Cognitive Behavioral Stress Management Effects on Cancer-Related Distress and Neuroendocrine Signaling in Breast Cancer: Differential Effects by Neighborhood Disadvantage

Abstract

Purpose.

Women residing in disadvantaged neighborhoods experience disparities in breast cancer (BC) survival which persist when accounting for individual-level socioeconomic/treatment factors. The chronic stress of living in a disadvantaged neighborhood may compound the stress of a new cancer diagnosis, leading to neuroendocrine dysregulation. Cognitive Behavioral Stress Management (CBSM) has shown efficacy at reducing distress and modulating neuroendocrine functioning, but it is unknown whether it is efficacious in this population.

Methods.

This is a secondary analysis of a randomized trial of 10-week group-based CBSM (versus a psycho-educational control) in women with nonmetastatic BC. The Area Deprivation Index (ADI) was calculated, and women were categorized as living in low (n=175) versus high disadvantage (n=50). Women completed a measure of cancer-related distress (Impact of Events Scale-Intrusions) and underwent blood draws to collect PM cortisol at baseline, 6-months, and 12-months. Hierarchical linear modeling tested whether condition predicted the slope of outcomes, and whether ADI moderated these relationships.

Results.

CBSM was associated with greater reductions in cancer-specific distress and cortisol, though these effects were not found to be moderated by ADI. Exploratory simple slope analyses showed that CBSM was associated with decreased cancer-related distress across ADI categories, while CBSM resulted in decreased cortisol among low ADI women only.

Conclusion.

CBSM reduced cancer-related distress across neighborhoods, but this was only accompanied by cortisol changes among those in advantaged neighborhoods. Neighborhood disadvantage may represent a particularly salient stressor that is distinct from cancer-specific distress. Future interventions targeting this population should consider modifications to increase relevance and accessibility.

Introduction

Significant advancements in the screening and treatment of nonmetastatic breast cancer over the past decades have resulted in a near 90% 5-year survival rate [1]; however, disparities in survival persist. Specifically, women with nonmetastatic cancer residing in under-resourced and disadvantaged neighborhoods are more likely to be diagnosed at later stage of disease [2], have more aggressive tumor subtypes [3, 4] and receive suboptimal (e.g., non-National Comprehensive Cancer Network [NCCN] concordant) treatment [5]. Even when accounting for these differences in screening and treatment, as well as individual-level socioeconomic factors, neighborhood disadvantage has been associated with worse breast cancer survival [5].

A diagnosis of breast cancer is often accompanied by significant psychological distress [6], which may lead to downstream physiological consequences, such as dysregulations in immune (e.g., increased inflammatory signaling) and neuroendocrine (hypothalamic pituitary adrenal [HPA] axis and sympathetic nervous system [SNS]) functioning [7-10]. Women living in disadvantaged neighborhoods may be especially vulnerable to these biological effects as the chronic stress of living in a marginalized neighborhood is compounded by the acute stress of a new cancer diagnosis [11]. For example, living in disadvantaged neighborhoods is associated with increased discrimination [12], toxicant exposure [13], and lack of access to healthcare services [14], all of which contribute to chronic stress. Chronic stress, in turn, results in dysregulation of the immune and neuroendocrine systems [7]. In the context of breast cancer, this may be reflected in changes in the tumor microenvironment that leads to more aggressive disease and worse survival [8, 15, 16]. This is supported by prior work among women with breast cancer demonstrating that greater neighborhood disadvantage is associated with greater inflammatory and SNS transcriptional activity in tumor samples [10], higher afternoon-evening (PM) serum cortisol in the postsurgical period [17], increased markers of biological aging, including allostatic load and global DNA methylation [18], and shorter recurrence-free survival, which was predicted by greater inflammatory and SNS expression in breast cancer tumors [10]. Together, this suggests that the chronic stress of living in disadvantaged neighborhoods gets “under the skin” and drives disparities in survival [17].

