Eudaimonic Well-Being and Tumor Norepinephrine in Epithelial Ovarian Cancer Patients

Abstract

Background

The impact of psychological well-being on physiologic processes involved in cancer progression remains unclear. Prior research has implicated adrenergic signaling in tumor growth and metastasis. Given that adrenergic signaling is influenced by both positive and negative factors, we examined how two different aspects of well-being (eudaimonic and positive affect) and psychological distress were associated with tumor norepinephrine (NE) in ovarian cancer patients.

Methods

Women with suspected ovarian cancer (N=365) completed psychosocial assessments pre-surgery and clinical information was obtained from medical records. Study inclusion was confirmed following histological diagnosis. Tumor NE was measured in frozen tissue samples using HPLC with electrochemical detection. We employed confirmatory factor analysis to model eudaimonic well-being, positive affect, and psychological distress, and structural equation modeling (SEM) to examine associations between these factors and tumor NE.

Results

Eudaimonic well-being, positive affect, and psychological distress, modeled as distinct but correlated constructs, best fit the data (i.e., compared to unitary or 2-factor models) (RMSEA=.048, CFI=.982, SRMR=.035). SEM analyses that included physical well-being, stage, histology, psychological treatment history, beta-blocker use, and caffeine use as covariates had good model fit (RMSEA=.052, CFI=.955, SRMR=.036) and showed that eudaimonic well-being was related to lower tumor NE (β=−.24, p=.045). In contrast, we found no effects for positive affect or psychological distress.

Conclusion

Eudaimonic well-being is associated with lower tumor NE, independent of positive affect and psychological distress. Because adrenergic signaling is implicated in tumor progression, increasing eudaimonic well-being may improve both psychological and physiologic resilience in ovarian cancer patients.

Introduction

Ovarian cancer has the highest mortality of all female reproductive system cancers and an overall 5-year survival rate of 45%.¹ Psychological distress is common among ovarian cancer patients, with high rates of anxiety and depression.² Because psychological distress has been associated with both poorer quality of life³ and disease progression in a variety of cancers,⁴ it is important to identify factors that can be targeted to modulate disease-related processes and improve clinical outcomes and quality of life.

Interest in psychological factors that may protect health has grown substantially in recent years.⁵ Positive affect, or the well-being associated with pleasurable engagement with one’s environment (i.e., hedonic well-being),⁶ has been related to psychosocial resilience, longevity, and better health in healthy populations, independent of negative affect.^7,8,9 The role of positive affect in cancer has been less well-characterized. One study found that positive affect was related to biomarkers and survival in patients with renal cell carcinoma,¹⁰ whereas another found detrimental effects of positive affect on inflammatory biomarkers.¹¹ Other studies have shown that eudaimonic well-being, or experiencing a deeper sense of meaning in life, fulfilling one’s potential, and accepting oneself,¹² is distinct from positive affect and may be a stronger predictor of health status.^13–15 Eudaimonic well-being has been negatively related to perceived stress, depression, and anxiety in cancer patients and survivors.^16,17

In healthy adults, positive affect has been associated with a more normalized diurnal cortisol rhythm and enhanced immune function, although results are inconsistent.¹⁸ Likewise, eudaimonic well-being has generally been related to healthier inflammatory profiles, cortisol rhythms, cardiovascular biomarkers, and sleep patterns.^14,19,20 Differential effects for these constructs have also been shown. For example, eudaimonic well-being (but not positive affect) predicts better cardiovascular health¹⁵ and lower plasma interleukin-6 in older adults.¹⁴ Eudaimonic well-being has also been related to reduced activation of the conserved transcriptional response to adversity (CTRA), a gene expression signature characterized by upregulation of pro-inflammatory genes and downregulation of genes associated with anti-viral immunity,²¹ whereas positive affect (hedonic well-being) was associated with the reverse.²²

One pathway by which psychosocial factors influence health is via activation of the sympathetic nervous system (SNS), with norepinephrine (NE) as a key neuroeffector molecule.²³ NE-induced adrenergic signaling can regulate multiple downstream biological processes involved in tumor growth and progression,^24–26 including angiogenesis,²⁷ invasion²⁸, and resistance to anoikis.²⁹ Previous studies have found a relationship between tumor NE levels and negative factors such as stress, depression, and social isolation.^30–32, but little is known about the role of positive affect or eudaimonic well-being in cancer. Positive affect and eudaimonic well-being have shown divergent relationships with measures of SNS activation in healthy populations.^15,18 The goal of this study, therefore, was to examine how positive affect, eudaimonic well-being, and psychological distress relate to tumor NE in ovarian cancer patients.

