Relationship of cancer-related fatigue with psychoneurophysiological (PNP) symptoms in breast cancer survivors

Abstract

Purpose:

Cancer-related fatigue (CRF) is a highly prevalent and debilitating symptom reported by breast cancer survivors (BCS). CRF has been associated with the co-occurrence of anxiety, depression, poor sleep quality, cognitive impairment, which are collectively termed as psychoneurophysiological (PNP) symptoms. CRF and these PNP symptoms are often reported during and after treatment with long-lasting distress. It is unclear how CRF and these PNP symptoms influence each other. This study aimed to explore predictive factors (i.e., PNP symptoms and social-demographic factors) of CRF, and test exploratory path models of the relationships of CRF with PNP symptoms (depression, anxiety, sleep disturbance, pain, and cognitive function) in BCS.

Methods:

This paper is part of a larger descriptive, correlational, and cross-sectional study. Validated and reliable instruments assessed CRF, depression, anxiety, sleep disturbance, pain, and cognitive function. Descriptive statistics, Pearson correlation, multiple linear regression models, and path analysis were employed.

Results:

Patients (N=373) who reported less bodily pain had worst CRF (r=−.45, p<.01). Significant predictors of CRF included depression, sleep disorder, bodily pain, perceived cognitive ability, and dispositional (state) optimism. Depression alone accounted for 31% of the variance in CRF. An integrative path model with bodily pain, neuropathic pain, CRF, and depression showed a good fit across different indices (CFI=0.993, RMSEA=0.047, 90% CI 0–0.12, SRMR=0.027).

Conclusions:

This study identified significant predictors of CRF and revealed a good fit mediation model with significant pathways for CRF, suggesting that a common etiology may underpin the co-occurrence of CRF with PNP symptoms (pain and depression). However, further investigation with longitudinal design is necessary to explore the causal relationships of these symptoms. Evidence-based strategies/interventions are needed to reduce or eliminate the burden of these symptoms on the lives of BCS.

1. Introduction

Being the second most frequent cancer among American women, breast cancer is a fairly common malignancy (American Cancer Society, 2023). The survival rate of breast cancer has increased due to better therapy and earlier diagnosis, yet breast cancer survivors (BCS) are still burdened by long-term physical and psychological side effects from the cancer and/or the cancer treatments they received (American Cancer Society, 2023). Treatments for breast cancer, for instance, may raise the risk of morbidity and lower quality of life due to the presence of unpleasant symptoms (Harris, 2021, Hickey et al., 2016, Shaitelman et al., 2015, Kim et al., 2012). Long after the course of cancer treatment has ended, BCS continue to report negative physical and psychological effects, like complaints of cancer-related fatigue (CRF), an extremely common, incapacitating, and chronic behavioral toxicity, experienced during and after cancer therapy, but has also been reported even during cancer remission (Bower et al., 2000, O’ Regan and Hegarty, 2017, Goedendorp et al., 2016, Wang et al., 2014, Wang, 2008, Minton et al., 2013).

The experience of CRF reported by BCS during and after treatment frequently co-occur with other symptoms such as depression, anxiety, pain, cognitive impairment, and sleep disruption, resulting in ongoing distress (Harris, 2021, Hickey et al., 2016, Shaitelman et al., 2015, George et al., 2020, Goedendorp et al., 2016). The co-occurrence of these symptoms with CRF, especially in higher severity, such as disturbed sleep, depressed mood, and anxiety establishes CRF as a potential index symptom critical in driving the clustering of these symptoms, collectively referred to as psychoneurophysiological (PNP) symptoms (Crane et al., 2020, Chow et al., 2019).

