Fatigue severity and avoidance among individuals with chronic disease: a meta-analysis

Abstract

Objective:

Fatigue is a common, debilitating symptom experienced by individuals with chronic disease. Avoidance, or the act of evading unwanted experiences, is associated with fatigue across chronic disease samples. The current study sought to determine the strength of association between fatigue severity and avoidance in individuals with chronic disease.

Methods:

PubMed, PsycINFO, CINAHL, and ProQuest Dissertations and Theses databases were searched. Eligible studies measured fatigue and avoidance in chronic disease samples. Sixty-six studies were included. Data analyses were conducted in Rstudio. A random effects model was employed, and a weighted mean effect size was computed for fatigue severity and avoidance. Mixed-effects meta-regression analyses were conducted to examine moderating variables, including patient, clinical, and measurement characteristics. Publication bias was examined using funnel plot, trim-and-fill, and p-curve.

Results:

The meta-analysis comprised of 71 unique patient samples from 66 studies. The total number of included participants was 13,024. A small, positive association was found between fatigue severity and avoidance, r(71) = .22, p < .001, 95% CI [.18-.27], SE = .02. There was also significant heterogeneity, Q(70) = 349.96, p <.001. Moderator analyses examining age, sex, illness duration, avoidance type, and disease sample were all non-significant. Regarding publication bias, trim-and-fill resulted in a modified weighted mean effect size (r(83) = .18, p < .001) and a p-curve analysis supported the evidential value of the current analysis.

Conclusion:

Findings support that among individuals with chronic disease, fatigue severity and avoidance are positively associated, which has implications for behavioral interventions in this population.

Introduction

Fatigue, characterized as a feeling of exhaustion or lack of energy and reduced capacity in functioning [1–3], is a nonspecific symptom experienced by most at some point across the lifetime. Fatigue is common among healthy individuals and frequently reported by patients with chronic conditions. Fatigue is prominent in many cancers, and prevalence of fatigue has been reported in patients after stroke [4] and patients with rheumatic diseases [5, 6], inflammatory bowel disease [7], multiple sclerosis [8], and systemic lupus erythematosus [9, 10], to name a few of many. More recently, as individuals recover from COVID-19, many experience persistent fatigue post-recovery [11].

Fatigue is also associated with chronic disease treatment, including chemotherapy, radiotherapy, hemodialysis, and immunotherapy [12]. Although biological mechanisms of fatigue have been investigated [13], no confirmatory biomarker exists, which complicates diagnosis and treatment. Although fatigue among healthy individuals is relieved through rest, individuals with chronic disease experience fatigue that is severe, persistent, and debilitating.

Fatigue experienced by those with chronic disease is associated with significant impairment across life domains, including physical and mental functioning [14] and reduced quality of life in children, adolescents, and adults [15–18]. It is linked to adverse outcomes such as disability, work absenteeism, and productivity loss [15, 19–22]. Patients with chronic disease frequently identify fatigue as one of the most distressing symptoms [23], adversely affecting daily life [24, 25], and more disruptive than pain [26]. Given that fatigue management impacts patient outcomes (e.g., disability and depression), measuring and attending to fatigue may be important in promoting early intervention. `

Better understanding fatigue maintenance could provide insight into how fatigue-related disability and impairment develop and whether there is a temporal window during which intervention is most effective. Given that treatment regimens may induce fatigue, reducing the impact of existent fatigue may be more practical. Considering the prevalence across numerous chronic diseases, identifying transdiagnostic risk factors may be one way to better understand fatigue maintenance. Transdiagnostic risk factors, or risk factors that are related to various diseases, are associated with significant health problems among those with chronic diseases, including chronic pain and distress [27]. Since having a chronic disease is associated with fatigue, targeting risk or maintenance processes may be helpful at fatigue onset, especially among cases where etiology is undetermined. One such process that may be a risk or maintaining factor for fatigue in individuals with chronic disease is avoidance.