Methods

Patients and Procedures

Patients were recruited at a clinic visit 1–2 weeks before diagnostic surgery for a pelvic mass suspected for ovarian cancer. Women with a history of cancer, non-ovarian primary tumor site, non-epithelial tumors, tumors of low malignant potential, or a comorbidity known to influence the immune system were excluded. Additional exclusion criteria included: less than 18 years, systemic corticosteroid use in the previous month, and currently pregnant. Psychosocial assessments were completed at home between the initial clinic visit and surgery. Tumor samples were fresh frozen in the surgical pathology lab as soon as possible following resection. The total sample included 365 women diagnosed with invasive primary epithelial ovarian, fallopian tube, or peritoneal cancer. Twelve patients had missing questionnaire data, so tests of the measurement model included 353 patients. Because maximum likelihood estimation uses all available data,^33,34 patients missing either tumor NE or questionnaires were still included in the predictive models (See CONSORT Table, Figure 1). All procedures were approved by institutional review boards at the University of Iowa, Washington University, and University of Miami.

Figure 1.

CONSORT diagram displaying patient inclusion.

Measures

Psychosocial measures

The Center for Epidemiological Studies-Depression Scale (CES-D) measures frequency of depressive symptoms over the previous week.³⁵ A stable four-factor structure has been identified, including depressed affect, positive affect, vegetative symptoms, and interpersonal relations.³⁶ Because cancer symptoms can resemble vegetative depression, we used the depressed mood subscale. The positive affect scale was also used. Scores on these subscales range from 0–15, with higher scores indicating more depression or positive affect. Adequate reliability was obtained for depression (α=.83) and positive affect (α=.77).

The Profile of Mood States short-form (POMS-SF) measures psychological distress, with subscales for fatigue, vigor, tension/anxiety, depression/dejection, confusion, and anger/hostility.³⁷ This factor structure has been validated in cancer patients.³⁸ The vigor subscale consists of 6 items assessing the frequency of emotions such as “cheerful” and “lively”. The vigor and depression/dejection subscales both had excellent reliability (α=.88 and .90, respectively).

Cancer-specific intrusive thoughts and avoidance behavior were assessed with the Impact of Event Scale (IES).³⁹ This 15-item scale has been used to measure psychological distress and anxiety symptoms in cancer patients.⁴⁰ Participants indicated how often in the previous week they experienced intrusive thoughts and avoidance regarding their cancer, with higher scores reflecting greater psychological distress. This scale had excellent reliability (α=.89).

Three subscales from Ryff’s Psychological Well-Being Scales (PWBS) were used as measures of eudaimonic well-being.¹² The personal growth scale measures whether individuals feel they have psychologically developed and realized their potential. The purpose in life scale assesses whether one has goals and a sense of meaning in life. The self-acceptance scale assesses positive attitude toward the self and acceptance of both good and bad qualities. Each scale has 7-items with higher scores indicating greater well-being. Adequate reliability was obtained for each scale (α’s≥.74). This questionnaire was added to the protocol following initiation of the study; thus fewer participants completed the PWBS.

Self-reported symptoms were measured with the physical well-being subscale of the Functional Assessment of Cancer Therapy.⁴¹ This subscale assesses 7 somatic symptoms commonly experienced by cancer patients (e.g., lack of energy, pain) and is sensitive to disease stage. Items are rated over the previous week, with higher scores indicating fewer symptoms or greater well-being. This measure demonstrated good reliability (α=.84).

Demographic and clinical information

Demographic information was obtained by self-report and clinical information was obtained from medical records.

Tumor norepinephrine

Tumor NE was measured as previously described.⁴² Briefly, frozen tumor samples were pulverized, homogenized, and extracted prior to immediate measurement of catecholamine levels using high performance liquid chromatography with electrochemical detection (HPLC-ED).⁴³ Tumor NE was below detectable limits for 31% of patients and was set at the lowest detectable level of the assay (0.1 pg/mg). Tumor epinephrine and dopamine levels were below detectable limits for the majority of the sample; thus, these catecholamines were not included in analyses.