The clustering of CRF with other PNP symptoms have detrimental impact on the quality of life of cancer survivors (Berger and Mitchell, 2008, Pinto et al., 2013, Hauth et al., 2021), particularly contributing to a decline in treatment compliance and survival rates (Byar et al., 2006, Berger and Mitchell, 2008, Monga et al., 1999). However, investigating the direct and indirect linkages of CRF with these co-occurring PNP symptoms has just gained a new interest in survivorship research with the long-term goal of using these proven relationships as interventional targets to reduce or eliminate the distressing impact of these symptoms on the lives of BCS. Prior longitudinal study observed that depression, fatigue, and sleep disturbance reported by BCS were highly correlated both within and between treatment time points, where baseline fatigue (time 1) predicted post-chemotherapy depression (time 2), and postchemotherapy fatigue (time 2) predicted depression at follow-up (time 3) (Ho et al., 2015). Even while CRF has been found to be a significant risk factor for other behaviors, there is currently less research on how CRF and these co-occurring PNP symptoms interact and influence one another, particularly over time (Trudel-Fitzgerald et al., 2013b, Trudel-Fitzgerald et al., 2013a).

Therefore, this secondary analysis of a cross-sectional data is our initial attempt to (1) explore predictive factors (i.e., PNP symptoms and social-demographic factors) of CRF, and (2) test exploratory path models to establish the relationships of CRF with PNP symptoms in BCS. We are currently following these patients to determine longitudinal changes to the relationships of these co-occurring symptoms.

2. Theoretical Framework

This study is grounded on the revised Theory of Unpleasant Symptoms (TOUS) to answer its study aims. The TOUS is founded on the assumption that symptoms share enough similarities. The theory’s unique implication was that treating one symptom will help with treating others. Therefore, the revised theory of TOUS addresses the synchronous occurrence of several symptoms that may have an additive impact on symptom experience, distress, and performance (Lenz et al., 1997, Lenz et al., 1995). In TOUS, one’s propensity to experience a particular unpleasant symptom experience is influenced by three important components, physiological, psychological, and environmental/situational perspectives (Lenz et al., 1997, Lenz et al., 1995). Age, gender, aspects of the illness and therapies (such as clinical phases and various therapy regimens) are examples of the physiological component of the model. Mental health issues to include emotional and cognitive issues, comprise the psychological component. Situational components include social support and marital status. For this study, demographic-social-clinical traits such as age, race, marital and work status, level of education, cancer type/stage, and treatment were considered physiological and situational factors. The psychological factors included the PNP symptoms that co-occurred with CRF, such as anxiety, depression, pain, and perceived cognitive function. We hypothesized that these influencing factors—physiological, environmental, and psychological—would have either favorable or unfavorable effects on CRF.

3. Methods

3.1. Design, Study Participants, and Procedures

This paper is part of a broader descriptive, correlational, and cross-sectional study (ClinicalTrials.gov Identifier: NCT04611620) that looks at factors that influence cancer related-cognitive impairment reported by breast and colorectal cancer survivors. The Institutional Review Board (IRB) at Indiana University and the Melvin and Bren Simon Comprehensive Cancer Center both gave their approval for the parent study. Using IRB-approved marketing, prospective volunteers were recruited via social media sites like Facebook and online cancer resource websites like Pink Ribbon Connection, Dr. Susan Love Foundation, Colorectal Cancer Alliance, etc.

Inclusion criteria for the parent study included self-reports of: 21 years of age or older; 6 months post-adjuvant therapy and neo-adjuvant therapy for early stage (Stage I–III) colorectal cancer or breast cancer (apart from current Aromatase Inhibitor or Tamoxifen use for BCS); and cancer-related cognitive impairment. Prior to the administration of any study procedures, eligible participants signed written online informed consents and HIPAA REDCap^® forms. Only a subsample of the BCS population who completed all study outcomes was included in this analysis since the primary trial had close to 89% BCS enrollment.

Individual unique identifiers were assigned to participants after we received their informed consent. Then, utilizing the REDCap^® platform, a secure link to a series of survey questions was made available to participants. The participant’s current age, race/ethnicity, marital status, prior educational history, current work status, income, diagnosis and clinical stage, and cancer treatments (chemotherapy, surgery, and/or radiation therapy) were all asked in the demographic and clinical form. The following section discusses the study outcomes.