Avoidance

Avoidance is the universal tendency to evade unwanted experiences [28]. Individuals can engage in different types of avoidance, whether trying to prevent internal experiences such as thoughts or emotions [29], or reducing activity or movement [30]. Avoidance can be an adaptive response under many circumstances when applied in the short-term. For example, if an avid runner sprains an ankle, avoiding running temporarily until the injury is healed is likely recommended. Avoidance, however, can become maladaptive when rigidly applied over time. For instance, individuals who experience chronic pain that avoid physical activity for extended periods of time are at risk for increasing pain intensity, deconditioning, and depression [30].

As such, it is hypothesized that avoidance (broadly) by individuals with chronic disease experiencing fatigue can become problematic when used persistently. This is in line with previous avoidance models, which have theorized that the cost of engaging in persistent avoidance can impair one’s functioning [30]. Fatigued individuals with chronic disease in particular may be at a heightened risk of using avoidance long-term, which may be ineffective and lead to paradoxical results. As indicated earlier, the fatigue experienced by individuals with chronic disease is more severe and persistent than fatigue in healthy individuals, and it is typically not relieved by rest. In healthy individuals, choosing to avoid something (e.g., activity) in response to fatigue can be effective because once fatigue severity is reduced, engagement in daily activities can resume. Conversely, in an individual with chronic disease, the same form of avoidance (e.g., activity) will likely not lead to a reduction in fatigue severity. In fact, prolonged periods of rest may instead increase fatigue severity because fatigue is expected to be persistent in chronic disease samples. As such, persistent avoiding (e.g., experiential avoidance, avoidance of movement, avoidance of fatigue), by individuals with chronic disease may become maladaptive when strictly applied long-term.

Again, it is important to note that avoidance is a common behavior that is effective in some contexts, but pervasive, repeated avoidance is likely to interfere with pursuit of personal values. The association between fatigue and avoidance has been examined across chronic disease samples. For example, fatigue and avoidance have been significantly associated in patients with breast cancer [31], chronic fatigue syndrome, rheumatic diseases [32], and multiple sclerosis [33]. Studies examining the association between fatigue and avoidance in chronic disease samples use a range of different measures. For example, general avoidance of internal experiences (experiential avoidance) is measured with the Acceptance and Action Questionnaire-II [34] (sample item: “I worry about not being able to control my worries and feelings”). Chronic disease-specific avoidance is measured with the Brief Coping Orientation to Problems Experienced [35] questionnaire (sample item: “I’ve been giving up the attempt to cope”). Fatigue-specific avoidance is also measured, in modified scales, such as the modified Tampa Scale for Kinesiophobia [36] (sample item: “If I would try to overcome it, my fatigue would increase.”).

Objectives of the present study

Although there is ample research examining the relation between fatigue severity and avoidance, the strength of these associations across chronic disease samples has not yet been synthesized. Meta-analysis may inform interventions for individuals with chronic disease experiencing fatigue, including identification of avoidance as a key process. The current study assessed the strength of association between fatigue and avoidance in patients with chronic disease, and a small-to-moderate positive association was hypothesized. Moderating variables were examined to determine the role of patient, clinical, and measurement characteristics.

Method

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMSA) protocol was used [37] and is provided in supplementary material (Table S1). The study protocol was not registered.

Inclusion criteria

Studies meeting the following criteria were considered appropriate for inclusion: (1) publication in English; (2) human subjects; (3) diagnosis of a chronic physical disease (a disease that has lasted or is expected to last one-year or longer, or results in the need for continuing medical care or impairment to activities of daily living [38]); (4) studies including self-report questionnaires of fatigue severity and/or impact; (5) studies including self-report questionnaires of general, fatigue-, or disease-specific avoidance; (6) and the inclusion of Pearson correlations between fatigue severity and avoidance. Measures of fatigue severity and impact were both eligible for inclusion, as self-report questionnaires of fatigue impact are commonly used to identify greater levels of fatigue. Studies did not have to meet a clinically meaningful fatigue threshold of fatigue.

