE (29.4%) and bisexual (19.9%) respondents than among lesbian (8.7%) and gay (11.6%) respondents.
  - TGE survivors additionally reported higher levels of unmet need for mental health care (25.4%), dental care (30.4%), eyeglasses (31.3%), and specialty care (22.9%) compared with other groups. Cost-related medication behaviors followed a similar pattern, with TGE and bisexual respondents more frequently reporting delayed prescription fills, skipped doses, and requests for lower-cost medications.
  - Several items, including childcare-related barriers and caregiving responsibilities for another adult, were reported by relatively few respondents in certain groups and are therefore interpreted cautiously. Taken together, these descriptive patterns indicate a disproportionate burden of financial and structural healthcare barriers among bisexual and TGE cancer survivors.”

Figure 1 - please edit the figure to ensure the percentages can be better viewed.

Response: Thank you for edit suggestions for better clarity, we have improved Figure 1.

Figure 2 - seems to be missing.

Response: This figure will no longer be referenced and removed from the manuscript.

Discussion:

Response: Thank you for raising this important distinction. We do not interpret these findings as evidence that LGBTQ+ status itself causes healthcare utilization barriers. Rather, sexual orientation and gender identity are used in this study as stratifying characteristics to identify differential exposure to barriers within the broader context of cancer survivorship care. Many of the barriers observed, such as affordability, access to specialty care, and medication costs, reflect well-documented systemic challenges in survivorship care delivery. Our findings indicate that these system-level deficiencies disproportionately burden bisexual and transgender/gender expansive survivors. We have revised the manuscript to clarify that the analysis is descriptive and comparative and does not attempt to disentangle LGBTQ+-specific effects from broader structural limitations of survivorship care.
- Page 21, “These findings are within-group comparisons among LGBT cancer survivors they are not evidence that LGBT survivors experience greater healthcare utilization barriers than the general cancer survivor population. Many of the barriers observed may reflect broader systemic challenges in cancer survivorship care. The present analysis describes healthcare utilization barriers and explores how they are differentially experienced across LGBT subgroups. Sexual orientation and gender identity do not cause healthcare utilization barriers; however, healthcare and other systems, policies, and practices function in ways that may expose LGBT cancer survivors to conditions that result in healthcare utilization barriers.”

Response: We appreciate this important clarification and note that this issue is addressed in our response Reviewer 1’s related comment about adaptation of healthcare utilization measures. As described above, prescription-related barriers were assessed using general, binary items in the All of Us survey and were not specific to medication type or indication. Consequently, the data does not allow us to distinguish whether reported affordability challenges reflect cancer-related treatments, gender-affirming therapies, insurance coverage limitations, or other clinical or contextual factors. We have revised the Limitations section to explicitly acknowledge this measurement constraint and to caution against attributing prescription-related barriers to any single mechanism, particularly among transgender and gender-expansive survivors.

Response: The directionality of associations between psychosocial factors and healthcare utilization barriers warrants further investigation. Due to the cross-sectional design of the available data, we are unable to determine causal pathways in the current study. We revised the Discussion to more explicitly highlight this inverse relationship and emphasized that future research should involve longitudinal designs, temporally ordered measures, and linkage with clinical and pharmacy data to disentangle whether psychosocial stressors precede healthcare barriers or emerge as consequences of barriers.
- Page 23, “However, the cross-sectional nature of these data precludes determining whether psychosocial adversity precedes healthcare utilization barriers or emerges because of sustained structural barriers to survivorship care. These bidirectional pathways likely operate simultaneously and reinforce one another, underscoring the complexity of survivorship experiences among LGBT subgroups.
- Future research should prioritize longitudinal and mixed-methods designs to clarify the temporal ordering between psychosocial stressors and healthcare utilization barriers among LGBT cancer survivors. Linking repeated measures of discrimination, stress, and social support with clinical, pharmacy, and insurance data would allow investigators to distinguish whether psychosocial adversity drives disengagement from care or whether persistent structural barriers erode support networks and exacerbate stress over time. Such approaches are critical for identifying intervention points that effectively disrupt these reinforcing cycles.”