2.2. Study Measures

The 8-item short forms for the Patient-Reported Outcomes Measurement Information System (PROMIS^®) evaluated CRF and PNP symptoms such as anxiety, neuropathic pain, depression, and sleep issues. These standardized PROMIS^® short forms have received extensive validation for the assessment of symptoms across clinical populations (Pilkonis et al., 2011, Lai et al., 2011, Askew et al., 2016, Yu et al., 2012). Higher scores denote more severe/worse symptomology on the 5-point Likert scale, ranging from 0 to 4. The PROMIS^® reliability ranged from 0.72 to 0.99 for this specific secondary analysis. Bodily pain was measured by the SF-36’s two-item subscale, a widely used indicator of general health that has been validated in a variety of populations, including cancer patients (Ware and Sherbourne, 1992). For this analysis, reliability for the SF-36 bodily pain scale ranged from 0.80 to 0.94.

Participants’ perceptions of cognitive impairment were measured using the PROMIS^® Applied Cognition, version 1.0 short-form, 8-item subscales of Cognitive Abilities and Cognitive Concerns. For each subscale, the items were added up to produce a total score. Higher scores on the Cognitive Abilities subscale reflects greater cognitive capacity and higher scores in the Cognitive Concerns subscale suggests greater cognitive worries, and this form has been validated in cancer populations (Lai et al., 2014); Cella et al., 2010). For this analysis, the reliability of the PROMIS^® Applied Cognition ranged from 0.80 to 0.83.

The 10-item Life Orientation Test-Revised (LOT-R) was used to gauge optimism in both state and disposition (Scheier et al., 1994). On a 5-point Likert scale, where higher scores signal greater optimism, respondents were asked to rate their level or degree of agreement. The scale ranged from 0 (strongly disagree) to 4 (strongly agree). Based on LOT-R scores, the following clinically significant categories have been proposed: 0–13 = low optimism, 14–18 = moderate optimism, and 19–24 = strong optimism. (Celestine, 2019). For this analysis, LOT-R reliability ranged from 0.79 to 0.85.

3.3. Statistical Analyses

To determine which factors (depression, anxiety, neuropathic and bodily pain, sleep disturbance, and cognitive function) can be utilized to predict CRF in BCS, we first used descriptive statistics, bivariate Pearson correlation, and linear multiple regression models. To investigate the connections between pertinent variables and the outcome variable (CRF), correlation coefficients were used. The initial regression model only contained variables with a correlation coefficient greater than 0.3. We employed the stepwise technique to examine the significance of the predictors after incorporating all the predictors and two potential covariates (age and living with a partner/married or not) into the regression model. The final multiple regression model was comprised of only significant predictors. SPSS 28.0 (IBM Corporation, Armonk, NY) was used for all statistical analyses.

Second, we further investigated the potential connections between these important variables in a path model using path analysis. Based on several recent studies and our personal knowledge of CRF, three distinct models were investigated: 1) The first model was a simple mediation model using CRF as the mediator, based on a previous observation that bodily pain predicts CRF, and CRF predicts depression (Ho et al., 2015); 2) The second model was a more comprehensive model of fatigue, in which bodily pain and neuropathic pain predict fatigue, while fatigue and state optimism predict depression, and depression predicts cognitive abilities (Lukkahatai et al., 2016, Gu et al., 2023); 3) Based on models 1 and 2, we eliminated cognitive abilities as one of the model’s outcome variables and proposed a model that is simpler than model 2. In this model, we hypothesized that bodily pain and neuropathic pain predict CRF, and CRF predicts depression. The lavaan package (Rosseel, 2012), was used to conduct the path analysis in R version 4.3.1 (R Core Team, 2023).

4. Results

4.1. Sample Description

This analysis used a total of 373 BCS, as a subsample. The sociodemographic and clinical characteristics of the study sample are shown in Table 1. Majority of participants (348/373 or 93%) were White, married (255/373 or 68%), with an average age of 55.9 years (±9.8). More than half of the participants (244/373, 66%) received both chemotherapy and radiation therapy for their clinical cancer stage I (123/373, 33%) and stage II (160/373, 43%).

Table 1.

Sociodemographic and clinical characteristics of sample.