Cognitive behavioral stress management (CBSM) is a group-based psychosocial intervention aimed at decreasing the distress that accompanies a new breast cancer diagnosis [19]. Ten-session CBSM, which incorporates components of cognitive-behavioral therapy and relaxation training with an emphasis on cancer-specific stressors (e.g., appraisals and concerns about disease course, communicating with care team, coping with side effects), has resulted in enhanced quality-of-life and decreased distress as compared to a single-session psycho-education (PE) control seminar [20, 21]. Moreover, CBSM is associated with increased time to all-cause mortality, breast cancer mortality, and breast cancer recurrence over 11 years [22]. These survival effects may have been driven by biological changes associated with decreased stress: women receiving CBSM had significantly greater decreases in PM serum cortisol [23] and leukocyte metastatic and inflammatory gene expression [24] as compared to those receiving the PE control. Thus, CBSM may effectively dampen the effects of stress on cancer outcomes. However, it is unknown whether this intervention, which focuses primarily on cancer-specific stressors, is effective at reducing distress and downstream cortisol effects among women experiencing the additional chronic stress of living in marginalized neighborhoods.

For the current study, we used prior trial data to test in secondary analyses whether CBSM, as compared to a single-session PE control condition, had differential effects on cancer-related distress and PM serum cortisol levels among women in advantaged versus disadvantaged neighborhoods in South Florida. We hypothesized that CBSM would effectively enhance cancer-specific distress across both low and high disadvantaged neighborhoods, but that the intervention would only improve cortisol among women living in low disadvantage neighborhoods, as it may not sufficiently target the complex psychosocial stressors beyond cancer-related stress that women living in disadvantaged neighborhoods experience.

Methods

Study Design

This is a secondary analysis of a prior randomized controlled trial (RCT) investigating the effects of 10-week CBSM as compared to a 1-day psycho-educational (PE) control (ClinicalTrials.gov Identifier: NCT01422551). The original study was approved by an institutional review board at the University of Miami. All participants signed informed consent prior to participation. Survey data from three time points were used for the present analyses: post-surgery and prior to randomization into study conditions (T1), approximately 6 months (T2), and 12 months (T3) following study entry. Women also completed a blood draw at each of these timepoints to collect serum cortisol, as outlined below.

Participants

A total of 240 women who were 2-8 weeks post-surgery for nonmetastatic breast cancer were enrolled in the original trial. Exclusion criteria for the original study included lack of English language fluency, initiation of chemotherapy or radiation treatment, a history of prior cancer, severe psychiatric illness, an acute or chronic comorbid condition, or unwillingness to be randomized to study conditions.

Study Conditions

CBSM is a group-based, manualized 10-week intervention that incorporates both cognitive behavioral therapy (CBT) (e.g., cognitive restructuring, coping effectiveness training) and relaxation training (diaphragmatic breathing, guided imagery, progressive muscle relaxation, mindfulness) to target the distress that may accompany a new cancer diagnosis. The control condition is a 1-day psychoeducational group seminar on general health education (e.g., importance of healthy diet, treatment options) and basic descriptions of some stress reduction techniques akin to a self-help seminar. Additional details on the parent trial have been reported in detail elsewhere [25].

Measures

Participant characteristics.

Participants self-reported sociodemographic and medical information, which was verified by medical record review. For the current study, disease stage at diagnosis (0-III) and age at diagnosis were used as covariates in all models. Covariates were chosen due to both their demonstrated relationship with distress [26, 27] and cortisol [28, 29] in the context of breast cancer, and to be in line with prior analyses from this trial [17]. Covariates were kept to a minimum to avoid model overfitting and enhance parsimony [30].

Neighborhood disadvantage.