Clinical and policy implications:

This section can be improved alongside the reporting of results comparing these subgroups with the general cancer sample.

Response: Thank you for the opportunity to clarify. We have revised the Discussion section as follows:
- Page 24, “LGBTQ+ cancer survivors are a diverse and growing population; in 2024 the American Cancer Society (Lesbian, Gay, Bisexual, Transgender, Queer (LGBTQ+) People and Cancer Fact Sheet for Health Care Professionals, n.d.) estimated that 160,000 LGBTQ + people were diagnosed with cancer and nearly 50,000 would die from cancer. The National Cancer Institute (Baker, 2025) (Baker, 2025), American Society of Clinical Oncology (ASCO) (Kamen et al., 2022), The American Cancer Society (Lesbian, Gay, Bisexual, Transgender, Queer (LGBTQ+) People and Cancer Fact Sheet for Health Care Professionals, n.d.), the National Comprehensive Network (Mullins et al., 2025) and others (Cathcart-Rake et al., 2024; Kamen et al., 2022) all recommend regular and systematic SOGI data collection in cancer care. In their 2022, report the National Academies demonstrated the feasibility and acceptability of SOGI data collection by patients (Committee on Measuring Sex, Gender Identity, and Sexual Orientation et al., 2022). SOGI data collection is a foundational requirement for monitoring LGBTQ+ cancer survivors’ unique needs and experiences in oncology practice and research (Baker, 2025; Kamen et al., 2022). SOGI data collection is also vital to building trust and rapport between patients and providers (Baker, 2025) and for ensuring every patient receives the most appropriate care and treatment. However, to avoid the risk of increasing discrimination committed against LGBTQ+ cancer survivors, SOGI data collection must occur in concert with several key systems considerations, including anti-discrimination policies and consequences for discriminatory behaviors, identifying and utilizing best practice SOGI measures, health system readiness, electronic health record updates, mandatory and ongoing cultural humility training for providers and staff, and patient education concerning why SOGI data are collected, as well as privacy and confidentiality protections (Baker, 2025).

Thank you for the opportunity to review this paper and I hope that it gets published soon.

Reviewer 2 Report

Comments and Suggestions for Authors

Introduction: The authors should consider citing a more recent publication by Boehmer et al (2023) on “Cancer survivors' health behaviors and outcomes: a population-based study of sexual and gender minorities” which explicitly states that there is a paucity of data on cancer survivorship in this population and there is a gap as to whether health outcomes differ within subgroups of this population. This current manuscript contributes to filling this gap.

Minor: The use of the word "typology" is quite excessive in the same paragraph. The authors should consider a synonymous word.

Research Design/Statistical Analysis: I recommend logistic regression analyses to control for sociodemographic factors including age, race, etc. While a lot of these factors can confound the results of this study, the bivariate analyses used does not control for the effects of such factors. For instance, it is easy for one to think that the disparities in HCU within groups was due to age instead of sexual orientation. Unless the authors are able to prove beyond doubt that the agglomerative hierarchical cluster analysis they utilized was able to control for these effects, I suggest they conduct the logistic regression analysis.

Results and Conclusion: The authors should please proofread the manuscript to correct some minor errors. For instance…lines 201 and 202…”Skin cancer was most common among gay survivors (52.9%), followed by lesbians (27.5%) and bisexuals (37.7%)" should rather be “Skin cancer was most common among gay survivors (52.9%) followed by bisexuals (37.7%) and lesbians (27.5%)". Line 168 missed a citation and where is Figure 2….. line 302?

Author Response

Reviewer 2

Introduction: The authors should consider citing a more recent publication by Boehmer et al (2023) on “Cancer survivors' health behaviors and outcomes: a population-based study of sexual and gender minorities” which explicitly states that there is a paucity of data on cancer survivorship in this population and there is a gap as to whether health outcomes differ within subgroups of this population. This current manuscript contributes to filling this gap.