Analytic procedures

Initial analyses were conducted using SPSS (Version 21, Armonk, NY). Distributions were examined for normality and outliers (>3× the interquartile range), and no outliers were found. Tumor NE values were log-transformed due to non-normality. First, the measurement model was examined using confirmatory factor analysis (CFA) with the following indicators: purpose in life, self-acceptance, and personal growth (for eudaimonic well-being); vigor and positive affect (for positive affect); and depressed affect, depression-dejection, and cancer-specific distress (for psychological distress). Physical well-being was included in the CFA as a covariate of each factor to adjust for the influence of physical symptoms on reports of psychological states.⁴⁴ To determine if eudaimonic well-being, psychological distress, and positive affect are independent constructs or opposite poles of a unitary construct,⁴⁵ we evaluated a model including only one latent variable with all manifest variables as indicators. Similarly, to determine if positive affect and eudaimonic well-being are independent constructs, a 2-factor model with a general positive/well-being latent variable along with the psychological distress factor was tested.

After a measurement model was adequately specified, a structural equation model (SEM) was assessed with tumor NE as an endogenous outcome regressed on the three latent variables. Tumor stage, daily caffeine intake, use of beta-blockers, and physical well-being were included as a priori covariates, due to known relationships with adrenergic biomarkers and to control for potential influences of tumor burden and physical disability. To determine additional covariates, bivariate correlations were examined between tumor NE and body mass index, age, sleep quality, grade, histology, education, ethnicity, smoking history, psychological treatment history, and presence of medical comorbidities. Variables significantly correlated with tumor NE used as covariates and included histology and psychological treatment history. Tumor NE was regressed on these covariates, and correlations between covariates and latent factors were modeled.

All analyses were conducted in Mplus (Version 6.12, Los Angeles, CA) using maximum likelihood estimation with robust standard errors to account for non-normality and estimate parameters with missing data present.⁴⁶ Maximum likelihood estimation uses all available data, which provides a statistical advantage in terms of both power and reducing bias in parameter estimates.^33,34 Overall model fit was evaluated by examining the comparative fit index (CFI),⁴⁷ the root mean square error of approximation (RMSEA)⁴⁸, and the standardized root-mean-squared residual (SRMR) and estimates from the standardized model were reported. Recommended cut-off points⁴⁹ of .95 for CFI, .06 for RMSEA, and .08 for SRMR were used to evaluate model fit.⁴⁹). Non-nested models were compared by evaluating the difference between the Bayesian Information Criterion (BIC) for each model using Raftery’s guidelines (0–2=weak, 2–6=positive, 6–10=strong, >10=very strong).⁵⁰

Results

Sample Characteristics

Demographic and clinical characteristics of the sample are presented in Table 1. Average age at diagnosis was 59.8 years. Participants were primarily white, non-Hispanic, and married or living with a partner. Most participants were diagnosed with high grade (86%, n=315), advanced stage (72%, n=263) tumors. Bivariate correlations between study variables are displayed in Table 2.

Table 1.

Participant’s Demographic and Clinical Characteristics (N=365)

Characteristic	N (%)
Age (years)	M=59.76, SD=11.71
Race
American Indian/Alaskan Native	2 (.5)
Asian	3 (.8)
Pacific Islander	0
Black/African American	9 (2.5)
White	346 (94.8)
Missing data	5 (1.4)
Ethnicity
Non-Hispanic	322 (88.6)
Hispanic	15 (4.1)
Unknown/Missing data	28 (7.6)
Education
High school or less	133 (36.5)
Trade school/some college	110 (30.1)
College graduate	73 (20)
Postgraduate degree	31 (8.5)
Missing data	18 (4.9)
Relationship status
Married or living with partner	235 (64.4)
Single, separated, widowed, divorced	127 (34.8)
Missing data	3 (.8)
Cancer stage
I	72 (19.7)
II	26 (7.1)
III	229 (62.7)
IV	34 (9.3)
Missing data	4 (1.1)
Grade
Low	46 (12.6)
High	315 (86.3)
Missing data	4 (1.1)
Histology
Serous	263 (72.1)
Endometrioid	35 (9.6)
Mucinous	15 (4.1)
Clearcell	24 (6.6)
Other/unknown/missing data	28 (7.7)
Surgical debulking
Optimal	262 (71.8)
Suboptimal	92 (25.2)
Missing data	11 (3)

Table 2.