Unpublished studies, including dissertations and theses, were included to reduce publication bias [39]. Only empirical studies were examined. If an experimental, longitudinal, or intervention study met the above criteria, only baseline or pre-intervention data was utilized. If a study included appropriate fatigue and avoidance measures, but did not report Pearson correlations, the corresponding author was contacted to request data. Contacted authors were asked to provide other relevant data. Only the effect size of Pearson correlations between fatigue and avoidance were included due to recent research suggesting that prior conversions from beta-coefficient to correlations [40] can produce large biases [41].

Exclusion criteria

Studies were excluded if published in a language other than English, the sample were non-human subjects (i.e., animals), participants were nonchronically ill, or the study included experimentally-induced fatigue. Qualitative studies and case studies were excluded.

Chronic psychological conditions (e.g., depression) were excluded. Although depression and fatigue are theorized to have similar biological mechanisms [13, 42], and fatigue and loss of energy are included in the diagnostic criteria for depression [43], the onset of depression can also be in response to chronic disease. To avoid overlap in conceptualizing depression as an inciting event and as an adverse outcome, chronic psychological conditions were excluded.

Search Strategy

A Psychology and Education librarian assisted with the database search strategy. The final search was conducted on June 28, 2021, in the following databases: PubMed, PsycINFO, CINAHL (EBSCO), and ProQuest Dissertations and Theses. The PubMed search strategy, including key word searches and MeSH terms, can be found in the supplementary material (Table S2). Search strategies for other databases are available upon request. All results of the search were then uploaded to Rayyan [44], a free web-based tool for screening records for inclusion in systematic reviews. Duplicate citations were removed. Initial screening of title and abstracts was conducted by the first author (JLA). If the title and abstracts appeared to meet inclusion criteria, the full text was then examined for eligibility. See Figure 1 for the PRISMA flow diagram.

Figure 1.

PRISMA flow diagram.

Selection of predictor variables

Measures of fatigue severity and impact were included. Most measures provided a total or global score of general fatigue. Some measures included subscales, such as physical or mental fatigue. In cases where both total and subscale scores were reported, total score was used. If multiple subscales of the same measure were reported, but not the total score, the average across subscales was computed. Studies reported visual analog scales (VAS) of fatigue, validated measures, or measures modified to be fatigue-specific. Table S3 summarizes all the fatigue scales reported in the current analysis.

General avoidance and fatigue-specific avoidance measures were included. Measures related to other specific symptom, such as pain, were excluded, whereas measures of avoidance broadly related to one’s chronic disease were included. Table S4 summarizes the avoidance scales reported in the current analysis.

Selection of moderator variables

Moderator variables for the current meta-analysis included patient, clinical, and measurement characteristics such as age, sex, illness duration, disease type, and type of avoidance measure. Data were extracted from all included studies or requested by authors if not reported. Mean age and percentage of females in the sample were examined. Both mean (preferred) or median illness duration were examined. Type of chronic disease sample was tested as a moderator; only chronic disease samples with three or more studies were examined. Finally, type of avoidance was examined (i.e., general, fatigue-, or disease-specific).

Data extraction

The first (JLA) and second authors (MVB) completed the data extraction of eligible studies. To assess for inter-rater reliability, both authors double-coded a random selection of 30% of the included studies. Inter-rater reliability fell within the strong range (K = .81) [45]. All disagreements were discussed until 100% agreement was achieved. The following variables were coded from each study if the data were available: (a) type of chronic disease sample; (b) sample size (N); (c) age of sample (mean or median); (d) sex (percent female); (e) illness duration of chronic disease in years (mean or median); (f) fatigue measure(s); (g) avoidance measure(s); and (h) Pearson correlation(s) between the fatigue severity and avoidance measure(s).