Response: The introduction has been revised to cite this publication.
- Page 2, “Recent population-based research has explicitly highlighted the lack of survivorship data for sexual and gender minority populations and identified a critical gap in understanding heterogeneity across subgroups defined by sexual orientation and gender identity(Boehmer et al., 2013). Few studies use large national datasets to examine how psychosocial stressors and healthcare utilization barriers co-occur or cluster among LGBT cancer survivors.”

Minor: The use of the word "typology" is quite excessive in the same paragraph. The authors should consider a synonymous word.

Response: Thank you for catching the redundancy, the repeated “typology” uses have been converted to synonyms.

Research Design/Statistical Analysis: I recommend logistic regression analyses to control for sociodemographic factors including age, race, etc. While a lot of these factors can confound the results of this study, the bivariate analyses used does not control for the effects of such factors. For instance, it is easy for one to think that the disparities in HCU within groups was due to age instead of sexual orientation. Unless the authors are able to prove beyond doubt that the agglomerative hierarchical cluster analysis they utilized was able to control for these effects, I suggest they conduct the logistic regression analysis.

Response: We have now conducted a multinomial logistic regression analysis modeling healthcare utilization (HCU) barrier cluster membership (low, moderate, high) as the outcome, with the low-barrier cluster as the reference category. The models were adjusted for age at analysis, age at diagnosis, sexual orientation, gender identity, race, education, income, marital status, perceived discrimination, stress, social support, and quality of life. This approach allowed us to assess whether observed subgroup differences in HCU burden persisted after accounting for key demographic and psychosocial confounders.
- Results indicate that sexual orientation remained a significant predictor of cluster membership after adjustment, independent of age and other sociodemographic factors. In particular, lesbian and gay survivors had significantly lower odds of membership in the moderate and high HCU barrier clusters compared with bisexual survivors, while income, discrimination, stress, and quality of life were also robustly associated with higher HCU burden. Importantly, age at analysis was associated with lower odds of high-barrier cluster membership, suggesting that age alone does not explain the observed disparities across sexual orientation subgroups.
- These findings demonstrate that the heterogeneity identified through cluster analysis is not solely attributable to age or other sociodemographic differences and that subgroup disparities persist even under multivariable adjustment. We have incorporated these results into the Results section and included full model estimates in the Supplementary Materials.
- Results section has been revised,
  - Pages 8-10, “Multinomial Logistic Regression Predicting Healthcare Utilization Barrier Clusters
  - Table 5 presents results from the multinomial logistic regression examining demographic, socioeconomic, and psychosocial correlates of membership in the moderate- and high–HCU barrier clusters relative to the low-barrier reference group. Several factors were independently associated with elevated odds of high-barrier cluster membership, while others showed limited or no association after multivariable adjustment.
  - Sexual orientation remained a consistent and independent predictor of HCU burden. Compared with bisexual survivors, both lesbian and gay survivors exhibited lower odds of membership in higher-barrier clusters. Gay survivors had significantly lower odds of belonging to the high-barrier cluster (AOR = 0.65, 95% CI: 0.43–0.98), while lesbian survivors had significantly reduced odds of placement in the moderate-barrier cluster (AOR = 0.61, 95% CI: 0.45–0.83). Although the point estimate for lesbian survivors in the high-barrier cluster also suggested reduced odds, this association did not reach statistical significance. Collectively, these findings reinforce descriptive patterns indicating that bisexual survivors are disproportionately represented among those experiencing greater structural and cost-related barriers to care.
  - Income demonstrated a strong and specific association with high-barrier status. Higher income was associated with substantially lower odds of membership in the high-barrier cluster (AOR = 0.81, 95% CI: 0.75–0.89), reflecting a pronounced socioeconomic gradient in severe HCU burden. In contrast, income was not significantly associated with membership in the moderate-barrier cluster, suggesting that financial resources most clearly differentiate survivors facing the most intensive access challenges rather than intermediate levels of burden.
  - Psychosocial stressors emerged as some of the most robust predictors of cluster membership. Each one-unit increase in everyday discrimination score was associated with higher odds of belonging to both the moderate-barrier (AOR = 1.03, 95% CI: 1.01–1.04) and high-barrier clusters (AOR = 1.04, 95% CI: 1.02–1.06). Perceived stress demonstrated an even stronger and graded relationship: higher stress levels were associated with increased odds of placement in the moderate-barrier cluster (AOR = 1.02, 95% CI: 1.00–1.04) and markedly higher odds of membership in the high-barrier cluster (AOR = 1.06, 95% CI: 1.03–1.09). These results align with earlier cluster-level findings showing that survivors facing the greatest access barriers also report the highest cumulative psychosocial strain.
  - In contrast, social support did not independently predict cluster membership. After adjustment for discrimination, stress, income, and other covariates, social support coefficients were near null for both moderate- (AOR = 1.00, 95% CI: 0.99–1.01) and high-barrier clusters (AOR = 0.99, 95% CI: 0.98–1.01). This suggests that while social support varies descriptively across identity groups and clusters, it does not independently offset the effects of structural and psychosocial stressors driving HCU burden.
  - Gender identity showed elevated but non-significant associations with higher-barrier cluster membership. TGE survivors demonstrated higher point estimates for both moderate (AOR = 1.35) and high-barrier clusters (AOR = 1.33), though confidence intervals were wide and crossed unity. These findings likely reflect limited statistical power rather than absence of effect and are directionally consistent with descriptive analyses showing disproportionate healthcare access challenges among TGE respondents.
  - Race, education, and marital status generally did not exhibit consistent or statistically significant associations with cluster membership after multivariable adjustment. Educational attainment below an advanced degree was associated with lower odds of high-barrier membership for those with a high school education or less (AOR = 0.52, 95% CI: 0.28–0.97), though this pattern did not extend uniformly across education categories. Racial comparisons did not demonstrate statistically significant differences in high-barrier membership, and marital status was not independently associated with cluster placement.
  - Overall, the regression results identify a concentrated pattern of vulnerability characterized by bisexual identity, lower income, and elevated discrimination and stress. These factors remained strong predictors of HCU burden even after adjustment for demographic characteristics and social support, underscoring a healthcare access landscape shaped less by interpersonal resources and more by cumulative structural and psychosocial disadvantage among LGBT cancer survivors.”