The Area Deprivation Index (ADI), a validated measure of neighborhood disadvantage [31] was calculated for women in the current trial using their provided home address at diagnosis and the ADI mapping atlas (https://www.neighborhoodatlas.medicine.wisc.edu/mapping). The ADI is comprised of several factors related to block-level socioeconomic status, including income/employment (e.g., income disparity; percentage employed population aged 16 years or older in white-collar occupations), housing (e.g., median home value; percentage owner-occupied housing unit), and household characteristics (e.g., percentage single-parent households with children younger than 18). For the current analyses, we used 2015 ADI for the state of Florida. State deciles are typically categorized as tertiles, with the lower tertile containing values of 1-3 representing the least disadvantaged [32]. To obtain adequate sample sizes for our planned analyses, we collapsed the highest two tertiles to compare low neighborhood disadvantage (ADI 1-3) versus high disadvantage (ADI 4-10), as has been done in our prior work [5, 17]. This categorization is supported by prior work in a sample of over 5,000 women with breast cancer, which demonstrated that women living in the top two most disadvantaged tertiles had worse survival as compared to the most advantaged tertile [5], suggesting that there is an effect of neighborhood disadvantage on health outcomes at both an intermediate and more severe level.

Cancer-specific distress.

Cancer-specific distress (e.g., unwanted thoughts regarding treatment, recurrence, and effects of cancer) was measured using the 7-item Impact of Events Scale- Intrusion Subscale (IES-I) [33]. The 7-item subscale assesses the degree to which women were distressed by thoughts surrounding their cancer in the past week (e.g., “Other things kept making me think about it”) on a Likert scale of Not at all (0) to Extremely (4). Responses were averaged for the subscale score. The internal consistency of the IES-I was high (alpha > .90).

Cortisol.

We collected blood samples via venipuncture for serum assays from participants at a single timepoint: in the late afternoon to evening (between 4:00-6:30 PM). This timing was chosen to control for circadian fluctuations and due to the theoretical importance of evening-time cortisol elevations. Cortisol follows a diurnal curve, peaking at awakening and declining throughout the day [34]. Thus, elevated cortisol levels in the evening may represent dysregulation of the typical diurnal curve. Peripheral blood was collected in red-topped vacutainer tubes, and serum was separated from cells when centrifuged [23]. Cortisol levels were calculated by competitive enzyme-linked immunosorbent assay (ELISA). Women were instructed to refrain from the consumption of alcohol, caffeine, or recreational drugs the day of collection.

Analysis

Hierarchical linear modeling (HLM) based on maximum likelihood estimation was used to test the role of CBSM receipt on the trajectory of psychological distress and cortisol over the first-year post-surgery, confirming prior reported results from the sample using alternative statistical methods (e.g., repeated measures ANOVA, latent growth modeling), and whether these effects are moderated by neighborhood disadvantage. HLM differs from traditional longitudinal analyses as it accounts for both between and within-person differences [30]. In addition, HLM does not assume an equal number of observations and uses all available data, and as such has increased power and less biased results [30].

A bottom-up modeling process was used for each model (Hox et al., 2018). We first ran an unconditional intercept only model with the random effect of person. We then added time from baseline assessment (coded as months since baseline) as a fixed effect to determine if there was systematic change in effect over the year following diagnosis. We used a likelihood ratio test (LRT) to determine whether this model significantly improved fit versus the unconditional model. We then used LRT to test random effects of linear time and quadratic time to determine the model structure. We then added study condition, the previously stated covariates, and the interaction of interest (time x condition) as fixed effects to the model. Finally, to test moderation by level of neighborhood disadvantage, a three-way interaction was added to the model as fixed effects (time x condition x ADI). In the case of a significant interaction, simple slopes were probed to aid interpretation.

Assumptions of homoscedasticity of variance and normally distributed level 1 and 2 residuals were visually assessed using plots of the final models, and all assumptions were met.