Characteristics	N	Mean (S.D) / Frequency (%)
Sociodemographic Characteristics
Age	372	55.9 (9.8)
Race	373
White	348	93%
Black	12	3%
Other	13	4%
Marital Status
Single	33	9%
Living with Partner	17	5%
Married	255	68%
Windowed	16	4%
Divorced	45	12%
Other / Prefer not to answer	7	2%
Highest Education Level
High school graduate	23	6%
Associate’s degree or equivalent	84	23%
Undergraduate / Bachelor’s degree or equivalent	132	35%
Master’s degree or equivalent	100	27%
PhD or equivalent	34	9%
Employment Status
Full-time (> 35 hrs/wk)	169	45%
Part-time (< 20 hrs/wk)	58	15%
Homemaker	14	4%
Retired	96	26%
Unemployed	25	7%
Other / Prefer not to answer	11	3%
Cancer Stage
Stage I	123	33%
Stage II	160	43%
Stage III	76	20%
Unsure	14	4%
Chemo and Radiation therapy
None	22	6%
Chemotherapy only	87	23%
Radiation only	20	5%
Both Chemotherapy and radiation	244	66%

4.2. CRF and Psychoneurophysiological (PNP) symptoms

Table 2 lists the mean scores and standard deviations for CRF and other PNP symptoms. According to published clinically significant cut-offs, the PROMIS^® short forms’ CRF, depression, anxiety, and sleep disturbance scores were greater than those of the pertinent reference population, which has a T score of 50 (National Institutes of Health, 2022). The mean PROMIS^® score for perceived cognitive abilities (T score = 40.71) was about one standard deviation lower (worse) than the average T score of the relevant reference population, which is 50. The subsample of BCS used in this research had a mean dispositional/state optimism score of 14.9 (±4.7), which indicates moderate optimism (Celestine, 2019).

Table 2.

Description of cancer- psychoneurophysiological (PNP) symptoms

Symptoms	Questionnaire	N	Mean (S.D.)	T Score (Mean raw score rounded)
Lower scores=Better
Fatigue	PROMIS 8a Adult SF	373	24.3 (8.6)	57.5*
Depression	PROMIS 8a Adult SF	373	16.6 (7.4)	55.9*
Anxiety	PROMIS 8a Adult SF	373	18.7 (6.9)	57.4*
Sleep Disturbance	PROMIS 8a Adult SF	373	23.9 (8.3)	56.2*
Neuro Pain	PROMIS 5a Adult SF	373	9.4 (5.1)	48.8
Cognitive Concerns	PROMIS 8a Adult SF	373	26.9 (7.6)	43.3
Higher scores=Better
Cognitive Abilities	PROMIS 8a Adult SF	373	21.0 (6.9)	40.71*
State Optimism	Life Orientation Test-Revised	373	14.9 (4.7)**
Bodily Pain	SF-36 Subscale	373	57.7 (25.6)

Note: SF = short form; SD = standard deviation

^*T-score > or < 50. When interpreting Patient Reported Outcome Measurement Information System (PROMIS) T-Scores, a score of 50 is the mean of a “relevant reference population” with a standard deviation of 10. So, 40 is one standard deviation below mean, and 60 is one standard deviation above mean. (https://www.healthmeasures.net/score-and-interpret/interpret-scores/promis)

^**Life Orientation Test – Revised scores of 14 – 18 are categorized to have moderate optimism (https://positivepsychology.com/life-orientation-test-revised/).

4.3. Correlations between CRF and PNP symptoms

The worst CRF was reported by BCS with high anxiety (r = .47, p < .01) and high depression (r = .56, p < .01), as shown in Table 3. Similarly, individuals who reported having the worst CRF (both p < .01) also reported having low perceived cognitive abilities (PROMIS^® applied cognitive ability, r = −.49), as well as significant cognitive concerns (PROMIS^® applied cognitive concern, r = .46). It is interesting to note that patients with lower levels of bodily pain had worse CRF (r = −.45, p < .01).

Table 3.