Results and Conclusion: The authors should please proofread the manuscript to correct some minor errors. For instance…lines 201 and 202…”Skin cancer was most common among gay survivors (52.9%), followed by lesbians (27.5%) and bisexuals (37.7%)" should rather be “Skin cancer was most common among gay survivors (52.9%) followed by bisexuals (37.7%) and lesbians (27.5%)". Line 168 missed a citation and where is Figure 2….. line 302?

Response: We thank the reviewer for noting this error. We have revised the sentence to reflect the correct ordering of prevalence across groups.
- Thank you for noting these issues. The missing citation at line 168 has been added. Figure 2 has been removed from the manuscript, as the information it conveyed is fully presented in Table 4, and all references to Figure 2 have been deleted accordingly.

Reviewer 3 Report

Comments and Suggestions for Authors

This is exciting work to see come out of the All of Us Research program and a relatively large study of LGBTQ cancer survivors with numerous psychosocial metrics to analyze. The article focuses on providing a typology for barriers to healthcare utilization in a population of LGBTQ cancer survivors disaggregated by sexual orientation or gender identity. The language shows sensitivity and consideration with only a few instances that could be reworded (e.g. replace "targeted" with "tailored" throughout, use of person-first language to replace "bisexuals" with "bisexual cancer survivors"). The introduction touches on the concept of applying an intersectional framework. However, it isn't clear what intersections are being included in this study. Specifically, I would like more discussion on the notable differences in age (there is nearly a decade difference between some groups), race, and gender. For example, the financial burden of cancer and healthcare is dramatically different for someone age 65 (and eligible for Medicare) than for someone age 55. Lesbian women were 91% white, can some of these differences in barriers be explained by fewer experiences of racial discrimination? I did not find Figure 2, was it included?

Major comments:

Although the use of an agglomerative hierarchical cluster analysis feels novel, I am not sure what this contributes that a more simple tertile of the summation of HCU barriers could not?
Is it valid to treat all barriers equally? Especially because there is some redundancy particularly for questions regarding medication.
Was there information on timing of cancer diagnosis? How long had it been since participants had been diagnosed with cancer?
Could Table 3 include the possible ranges of values for each measure?