Results

Descriptive Statistics

Of the 240 women enrolled in the original trial, 225 provided home street addresses (e.g., versus a PO Box) from which the study team was able to collect metrics of neighborhood disadvantage, and thus were included in the present analyses. The mean age at diagnosis was 50.4 years (range 23-70), and the majority (63.4%) were non-Hispanic White. Most women were diagnosed with stage I (38.1%) or II (39.6%) disease. Though participants spanned the full range of ADI (1-10), women were predominantly living in more advantaged neighborhoods (low ADI=175, 77.8%). There was no difference in CBSM session attendance by ADI (t(103)=−.208, p=.836). As has been previously reported [17], higher ADI was associated with greater serum cortisol at baseline (t(120)=.235, p=.020). In addition, we found that higher ADI was associated with less cancer-specific stress at baseline (t(222)=1.80, p=.037). See further descriptive statistics in Tables 1 and 2. See Figure 1 for CONSORT flow diagram.

Table 1.

Baseline Medical and Demographic Variables by ADI

Variable	ADI 1-3 (n=175)	ADI 4-10 (n=50)	Test Statistic	P-Value
Age at diagnosis	50.76(8.88)	49.28(9.65)	t(223) = 1.02	.309
Race/Ethnicity
Non-Hispanic White	119(68.0%)	25(50.0%)	χ2(3) = 15.73	.001**
Hispanic	43(24.6%)	13(26.0%)
Black	9(5.1%)	11(22.0%)
Asian	4(2.3%)	0(0.0%)
Stage of Disease:			χ2(3) = 5.27	.153
Stage 0	24(13.7%)	10(20.0%)
Stage 1	72(41.1%)	13(26.0%)
Stage 2	62(35.4%)	24(48.0%)
Stage 3	15(8.6%)	3(6.0%)
Household Income (thousands)	83.32(73.85)	68.83(44.60)	t(223) = 1.26	.104
Years Education	15.74(2.46)	15.00(1.98)	t(223) = 1.96	.025*
CBSM Session Attendance	7.10(2.22)	7.21(2.87)	t(103) = −.208	.836
Cancer-Related Distress	14.17(8.17)	11.80(8.50)	t(222) = 1.80	.037*
Serum Cortisol (μg/dL)	7.95(4.13)	10.80(9.17)	t(120) =.235	.020*

Parameter estimate standard errors listed in parentheses

^**p<.01

^*p<.05.

Table 2.

Baseline Medical and Demographic Variables by Condition

Variable	PE (n=114)	CBSM (n=111)	Test Statistic	P-Value
Age at diagnosis	51.24(9.05)	49.60(9.03)	t(223) = 1.36	.177
Race/Ethnicity			χ2(3) = 4.05	.256
Non-Hispanic White	72(63.7%)	72(64.9%)
Hispanic	27(23.9%)	29(26.1%)
Black	10(8.8%)	10(9.0%)
Asian	4(3.5%)	0(0.0%)
Stage of Disease:			χ2(3) = 2.81	.420
Stage 0	18(15.8%)	16(14.4%)
Stage 1	48(42.1%)	37(33.3%)
Stage 2	40(35.1%)	46(41.4%)
Stage 3	7(6.1%)	11(9.9%)
Household Income (thousands)	80.62(69.75)	79.69(67.48)	t(196) = −.095	.925
Years Education	15.44(2.28)	15.71(2.53)	t(223) = .206	.469
Cancer-Related Distress	13.25(8.33)	14.05(8.24)	t(222) = −.720	.472
Serum Cortisol	7.78(3.45)	9.61(7.67)	t(121) = −1.73	.086

Parameter estimate standard errors listed in parentheses.

^**p<.01

^*p<.05.

Figure 1.

CONSORT Flow Diagram.