Analytic strategy

Data analysis was conducted using Rstudio software (Version 4.0.2), using the metafor package [46]. A random effects model was employed because differences in effect sizes were expected, given that studies include different patient types and different measures [47]. Evidence suggests that random effects models provide greater accuracy in measurement than fixed effects models [48]. All Pearson correlations were converted into Fisher’s z, and a single weighted mean effect size was computed for fatigue severity and avoidance. Publication bias was examined using both visual and objective measures, including a funnel plot [49, 50], Egger’s regression test [51], Begg and Mazumdar’s rank correlation test [52, 53], Duval and Tweedie’s trim-and-fill [54–56], and p-curve analysis [57]. Heterogeneity was examined with Cochrane’s Q and I² [58]. Moderator analyses were conducted using mixed-effects meta-regression, which is appropriate for continuous and categorical outcomes [59].

Results

Study and patient characteristics

Details of the included studies can be found in Table S5. A total of 66 unique studies were included in this meta-analysis, with 64 published between 1990 and 2021. One of the included studies is currently in press [60] while another is from baseline data of an ongoing, unpublished trial [61]. Six of the included studies were doctoral dissertations/theses [62–67].

Several studies [68–73] utilized multiple fatigue severity and/or avoidance measures. Two [70, 71] included two measures of fatigue severity: a validated measure (the Checklist of Individual Strength) and a VAS; the correlation between avoidance and the validated measure was retained to be maximally conservative, as a stronger correlation was observed with the VAS. Two utilized multiple validated fatigue severity measures [69, 73] and two utilized multiple validated avoidance measures [68, 74]. Scale reliabilities were not reported and thus could not be compared, so weakest correlations were retained. More specifically, for the studies with multiple fatigue severity measures, a disease-specific fatigue measure (the Irritable Bowel Disease Fatigue Scale) was retained over the Multidimensional Fatigue Inventory [69], while the Fatigue Scale of Motor and Cognitive Functions was retained over the Chalder Fatigue Questionnaire [73]. For the studies with multiple avoidance measures, the Impact of Events Scale was retained over Cognitive and Behavioral Responses to Symptoms Questionnaire [68], while the Illness Management Questionnaire was retained over the COPE questionnaire [72]. Overall, a variety of avoidance and fatigue severity measures were reported (see Tables S3 and S4).

Several studies [65, 75–77] examined more than one chronic disease sample; when available, data for each sample were examined separately. Thus, the current meta-analysis is comprised of 71 unique patient samples from 66 studies. The total number of included participants is 13,024. Sample size ranged from 21 to 1,127 participants (M = 183.40, SD = 197.23). The average percentage of females in each sample was 72.58% (range: 0 to 100). The average age was 46.88 years old (SD = 11.93, range: 14.60 to 67.80); four studies (6.06%) did not include mean age. The average illness duration (mean and median) was 8.42 years (SD = 8.19, range: .05 to 55); 24 studies (36.36%) did not report illness duration.

Nineteen chronic disease samples were represented in the 66 studies. Most studies examined patients with cancer (n=23) [31, 62, 66, 68, 78–96], spanning breast, prostate, gynecologic, lung, gastrointestinal, colorectal, leukemia, lymphoma, and brain cancer. Fifteen studies examined individuals with multiple sclerosis [61, 65, 73, 97–108]. Fourteen studies examined individuals with chronic fatigue syndrome, myalgic encephalomyelitis and/or chronic fatigue immune dysfunction [32, 60, 63–65, 70–72, 76, 109–113], five with rheumatoid arthritis [65, 75, 114–116]; and two studies with fibromyalgia [65, 77], irritable bowel disease [69, 77], systemic lupus erythematosus [65, 67], and heart disease or heart failure [117, 118]. One study was included examining: end-stage kidney disease [119], vulvodynia [120], asthma [76], antineutrophil cytoplasmic antibody associated vasculitis [75], atrial fibrillation [121], stroke [122], non-specific chronic low back pain [123], and late-onset sequelae of poliomyelitis [124]. One study combined a group of autoimmune rheumatic diseases (e.g., connective tissue disease, spondyloarthropathy, and systemic sclerosis) [116], whereas another included two separate groups, one comprised of patients with autoimmune conditions, and the other with a functional somatic syndrome [65].