Minor comments:

Line 168 the citation is missing.
Did you assume that all respondents who identified as gay were male?
Line 259 says that TGE individuals reported the lowest scores across nearly all support domains yet gay cancer survivors had the lowest mean.

Author Response

Reviewer 3

his is exciting work to see come out of the All of Us Research program and a relatively large study of LGBTQ cancer survivors with numerous psychosocial metrics to analyze. The article focuses on providing a typology for barriers to healthcare utilization in a population of LGBTQ cancer survivors disaggregated by sexual orientation or gender identity. The language shows sensitivity and consideration with only a few instances that could be reworded (e.g. replace "targeted" with "tailored" throughout, use of person-first language to replace "bisexuals" with "bisexual cancer survivors"). The introduction touches on the concept of applying an intersectional framework. However, it isn't clear what intersections are being included in this study. Specifically, I would like more discussion on the notable differences in age (there is nearly a decade difference between some groups), race, and gender. For example, the financial burden of cancer and healthcare is dramatically different for someone age 65 (and eligible for Medicare) than for someone age 55. Lesbian women were 91% white, can some of these differences in barriers be explained by fewer experiences of racial discrimination? I did not find Figure 2, was it included?

Response: We have removed text that communicates that any intersectional approaches were used.
- To address this concern, we have expanded the Discussion to more explicitly acknowledge how differences in age distribution, racial composition, and gender identity across subgroups may contribute to observed patterns of healthcare utilization barriers.
- Page 22, “Observed differences in healthcare utilization barriers across LGBT subgroups may reflect variation in age distributions, racial composition, and gender identity; however, this analysis did not apply formal intersectional methods capable of fully capturing overlapping identities. In multinomial models adjusting for age, race, income, education, and marital status, sexual orientation remained a consistent independent predictor of HCU burden, with bisexual cancer survivors showing disproportionately higher odds of membership in both moderate- and high-barrier clusters. Income emerged as a strong protective factor against high-barrier membership, highlighting a pronounced socioeconomic gradient in severe healthcare access challenges.
- Psychosocial stressors demonstrated robust and graded associations with HCU burden. Higher levels of everyday discrimination and perceived stress were independently associated with increased odds of placement in higher-barrier clusters, even after adjustment for sociodemographic factors. In contrast, social support did not independently predict cluster membership and did not attenuate associations for discrimination, stress, or income, suggesting limited buffering capacity in the context of sustained psychosocial and structural strain.
- Gender identity showed elevated but non-significant associations for transgender and gender-expansive survivors, likely reflecting limited precision due to smaller subgroup sizes. Race, education, and marital status did not demonstrate consistent independent associations after adjustment, despite descriptive differences across subgroups. Taken together, these findings indicate that sexual orientation, income, and cumulative psychosocial stressors were the most consistent predictors of healthcare utilization barriers, outweighing demographic composition alone in shaping survivorship access patterns among LGBT cancer survivors.”
Major comments:

Although the use of an agglomerative hierarchical cluster analysis feels novel, I am not sure what this contributes that a more simple tertile of the summation of HCU barriers could not?

Response: We appreciate the reviewer’s thoughtful question regarding the added value of hierarchical cluster analysis compared with a simple summed score of healthcare utilization (HCU) barriers. While a summed or tertiled index captures the quantity of barriers experienced, the cluster-based approach was intentionally selected to capture qualitatively distinct patterns of barriers that are not interchangeable or equally salient in practice.
Specifically, the agglomerative hierarchical clustering approach identifies groups of survivors who experience co-occurring configurations of barriers (e.g., affordability-driven barriers versus widespread structural and cost-related barriers), rather than assuming that all barriers contribute equally and additively to a single latent construct. Our results demonstrate that survivors in the high-barrier cluster are not merely those with “more” barriers, but those experiencing systematically different and more severe combinations of cost, access, and delay-related obstacles. These patterns are clinically and policy-relevant, as they imply different intervention targets than would be suggested by a simple barrier count.
Importantly, the cluster solution revealed non-linear distributions of barriers across groups that would likely be obscured by tertile-based categorization of a summed score. Thus, the clustering approach provides additional insight into heterogeneity in healthcare access challenges that extends beyond barrier burden alone.