CBSM Effects on Cancer-Related Distress: Main and Moderator Effects

The final model of cancer-related distress included a fixed, linear effect of time. When adding the interaction of time x condition and covariates, we confirmed prior published findings [35] that women receiving CBSM had significantly greater decreases in cancer-related distress over the study period than did women receiving the control condition (B(SE)=−0.28(0.09), p=.003). There was also a main effect of age, such that younger women reported greater distress over the study period (B(SE)=−0.16(0.05), p=.001). The three-way interaction of ADI x condition x time was added to the model, and did not yield a significant effect (B(SE)=−0.04(0.05), p=.393), indicating that the significant effect of CBSM on cancer-specific distress is not moderated by neighborhood disadvantage. See Table 3 for final model statistics of cancer-related distress.

Table 3.

Fixed and Random Effects for Final Models Predicting Cancer-Related Distress

Parameters	Main Effect Model	Moderation Model
Regression coefficients (fixed effects)
Intercept	20.87(2.81) **	19.49(3.27) **
Time	−0.45(0.07) **	−0.14(0.15)
Condition: CBSM	0.40(0.97)	0.72(2.14)
ADI: Low Disadvantage	-	2.71(1.81)
Age at Diagnosis	−0.16(0.05) **	−0.18(0.05) **
Stage of Disease: Ref = DCIS
Stage 1	−0.22(1.23)	0.20(1.31)
Stage 2	−0.21(1.25)	−0.09(1.33)
Stage 3	0.34(1.86)	0.71(1.97)
Time * Condition	−0.28(0.09) **	−0.50(0.20) *
Time * ADI	-	−0.36(0.17) *
Condition * ADI	-	−.26 (2.42)
Time * Condition * ADI	-	0.23(0.23)
Variance Components (random effects)
Residual	31.22(5.59)	29.98(5.43)
Intercept	28.32(5.32)	29.51(5.43)
Model summary
Deviance Statistic	4141.49	3863.95

Parameter estimate standard errors listed in parentheses

^**p<.01

^*p<.05.

CBSM Effects on Cortisol: Main and Moderator Effects

The final model of PM cortisol included a random, linear effect of time. When adding the interaction of time x condition and covariates, we confirmed prior published findings [23] that women receiving CBSM had significantly greater decreases in cortisol over the study period than did women receiving the control condition (B(SE)=−0.26(0.11), p=.014). The three-way interaction of ADI x condition x time was added to the model and did not yield a significant effect (B(SE)=−0.14(0.28), p=.622), indicating that the significant effect of CBSM on evening cortisol is not moderated by neighborhood disadvantage. See Table 4 for final model statistics of cortisol.

Table 4.

Fixed and Random Effects for Final Models Predicting Serum Cortisol

Parameters	Main Effect Model	Moderation Model
Regression coefficients (fixed effects)
Intercept	7.30(2.35) **	19.49(3.27) **
Time	0.10(0.07)	0.01(0.19)
Condition: CBSM	1.98(0.94)	0.12(2.16)
ADI: Low Disadvantage	-	−0.98(1.87)
Age at Diagnosis	0.01(0.04)	0.01(0.05)
Stage of Disease: Ref = DCIS
Stage 1	0.26(0.92)	−0.07(1.00)
Stage 2	0.33(0.94)	0.01(1.01)
Stage 3	−3.54(2.07)	−4.21(2.15)
Time * Condition	−0.26(0.21) *	−0.14(0.25)
Time * ADI	-	0.11(0.21)
Condition * ADI	-	2.41 (2.42)
Time * Condition * ADI	-	−0.13(0.28)
Variance Components (random effects)
Residual	16.97(4.12)	17.08(14.13)
Intercept	12.83(3.58)	13.41(3.66)
Time	0.08(0.27)	0.09(0.29)
Model summary
Deviance Statistic	1931.72	1857.51

Parameter estimate standard errors listed in parentheses

^**p<.01

^*p<.05.