Bivariate Correlations, Means, and Standard Deviations of Latent Variable Indicators and Tumor Norepinephrine

Variable	1	2	3	4	5	6	7	8	9	10
1. Personal Growth (PWBS)	--
2. Purpose in life (PWBS)	.74***	--
	188
3. Self-Acceptance (PWBS)	.65***	.63***	--
	188	188
4. Positive affect (CES-D)	.37***	.35***	.39***	--
	186	187	186
5. Vigor (POMS-SF)	.31***	.32***	.29***	.50***	--
	181	182	181	326
6. Depressed affect (CES-D)	−.25***	−.25***	−.34***	−.49***	−.27***	--
	186	187	186	345	327
7. Depression-dejection (POMS-SF)	−.39***	−.33***	−.38***	−.52***	−.38***	.73***	--
	181	182	181	326	333	327
8. Cancer-specific distress (IES)	−.25**	−.18**	−.23**	−.35***	−.24***	.59***	.63***	--
	179	180	179	326	328	327	328
9. Physical well-being (FACT)	.26***	.30***	.26**	.49***	.47***	−.37***	−.48***	−.23***	--
	187	188	187	334	322	335	322	322
10. Tumor NE (log10, pg/mg)	−.21**	−.07	−.04	−.06	−.08	.04	.10	.02	−.12*	--
	151	151	151	272	263	273	263	262	267
N	188	189	188	345	334	346	334	334	339	290
M	38.03	38.26	38.21	7.50	8.32	3.49	6.87	26.92	19.41	.07
SD	6.81	7.33	7.97	3.00	5.45	3.34	6.64	14.79	6.39	1.01

Note:. N for each correlation is displayed below in italics and varies due to missing data for each pair of measures.

^*p<.05

^**p<.01

^***p<.001

Measurement Model

The initial CFA testing the hypothesized structure of eudaimonic well-being, positive affect, and psychological distress is presented in Table 3. This model had excellent fit (RMSEA=.048, CFI=.982, SRMR=.035, BIC=15740) and all factor loadings were significant (p<.001). Moreover, the three-factor model provided substantially better fit to the data than the unitary model (RMSEA=.177, CFI=.71, SRMR=.123, BIC=16002) and the two-factor model (RMSEA=.113, CFI=.891, SRMR=.097, BIC=15822). These results indicate that eudaimonic well-being, positive affect, and distress are correlated but independent constructs in this population.

Table 3.

Factor Loadings of Latent Variables from the Confirmatory Factor Analysis Model

Indicator	Standardized estimate	Standard error	R2
Eudaimonic well-being
Personal growth	.87	.03	.76
Purpose in life	.84	.04	.70
Self-acceptance	.77	.05	.59
Positive affect
Positive affect	.77	.04	.60
Vigor	.64	.05	.41
Distress
Depressed mood	.80	.03	.64
Depressed-dejection	.93	.02	.86
Cancer-specific distress	.69	.04	.47

Note: All factor loadings are significant at p <.001.

Structural Equation Model Predicting Tumor Norepinephrine

The final SEM is shown in Figure 2. Prior to model testing, we examined bivariate correlations between tumor NE and potential covariates. All a priori covariates were significantly related to tumor NE in bivariate correlations, including cancer stage (N=289, r_pb=.16, p=.01), beta-blockers (N=290, r_pb=−.14, p=.02), average daily caffeine intake (N=272, r=.16, p=.01), and physical well-being (r=−.12, p=.05). Histology (N=290, r_pb= .19, p=.001), and psychological treatment history (N=274, r_pb=−.13, p=.03) were also related to tumor NE and thus included in the SEM as covariates. The SEM model had good fit (RMSEA=.052, CFI=.955, SRMR=.036, BIC=19816). Of the covariates included, greater daily caffeine use was related to higher tumor NE (β=.16, p=.02), while psychological treatment history was related to lower tumor NE (β=−.21, p=.002). Relations between tumor NE and beta-blocker use (β=−.14, p=.06) and physical well-being (β=−.17, p=.09) were marginally significant, while cancer stage and histology (β=.11, p=.09) were unrelated (β=.10, p=.14).

Figure 2.

Structural equation model predicting tumor norepinephrine from latent variables and covariates. Indicators of latent variables, and correlations between control and latent variables, are removed for simplicity. * indicates p<.05

Several covariates were also related to the latent constructs. Advanced stage disease was associated with lower positive affect (estimate=−.16, p=.02), but not with eudaimonic well-being (estimate=.03, p=.74) or distress (estimate=.09, p=.09). Beta-blockers and caffeine use were unrelated to all latent variables. Tumor histology was unrelated to eudaimonic well-being and psychological distress, but patients with non-serous tumors reported more positive affect (estimate=−.17, p=.005). History of psychological treatment was related to lower eudaimonic well-being (estimate=−.19, p=.014) and greater psychological distress (estimate=.23, p<.001). Physical well-being remained significantly associated with eudaimonic well-being (estimate=.33), positive affect (estimate=.66), and distress (estimate=−.48) (p’s<.001), and the covariances between factors remained significant, with relationships similar to those in the measurement model.