Is it valid to treat all barriers equally? Especially because there is some redundancy particularly for questions regarding medication.

Response: This is a very thoughtful concern regarding whether it is appropriate to treat all HCU barriers equally, particularly given some conceptual overlap among medication-related items. This consideration was central to our analytic decision.
The purpose of the cluster analysis was not to estimate a weighted latent severity score, but rather to identify patterns of co-occurring barriers experienced by cancer survivors. In this context, barriers were treated equally to avoid imposing a priori assumptions about their relative importance, severity, or clinical impact, assumptions that are not well established in the literature, particularly for LGBT cancer survivor populations. Assigning differential weights would require strong empirical or theoretical justification, which is currently lacking for many access and affordability barriers.

Was there information on timing of cancer diagnosis? How long had it been since participants had been diagnosed with cancer?

Response: We appreciate this question. Direct measures of time since cancer diagnosis were not available in the All of Us survey data used for this analysis. However, to partially address timing-related differences in survivorship experiences, we incorporated age at cancer diagnosis (categorized age group at diagnosis) as a proxy measure.
Age group at diagnosis was added to the descriptive and bivariate analyses, as well as to the multinomial logistic regression models examining healthcare utilization barrier cluster membership. This allowed us to account for variation related to life stage at diagnosis, which may influence insurance coverage, healthcare access, and survivorship needs, even in the absence of precise time-since-diagnosis data.
We have clarified this approach and its limitations in the Methods and Discussion sections, noting that age at diagnosis serves as an imperfect proxy and that future studies with longitudinal follow-up are needed to directly examine survivorship trajectories over time.
- Methods Page 5, “Direct measures of time since cancer diagnosis were not available in the All of Us survey data. To partially account for timing-related differences in survivorship experiences, we incorporated age at cancer diagnosis (categorized age group at diagnosis) as a proxy measure. Age group at diagnosis was included in descriptive analyses, bivariate comparisons, and multinomial logistic regression models to account for variation related to life stage at diagnosis, which may influence insurance coverage, healthcare access, and survivorship needs.”

Could Table 3 include the possible ranges of values for each measure?

Response: Thank you for this suggestion, the value range has been added to the table for each measure.

Minor comments:

Line 168 the citation is missing.

Response: Thank you for this catch, it has been added in.

Did you assume that all respondents who identified as gay were male?

Response: We did not. However, during descriptive analysis it was observed that 95% of gay survivors identified as males.

Line 259 says that TGE individuals reported the lowest scores across nearly all support domains yet gay cancer survivors had the lowest mean.