Exploratory Data Analyses

Though we did not find significant three-way interactions of ADI x condition x time (i.e., moderation) in our outcomes of interest, we conducted exploratory analyses to investigate the effects of CBSM over time within the low ADI versus high ADI categories. Although it is not standard practice to probe simple slopes in the absence of a significant three-way interaction [30], we conducted these exploratory tests due to the theoretical and practical importance of understanding the relationship between CBSM and time (e.g., intervention efficacy) separately among women in high versus low disadvantage, especially in the context of data demonstrating women in high ADI neighborhoods have worse distress [11, 17], greater serum cortisol levels [17, 18], and poorer breast cancer outcomes [5]. In addition, probing the simple slopes without the inclusion of a three-way interaction simplifies the model, decreasing the probability of model misspecification and overfitting [30], which may have limited the above reported findings given our limited sample size. HLM was again used for these exploratory analyses to enhance power, given that all available data was used to estimate slopes [30]. This is especially important in these analyses given the smaller sample sizes.

CBSM Effects in Low Neighborhood Disadvantage Subsample

Cancer-related distress.

Among women in low ADI neighborhoods who completed at least two timepoints of the IES-I (N=173), we reran the model testing condition x time effects on distress controlling for age and stage. CBSM was associated with a greater decrease in cancer-specific distress as compared to the control condition (B(SE)=−0.49(0.18), p=.006).

Cortisol.

Among women in low ADI neighborhoods who completed at least two timepoints of cortisol collection (N=97), we reran the model testing condition x time effects on cortisol controlling for age and stage. CBSM was associated with a greater decrease in cortisol as compared to the control condition (B(SE)=−0.27(0.13), p=.038).

CBSM Effects in High Neighborhood Disadvantage Subsample

Cancer-related distress.

Among women in high ADI neighborhoods who completed at least two timepoints of the IES-I (N=50), we reran the model testing condition x time effects on distress controlling for age and stage. CBSM was associated with a greater decrease in cancer-specific distress as compared to the control condition (B(SE)=−0.27(0.11), p=.016).

Cortisol.

Among women in high ADI neighborhoods who completed at least two timepoints of cortisol collection (N=26), we reran the model testing condition x time effects on cortisol controlling for age and stage. CBSM was not associated with the change in cortisol as compared to the control condition (B(SE)=−0.19(0.19), p=.349).

Figures 2 and 3 represent the simple slopes of cancer-specific distress and cortisol by ADI.

Figure 2.

Cognitive Behavioral Stress Management Effects on Cancer-Related Distress in High versus Low Neighborhood Disadvantage.

Note. CBSM was associated with a greater decrease in cancer-related distress as compared to the control condition among both women in high disadvantaged (B(SE)=−0.49(0.18), p=.006) and low disadvantaged (B(SE)=−0.27(0.11), p=.016) neighborhoods.

Figure 3.

Cognitive Behavioral Stress Management Effects on Serum Cortisol in High versus Low Neighborhood Disadvantage.

Note. Among women in low Area Deprivation Neighborhoods (ADI), CBSM was associated with a greater decrease in PM serum cortisol as compared to the control condition (B(SE)=−0.27(0.13), p=.038). There was no effect of CBSM on the trajectory of serum cortisol among women in high ADI neighborhoods (B(SE)=−0.19(0.19), p=.349).

Discussion

The present study examined whether a cognitive behavioral stress management (CBSM) intervention for women with early-stage breast cancer had equivalent effects on self-reported cancer-specific distress and PM serum cortisol across women living in advantaged versus disadvantaged neighborhood environments. Our analyses confirmed prior work [23, 35] suggesting that the intervention reduces cancer-specific distress and serum cortisol over the year following diagnosis as compared to a single-session psycho-education seminar. Contrary to our hypotheses, these effects were not moderated by ADI. However, given concerns about multilevel complexity/overfitting within the full moderation models due to sample size constraints, exploratory analyses were conducted to probe the simple slopes of the effect of CBSM separately in women residing in high versus low disadvantaged neighborhoods. Here, we found that while cancer-related distress decreased in both women in high and low disadvantaged neighborhoods, only women residing in low disadvantaged neighborhoods experienced reductions in serum cortisol. Thus, CBSM was efficacious at reducing cancer-related distress across groups, but only women living in more advantaged neighborhoods experienced downstream neuroendocrine effects.