Lastly, and most importantly, we examined associations between the latent variables and tumor NE. Eudaimonic well-being was significantly related to lower tumor NE (β=−.24, p=.045). In contrast, psychological distress (β=.08, p=.45) and positive affect (β=.24, p=.17) were both unrelated to tumor NE (see Figure 2).

Discussion

The key finding of this study was that higher levels of eudaimonic well-being were found to be associated with lower tumor NE in ovarian cancer patients, while controlling for positive affect, psychological distress, and several demographic and clinical covariates. In contrast, positive affect and psychological distress were unrelated to tumor NE. In addition, consistent with prior research,^15,19,22,51 we found that positive affect (hedonic well-being) and eudaimonic well-being were correlated but distinct constructs. This suggests that the beneficial effects of eudaimonic well-being are independent of the influence of other factors, including positive affect. Based on these findings, we conclude that a deeper sense of well-being in cancer patients may be more physiologically protective than positive or pleasant emotions. A cancer diagnosis often provokes existential concerns, including a struggle to maintain self-identity and meaning in life.⁵² Eudaimonic well-being, which entails acting in accordance with one’s deepest values, appears to be more indicative of positive adjustment in the context of this struggle.⁵¹ For example, meaning and peace, possible indicators of eudaimonic well-being, have been associated with better QOL and less depression and anxiety in ovarian cancer patients.⁵³

The present findings may suggest a potential benefit for patients with greater eudaimonic well-being, given that beta-adrenergic signaling upregulates multiple pathways involved in tumor progression.²⁴. For example, we have previously reported associations of elevated tumor NE in ovarian cancer patients with activated focal adhesion kinase (FAK), which promotes tumor cell survival during metastasis. FAK in turn was associated with poorer overall survival.²⁹

These findings extend prior research on psychosocial factors and catecholamines in the tumor microenvironment. Data from an orthotopic model of ovarian cancer indicate that restraint-stressed mice show elevated tumor NE, along with greater tumor weight and nodules.⁵⁴ We have previously reported that ovarian cancer patients with greater social isolation have higher levels of NE in both tumor and ascites,⁵⁵ and that a composite of high depression and low social support is also associated with elevated tumor NE.^42,56 The present study extends this work by demonstrating that a broader sense of well-being predicts lower tumor NE. This study is also the first to examine the physiologic relevance of these factors in the context of cancer. The differential effects observed for positive affect and eudaimonic well-being are not surprising given that positive affect can have variable effects on the SNS, depending on both environmental and subject-specific factors.¹⁸ This study thus adds to the literature demonstrating the benefits of eudaimonic well-being, distinct from positive affect.⁵⁷

These findings have potential implications for clinical care. For example, assessing eudaimonic well-being can be done quickly, and individuals with low well-being may be candidates for psychosocial intervention. Additionally, the data suggest that interventions targeting eudaimonic well-being, such as Acceptance and Commitment Therapy,⁵⁸ mindfulness interventions,⁵⁹ and positive psychology approaches emphasizing gratitude and patient strengths,⁶⁰ may be more beneficial for enhancing physiological resilience than those targeting positive affect or psychological distress.⁵⁷

Limitations

These findings do not confirm causality between well-being and tumor NE. However, the methodological strengths, including the simultaneous examination of eudaimonic well-being, positive affect, psychological distress, provide a fine-tuned discrimination between these constructs. Although epinephrine and dopamine may also impact processes involved in tumor progression,^25,61 we found negligible levels of both; thus, these catecholamines were not included in analyses. The statistical approach enabled patients with missing data to be included.³³ However, 20.5% of this sample had questionnaire but not tumor NE data, which may have affected our ability to detect weaker associations between study variables. Lastly, while the tumor NE has been indirectly related to survival (e.g., via increased FAK), its direct effect on survival is unknown.

In conclusion, eudaimonic well-being, but not positive affect or psychological distress, is associated with tumor NE in ovarian cancer patients. Because ovarian cancer diagnosis and treatment is often accompanied by psychological distress, and because adrenergic signaling is known to play a role in tumor progression, interventions that increase eudaimonic well-being may have important clinical effects in cancer patients. Additional research is warranted to examine how eudaimonic well-being influences quality of life and disease progression in ovarian and other cancers.