Thank you for noting this important clarification. At the item level, gay cancer survivors frequently reported higher values across several individual social support domains. However, the composite social support score was calculated as an individual-level average across all available items, which resulted in a slightly lower overall mean for gay respondents when aggregated.
Importantly, gay cancer survivors also had the lowest mean scores for daily discrimination and perceived stress, indicating lower levels of these stressors, whereas transgender and gender-expansive (TGE) survivors reported the highest mean levels of discrimination and stress. With respect to social support, gay cancer survivors had the lowest overall composite mean, followed closely by TGE survivors.
- We have revised the Results section to explicitly distinguish between item-level patterns and composite score interpretations to avoid confusion.
  - Pages 14-15, “As shown in Table 4, transgender/gender expansive (TGE) and bisexual cancer survivors reported significantly higher composite scores for daily discrimination and perceived stress, and lower scores for social support, compared to lesbian and gay peers. On the discrimination composite, TGE survivors had the highest mean score (15.63, SD=10.56), followed by bisexual survivors (10.44, SD=8.71), while lesbian and gay survivors reported substantially lower values (7.68, SD=6.67 and 7.77, SD=7.92, respectively).
  - To illustrate item-level patterns within these composites, TGE respondents reported the highest mean frequency for nearly every discrimination item, including being threatened or harassed (m=1.31, SD=1.31), being treated with less respect (m=2.26, SD=1.41), and being perceived as dishonest (m=1.29, SD=1.49). Bisexual respondents also reported elevated frequencies of mistreatment compared to lesbian and gay participants, while lesbians generally reported the lowest levels of daily discrimination experiences.
  - Composite perceived stress scores followed a similar pattern: TGE survivors reported the highest mean (19.15, SD=7.99), bisexual survivors were also elevated (17.11, SD=8.56), and lesbian and gay survivors reported lower values (13.21, SD=7.60 and 13.21, SD=7.83). Item-level responses reflected these differences, with TGE survivors more frequently endorsing feelings of nervousness or stress (m=2.55, SD=1.14), lack of control over important things (m=2.11, SD=1.15), and inability to cope with demands (m=2.01, SD=1.07). Bisexual respondents showed similar elevated stress profiles compared to lesbian and gay respondents.
  - Social support composites also revealed disparities. Lesbian survivors reported the highest overall support (61.27, SD=10.73), while gay (57.69, SD=12.77), bisexual (59.51, SD=12.02), and TGE (58.34, SD=11.98) survivors reported lower values. Item-level indicators showed that lesbians consistently reported the highest perceived support, such as having someone to help if confined to bed (m=3.05, SD=1.14), take them to the doctor (m=3.23, SD=1.03), or assist with daily chores (m=2.93, SD=1.27). In contrast, TGE individuals reported the lowest scores across nearly all support domains. TGE and bisexual participants were also more likely to report feelings of isolation, lack of companionship, and having no one to turn to, while lesbian and gay individuals reported lower frequencies of these experiences.
  - Overall, composite scores and item-level patterns consistently demonstrated that TGE and bisexual cancer survivors experienced greater exposure to discrimination, elevated stress, and reduced social support compared to their lesbian and gay counterparts.”

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The responses are very nicely written at help clarify some of my main concerns regarding the article. With these responses, I would be in favor of accepting the article with minor revisions. A few requests:

Include the proportions of male and female for each SOGI group. There is ample evidence that barriers vary for bisexual females versus males, and similarly for transgender individuals.
Why could time since diagnosis not be calculated by simply taking the difference between age at diagnosis at age at the time of the survey?
Education does have a surprising and consistent pattern showing that lower education appears protective of higher barriers. It seems worth more discussion.

Author Response

Dear Academic Editor and Reviewer,

Thank you for your review of our manuscript, “Intersecting Barriers: Healthcare Access and Psychosocial Burden Among LGBT Cancer Survivors” for possible publication in your journal, Cancers. We have carefully considered reviewer feedback and revised our manuscript accordingly, these revisions are highlighted in the most recently submitted manuscript. A point-by-point response is provided below.