Correlations between cancer-related fatigue (CRF) and other psychoneurophysiological (PNP) symptoms

Means, standard deviations, and correlations with confidence intervals

Variable	N	M	SD	1	2	3	4	5	6	7
1. CRF	373	24.29	8.59
2. Anxiety	373	18.70	6.88	47** [.39, .54]
3. Bodily Pain	373	57.71	25.57	−.45** [−.53, −.37]	−.25** [−.34, −.15]
4. Depression	373	16.63	7.42	.56** [.48, .62]	.70** [.64, .74]	−.30** [−.39, −.21]
5. State Optimism (LOT-R)	373	14.92	4.74	−.32** [−.40, −.22]	−.48** [−.55, −.40]	.26** [.16, .35]	−.59** [−.65, −.52]
6. Cognitive Abilities (Aca score)	373	20.96	6.88	−.49** [−.56, −.40]	−.48** [−.55, −.39]	.32** [22, .41]	−.51** [−.58, −.43]	.38** [.29, .46]
7. Cognitive Concerns (PROMIS AppCog)	373	26.89	7.58	.46** [.38, .54]	46** [.38, .54]	−.35** [−.44, −.26]	46** [.38, .54]	−.29** [−.38, −.19]	−.75** [−.79, −.70]
8. Neuro Pain	373	9.44	5.10	.37** [.28, .45]	23** [.13, .32]	−.47** [−.54, −.38]	.28** [.18, .37]	−.18** [−.28, −.08]	−.19** [−.29, −.09]	.23** [.13, .33]

Note. M and SD are used to represent mean and standard deviation, respectively. Values in square brackets indicate the 95% confidence interval for each correlation. The confidence interval is a plausible range of population correlations that could have caused the sample correlation (Cumming, 2014).

^*Indicates p < .05.

^**indicates p < .01.

The correlations between cancer-related fatigue (CRF) and the predictive factors showed that CRF was related to depression, bodily pain, anxiety, perceived cognitive function (cognitive abilities and cognitive concerns), and dispositional/state optimism. LOT-R = Life Orientation Test-Revised assessed dispositional/state optimism, cognitive abilities was assessed by PROMIS^® Applied Cognitive Abilities, and general cognitive concerns was assessed by the PROMIS^® Applied Cognition - General Concerns.

4.4. Predictive factors of CRF

We presented that the outcomes of the multiple linear regression analysis utilizing CRF as the dependent variable in Table 4. In the final linear regression model, only the significant predictors were included. We considered age and marital status (married + living with partner versus others) as potential covariates because the subsample was 93% White and had an associate educational degree or higher (see Table 1). Both covariates (age and marital status) were excluded from the final model since they were not significant predictors in the first regression analysis. As shown in Table 4, depression, sleep disturbance, bodily pain, perceived cognitive ability, and dispositional (state) optimism were the major predictors of CRF. Approximately 7% of the variance in CRF was caused by bodily pain, while 31% of the variance in CRF was caused by depression alone. With a beta value of 0.36, the CRF score increased 0.36 standard deviations for every one standard deviation increase in depression. According to the standardized regression coefficients, depression had the strongest predictive ability of all the variables.

Table 4.

Predictors of cancer-related fatigue (CRF) - Regression results using fatigue as the criterion

Predictor	b	b 95% CI [LL, UL]	beta	beta 95% CI [LL, UL]	sr 2	sr2 95% CI [LL, UL]	r	Fit
(Intercept)	17.02**	[11.33, 22.72]
Depression	0.40**	[0.28, 0.52]	0.35	[0.24, 0.45]	.06**	[.03, .10]	.56**
Sleep Disturbance	0.21**	[0.12, 0.29]	0.20	[0.11, 0.28]	.03**	[.00, .06]	46**
Bodily Pain	−0.07**	[−0.10, −0.04]	−0.21	[−0.30, −0.12]	.03**	[.00, .06]	−.45**
Cognitive Abilities	−0.24**	[−0.35, −0.13]	−0.19	[−0.28, −0.10]	.03**	[.00, .05]	−.49**
State Optimism	0.20*	[0.03, 0.37]	0.11	[0.02, 0.20]	.01*	[−.01, .02]	−.32**
Neuro Pain	0.19**	[0.05, 0.34]	0.11	[0.03, 0.20]	.01**	[−.00, .02]	.37**	R2 = .480** 95% CI [.40,.53]

Note. N = 373. A significant b-weight indicates that both the beta-weight and semi-partial correlation are also significant; b represents unstandardized regression weights; beta indicates the standardized regression weights; sr² represents the semi-partial correlation squared; r represents the zero-order correlation; CI represents confidence interval; LL and UL indicate the lower and upper limits of a confidence interval, respectively.