Significant literature suggests a benefit of CBSM for patients with cancer [20]; however, there is limited data on three-way interactions (i.e., moderators) to better understand which groups benefit most from the intervention. Among women with breast cancer, having a higher level of distress at baseline was associated with greater benefit from the intervention [36]. In men with prostate cancer, marital status was found to be a moderator of CBSM effects on stress [37], such that partnered men had greater benefits from the intervention. While we do not present results indicating full moderation, our exploratory simple slope analyses do suggest that there may be differential effects of CBSM based on neighborhood disadvantage. To our knowledge, this is the first analysis testing the differential effects of CBSM based on socioeconomic factors among women with breast cancer.

These results have important implications for the development and delivery of psychosocial interventions aimed at improving outcomes among women in disadvantaged neighborhoods. This is a critical population on which to intervene to promote health equity, as living in a disadvantaged neighborhood is consistently related to worse breast cancer outcomes [5, 38]. These disparities persist even when accounting for individual-level factors including access to care, stage of diagnosis, and NCCN-guideline concordant treatment [5]. It has been proposed that the chronic psychological stress of living in a marginalized neighborhood may contribute to worse outcomes via changes in the neuroendocrine signaling process, including the HPA axis [17] and SNS [10], which impact the immune system and tumor microenvironment [7, 11, 15]. This theory is supported by cross-sectional work from this cohort demonstrating that women living in disadvantaged neighborhoods had elevated anxiety symptoms and serum cortisol in the post-surgical period as compared to women living in advantaged neighborhoods [17]. This theory is further supported by the present results: though women in disadvantaged neighborhoods experience improvement in their cancer-related distress, CBSM in its present form may not adequately address additional stressors women in marginalized neighborhoods experience (e.g., structural racism, police presence and safety concerns). In this population, reducing cancer-related distress may be necessary but not sufficient to blunt the hyperactive neuroendocrine stress response. Directing and tailoring stress management skills (relaxation, cognitive reframing, coping effectiveness, interpersonal skills) to deal with perceptions of chronic neighborhood-related distress (e.g., safety concerns, lack of social cohesion, perceived discrimination) may be critical to achieve downstream biological changes among women in marginalized neighborhoods.

In addition to cancer- and neighborhood-related distress, an important target of intervention may be promoting health behaviors. Greater neighborhood disadvantage has been associated with worse medication adherence in other populations, including liver transplant patients [39] and patients with type 2 diabetes [40]. In addition, neighborhood disadvantage is related to physical inactivity [41] and poor sleep quality [42]. While CBSM interventions have resulted in enhanced health behaviors (e.g., endocrine therapy medication adherence [43], improved sleep quality [44]), additional focus on problem-focused coping skills to overcome barriers to health behaviors prevalent in disadvantaged neighborhoods (e.g., lack of green space for exercise) may further improve health outcomes.

Overall, these results point to the potential of adapting CBSM for women living in disadvantaged neighborhoods. In addition to targeting stressors beyond cancer that may be especially salient for women in marginalized neighborhoods, it is also important to consider the delivery of the intervention. For example, women in disadvantaged neighborhoods may have greater difficulty attending in-person sessions due to geographic constraints, which may point to the value of a telehealth intervention. However, prior work in other disease groups (persons with HIV) have demonstrated that higher disadvantage neighborhoods are linked to less uptake of telehealth visits [45]. In the current study, ADI was not associated with attendance of the in-person sessions, suggesting that in-person sessions may be feasible. However, the high ADI subsample was underrepresented in this trial, and remote methods may facilitate increased access to such an intervention and clinical trial participation among this group by lessening logistical and financial barriers. Mixed-methods and qualitative investigation may be an important next step to better understand the unique needs of women with breast cancer in disadvantaged neighborhoods. This patient-centered approach to intervention development would help inform necessary tailoring of content and modality to maximize uptake and benefits, with co-design of modifications with patient partners from the target communities. Should a digital intervention be identified as the optimal modality of delivery, it will be important to investigate ways to optimize engagement with the digital intervention to enhance outcomes in this population (Walsh et al., 2024).