Reviewer 3

Comment: Include the proportions of male and female for each SOGI group. There is
ample evidence that barriers vary for bisexual females versus males, and
similarly for transgender individuals.
- Response: We agree that sex differences within sexual and gender minority subgroups are an important consideration. In response, we have revised Table 1 to present descriptive characteristics stratified by sex assigned at birth (female vs male) and report the proportions of individuals assigned female and male at birth within each SOGI group. Additionally, due to significant differences based on cross-classifying sex, sex assigned at birth was included in the Multinomial Regression Model. The Results section for Table 1 and Multinomial Regression have been revised as follows:
  - Page 11, “Sex assigned at birth varied substantially across sexual orientation and gender identity groups. Lesbian survivors were predominantly assigned female at birth (96.2%), whereas gay survivors were predominantly assigned male at birth (96.2%). In contrast, bisexual and TGE survivors demonstrated greater heterogeneity: 70.8% of bisexual survivors and 61.7% of TGE survivors were assigned female at birth, while 28.3% of bisexual and 34.9% of TGE survivors were assigned male at birth. Categories representing intersex or nonresponse were rare and suppressed due to small cell sizes.”
  - Results Pages 18-19, “Sexual orientation remained associated with HCU burden. Compared with bisexual survivors, lesbian survivors had significantly lower odds of moderate-barrier membership (AOR = 0.54, 95% CI: 0.39–0.75) and high-barrier membership (AOR = 0.58, 95% CI: 0.36–0.95). No statistically significant differences were observed for gay survivors relative to bisexual survivors. These findings reinforce descriptive patterns indicating that bisexual survivors are disproportionately represented among those experiencing greater structural and cost-related barriers to care. Sex assigned at birth was independently associated with HCU barrier membership. Compared with survivors assigned female at birth, those assigned male at birth had significantly lower odds of membership in both the moderate- (AOR = 0.66, 95% CI: 0.46-0.94) and high-barrier clusters (AOR = 0.53, 95% CI: 0.30-0.94). Estimates for individuals classified as intersex, none of these, or nonresponse were imprecise and not statistically significant due to small cell size”
- Due to limited sample sizes and All of Us privacy constraints, we were unable to estimate fully stratified interaction models (e.g., bisexual females vs bisexual males or transgender females vs transgender males) due to model failure from sparse group sizes. Nonetheless, the combination of sex-stratified descriptive analyses, and adjusted multinomial models provides a robust assessment of LGBT survivorship profiles while appropriately accounting for sex composition and avoiding overinterpretation of sparse subgroups.
Comment: Why could time since diagnosis not be calculated by simply taking the
difference between age at diagnosis at age at the time of the survey?
- Response: Time since diagnosis could not be calculated because age at diagnosis was not available as a continuous variable in the survey data. The only diagnosis-related timing variable available was age group at diagnosis, reported in broad categorical bands (e.g., 0–11, 12–17, 18–64, ≥65 years). As a result, subtracting age group from continuous age at survey would produce imprecise and potentially misleading estimates of time since diagnosis.
- To partially account for survivorship timing, we had included age group at diagnosis as a covariate in descriptive analyses and regression models. We have clarified this limitation in the Discussion section.
  - Pages 24-25, “Additionally, survivorship timing could not be modeled directly because age at diagnosis was not available as a continuous variable in the survey data. The only diagnosis-related timing measure available was age group at diagnosis, reported in broad categorical bands (e.g., 0–11, 12–17, 18–64, ≥65 years). As a result, time since diagnosis could not be calculated reliably by subtracting age group from continuous age at the time of survey, as doing so would yield imprecise and potentially misleading estimates. To partially account for survivorship timing, age group at diagnosis was included as a covariate in descriptive analyses and multivariable models.”
- Comment: Education does have a surprising and consistent pattern showing that
  lower education appears protective of higher barriers. It seems worth more
  - Response: We have expanded the Discussion to address this finding in greater depth. Importantly, the observed association should not be interpreted as evidence that lower educational attainment reduces healthcare barriers.
  - An alternative explanation may be that cancer survivors with advanced degrees may have higher expectations for care coordination, engage more intensively with healthcare systems, and encounter greater administrative complexity through increased use of specialty care, surveillance, and appeals processes. These patterns may increase the likelihood of identifying and reporting healthcare utilization barriers. In contrast, individuals with lower educational attainment may disengage earlier or normalize similar obstacles, resulting in lower reported barrier burden despite comparable structural constraints.
  - We now explicitly discuss this finding within frameworks of expectation–reality mismatch, healthcare system exposure, and minority stress, emphasizing that education does not function as a protective socioeconomic proxy once income is accounted for. This expanded discussion is included in the revised Discussion section.
    - Page 20, “The unexpected associations observed for income and education in the adjusted models warrant brief clarification. Higher income was consistently associated with reduced the odds of membership in the high-barrier cluster, which may underscore the importance of financial resources in buffering against utilization barriers such as affordability-related access challenges. In contrast, having a high school education or less was associated with lower odds of high-barrier membership relative to advanced degree holders, after accounting for income and psychosocial factors. This pattern may suggest that lower levels of educational attainment may reflect differences in healthcare system engagement, expectations, or exposure to administrative complexity. Importantly, these findings should not be interpreted as lower education being protective, but rather as suggesting the limits of educational attainment in offsetting structural and psychosocial barriers once financial resources are considered.”

Author Response File: Author Response.docx