^*indicates p < .05.

^**indicates p < .01.

4.5. Path Models showing relationships of CRF with PNP Symptoms

Figures 1 through 3 display the outcomes of the path model analysis. A significant path coefficient with a significant chi-square (p<.05) were present for each of the suggested paths. Model 1 had the fewest relevant variables, just CRF, bodily pain, and depression, according to a previous report (Ho et al., 2015). The goodness of fit indices consistently demonstrated that this model provided a good fit to the data (CFI = 0.997, RMSEA = 0.041, 90% CI 0 – 0.15, SRMR = 0.02). Based on prior research (Lukkahatai et al., 2016, Gu et al., 2023), model 2 was the most intricate of the three and contained, with the exception of sleep disturbance, the majority of the relevant predictors in the final regression model. However, this model performed poorly, in terms of how well the model fit the data (CFI = 0.895, RMSEA = 0.145, 90% CI 0.11 – 0.18, SRMR = 0.088). Our proposed model, Model 3, combined Models 1 and 2. The CFI value for this model was 0.993, the RMSEA was 0.047, with 90% Confidence Interval from 0 to 0.12, and the SRMR was 0.027.

Figure 1.

Results of Path Model 1:

Note. N = 373. → represents regression path with path coefficient. A simple mediation model using cancer-related fatigue (CRF) as the mediator. *** indicates p < .001.

Figure 3.

Results of Path Model 3:

Note. N= 373. ---- represents covariance path; → represents regression path with path coefficient. An integrated model of CRF. Bodily pain and neuropathic pain predict CRF, and CRF predicts depression.

*** indicates p < .001.

5. Discussion

As measured by PROMIS^®, the study participants generally had worse CRF and co-occurring PNP symptoms than the general public. Our key findings were as follows: (1) The severity of CRF was inversely related to perceived cognitive abilities and bodily pain, but positively correlated with anxiety, depression, and perceived cognitive concerns; (2) Depression was the strongest predictor of CRF in this population, followed by sleep disorders, bodily pain, and perceived cognitive abilities. The majority of these findings were in line with earlier studies, but this secondary analysis uncovered several unique relationships.

It has been clearly shown that depression and CRF are related in prior research (Bower et al., 2000, Hofman et al., 2007, Mendoza et al., 2010), including longitudinal studies (Andrykowski et al., 2010, Otte et al., 2010), and even meta-analyses (Brown and Kroenke, 2009, Donovan and Jacobsen, 2011). The significant correlation between these two factors raises the possibility that both behaviors may have a shared etiology. Inflammation is the most frequently suggested common etiology for both depression and CRF.(Saligan and Kim, 2012). One could contend that utilizing depression as a therapeutic target could also reduce or eliminate the burden of CRF. However, clinical accounts of prolonged fatigue following the remission of depression, in particular, have produced conflicting results in earlier investigations, raising the possibility that both variables are separate phenomena (Ang et al., 2017).

Similar to this, a prior study looked at the relationships between optimism and CRF in post-chemotherapy breast cancer patients. That investigation showed that despite the frequent co-occurrence of CRF and depression, only depression was associated with optimism while CRF was primarily related to the severity of physical symptoms (Levkovich et al., 2015). So, the negative regression coefficient between dispositional/state optimism and CRF found in this study (see Table 4, r = −.32, p < .001) may primarily reflect the association between depression and dispositional/state optimism and less so with CRF.