In addition to adapting CBSM, dissemination and implementation science is critical to promote health equity. For instance, recent work investigating digitized CBSM has demonstrated efficacy at reducing distress [46]. Digitized CBSM may enhance access, allow for flexible use among patients, and allow the limited amount of mental health professionals trained in cancer-related distress to focus on the most complex cases [47]. Further work aiming to disseminate and implement CBSM, in addition to adapting the intervention, is needed for women living in disadvantaged environments.

The current study has several limitations. Primarily, we were limited in sample size. Sample size constraints limited us in the number of covariates we were able to add to our model without overfitting, and we were unable to control for additional individual-level covariates (e.g., race, ethnicity, insurance status). However, prior work has indicated that ADI has an impact on outcomes above and beyond these individual-level factors [5]. It is important to note that we were particularly limited in sample size among women living in high disadvantage neighborhoods. This may be due to trial recruitment from a major cancer center, and limits generalizability as this is not representative of the general catchment area [5]. To overcome this, we utilized tertiles of ADI and collapsed the upper two tertiles to obtain sufficient power, as has been done in prior work [17]. In addition, we utilized hierarchical linear modeling, which uses all available data to estimate the slope (e.g., change) and is less biased than traditional longitudinal analyses. In addition, given concerns about sample size, we conducted exploratory analyses of the simple slopes separately among high versus low neighborhood disadvantage to better understand the effects of CBSM in each population. Future trials that are adequately powered to test the moderation of ADI on the effects of psychosocial interventions on psychological and biological outcomes among women with breast cancer are warranted. Finally, we are limited in the absence of a measure of perceived neighborhood stress (e.g., the Neighborhood Social Environment Adversity Survey [48]) or general stress (e.g., the Perceived Stress Scale [49]) from the parent trial, and thus were only able to measure ADI differences in cancer-specific distress. This would be an important construct to measure in a future, well-powered trial to better understand the differential ability of CBSM to reduce cancer-specific distress as well as general distress, which may be higher at baseline among women in more disadvantaged neighborhoods. This is especially important to understand in the context of our unanticipated finding that women in higher disadvantaged neighborhoods reported less cancer-specific distress at baseline, despite revealing greater clinician rated anxiety and greater serum cortisol in this timeframe as reported elsewhere [17].

Another study limitation is the use of a single blood draw to capture serum cortisol at each timepoint. Although evening-time cortisol may represent dysregulation as levels during this time should be low in individuals with a typical diurnal rhythm, cortisol may fluctuate throughout the day due to stressors in the environment [50]. Thus, a single snapshot may not best represent overall cortisol levels. Future work should capture cortisol at various times throughout the day (e.g., utilizing salivary cortisol collection), to examine changes in the diurnal slope.

Conclusions

CBSM effectively reduced cancer-related distress among women with breast cancer living in advantaged and disadvantaged neighborhoods, but this only resulted in downstream biological changes among women living in advantaged neighborhoods. Neighborhood disadvantage may represent a particularly salient stressor over the cancer care continuum that impacts perceived neighborhood distress, neuroendocrine functioning and health behavior engagement, leading to worse health outcomes. Future interventions aiming to provide psychosocial resources to decrease the effect of stress on health should specifically target women in highly disadvantaged neighborhoods and modify the content and delivery modality to increase relevance and accessibility for this population to reduce cancer health disparities.

Data Availability Statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.