The study’s noteworthy finding is the inverse correlation between bodily pain and CRF. Contrary to other research that suggested pain is a significant risk factor for CRF, patients who reported less bodily pain had worse CRF (Ma et al., 2020, Yildiz Kabak et al., 2020, Lamino et al., 2011). Very few studies have observed the inverse relationship of bodily pain and fatigue. To determine if central sensitization is linked to fatigue, regardless of the potential confounding effect of pain, one study from England involving roughly 2500 participants reported that central sensitization predicted fatigue, independent of whether there was pain (Druce and McBeth, 2019). Remember that central sensitization is a recognized mechanism for pain, defined by enhanced pain responses to physical stimuli, which are frequently unpleasant sensations (Yunus, 2007). Our study’s unique finding, along with the evidence that central sensitization can predict fatigue independent of the presence of pain, raises further questions about the potential contributions of various physiologic mechanisms, including inflammation and genetic predisposition.

We are aware that immunosuppression or inflammatory levels do not always correspond with severity of fatigue (Druce et al., 2016). Further research on the contributions of inflammation and immunosuppression to the reported PNP symptoms is warranted given that the BCS subsample included in this analysis may have varied degrees of inflammation and immunosuppression. Additionally, certain genetic polymorphisms, such as BDNF Val66Met, increase pain perceptions in chronic pain conditions (Vossen et al., 2010). Participants with CRF who do not experience bodily pain may be distinguished by identifying particular genetic variations in this subpopulation.

The path models we evaluated revealed interesting information. First, model 1 (Figure 1) showed that CRF was an effective mediator between bodily pain and depression, and CRF was a good mediator between bodily pain and depression, showing that bodily pain predicted depression indirectly through CRF, as was originally suggested by Ho et al. (Ho et al., 2015). Second, model 2 (Figure 2) was a comprehensive model based on earlier observations (Gu et al., 2023, Lukkahatai et al., 2016). Although it did not show a good fit to our data, it did help us discover that the model’s complexity made the model fit worse; necessitating the need to simplify the hypothesized model that can account for the relationships between CRF and other PNP symptoms. Third, a better model fit was obtained by combining models one and two while simplifying the model’s variables (Figure 3). This finding implies that CRF can predict depression and that both bodily and neuropathic pain can predict CRF. Furthermore, CRF mediates depression and pain (both bodily and neuropathic) in BCS. However, our existing cross-sectional data limited our ability to analyze and create a causal relationship between the variables.

Figure 2.

Results of Path Model 2:

Note. N= 373. ---- represents covariance path; → represents regression path with path coefficient. A comprehensive model of CRF. Bodily pain and neuropathic pain predict fatigue, while fatigue and state optimism predict depression, and depression predicts cognitive abilities.

*** indicates p < .001.

Limitations

The present study had several limitations. The analyses were unable to investigate causality between study variables because of the cross-sectional study design and the findings are less generalizable because of a homogeneous sample (93% White). The subjectivity of symptom reporting was another study disadvantage. Future research should include objective methods for evaluating symptoms and enrolling a more diverse sample, particularly to further study the negative correlation between CRF and bodily pain. In addition, it is crucial to take into account the psychological aspects that contribute to CRF, such as social support, affect, and catastrophizing.

6. Conclusions

The study concludes that depression, sleep disturbance, bodily pain, neuropathic pain, perceived cognitive ability, and dispositional optimism are important predictors of CRF. Furthermore, a good fit mediation model with significant pathways for CRF is shown, where CRF is a critical mediator of the relationships between bodily pain, neuropathic pain, and depression. Important considerations on the role of psychosocial factors in the development and persistence of CRF during cancer survivorship are critical to treat and manage this distressing condition affecting lives of patients and their families.

Our findings suggested that longitudinal assessments of late and long-term cancer-related behavioral toxicities, including PNP symptoms, must be included in follow-up evaluations. An important step in managing the comorbidity of symptoms may be depression screening. Evidence-based strategies/interventions are needed to reduce or eliminate the burden of CRF and other PNP symptoms in BCS.

Highlights.

• Worst fatigue is associated with high depression, anxiety, and cognitive concern.

• Low perceived cognitive ability and less bodily pain correlate with severe fatigue.

• Cancer-related fatigue is a key mediator between pain and depression.

• Evidence-based strategies are needed to palliate psychoneurophysiological symptoms.