Preliminary Analyses

Preliminary analyses were conducted to investigate outliers and influential cases [125, 126]. Two studies were identified as potential outliers and influential [82, 115] and were rechecked for coding accuracy. These studies observed correlations between fatigue severity and avoidance that were moderately-sized and negative (−.30 and −.28, respectively) and authors retained both studies to avoid biasing results. It is possible that these studies were outliers because of the patient samples; one included a sample of patients with cancer who were adolescents, and the other included a sample of patients with rheumatoid arthritis in the early stages.

Effect size analyses and heterogeneity

For primary analyses, an effect size and variance for each study was computed; then, a single weighted mean effect size was calculated (Table 1). The weighted mean effect size of the association between fatigue severity and avoidance was significant, r(71) = .22, p <.001, 95% CI [.18-.27], SE = .02. This finding indicates that among individuals with chronic disease, those with greater levels of fatigue report more engagement in avoidance, and those who report more engagement in avoidance report greater fatigue severity. Figure 2 illustrates a forest plot with the computed weighted mean effect, study effect sizes, and confidence intervals. Cochran’s Q tests of heterogeneity revealed significant variation between study effects, Q(70) = 349.96, p <.001, indicating between- and within-study variability and justifying moderator analyses. Of the total heterogeneity, I² indicated that 81.50% reflects true differences in effect size, or between-study differences. According to I² thresholds [127], this percentage reflects high variance. These findings support the use of a random-effects model and justify moderator analyses.

Table 1.

Effect size r estimates for the association between fatigue and avoidance.

	r	95% CI	z	k	Q	I2
Total	0.22	0.18 to 0.27	10.12***	71	349.96***	81.50
General avoidance	0.19	0.08 to 0.31	3.22**	15	81.79***	87.80
Fatigue-specific avoidance	0.28	0.21 to 0.35	7.50***	8	10.27	43.10
Disease-specific avoidance	0.23	0.18 to 0.28	8.77***	48	207.64***	79.74

^*p <.05

^**p < .01

^***p < .001

Figure 2.

Fatigue and avoidance among individuals with chronic disease – forest plot.

Weighted mean effect sizes were computed to examine the associations between fatigue severity and general, fatigue-, and disease-specific avoidance. Regarding fatigue severity and general avoidance, the weighted mean effect size indicated a positive association (r(15) = .19, p < .01). The weighted mean effect size for fatigue severity and fatigue-specific avoidance (r(8)= .28, p < .001) and disease-specific avoidance (r(48) = .23, p <.001) demonstrated positive associations.

Moderator analyses

Moderator analyses were conducted to investigate variables that may explain between-study and within-study variance (see Table 2). Specifically, mixed-effects meta-regression was used to examine whether the association between fatigue severity and avoidance was moderated by age, sex, illness duration, disease type, and type of avoidance. Findings revealed that age, sex, and illness duration did not have a significant moderating effect. The association between fatigue severity and avoidance was not significantly affected by avoidance type. Regarding chronic disease type, samples examined included cancer, multiple sclerosis, chronic fatigue syndrome, autoimmune rheumatic diseases (i.e., rheumatoid arthritis, systematic lupus erythematosus, connective tissue disease, and spondyloarthropathy [5]) and cardiovascular diseases (i.e., heart disease, heart failure, atrial fibrillation, and stroke). The association between fatigue severity and avoidance was not significantly moderated by chronic disease type.

Table 2.

Tests of mixed-effects meta-regression for moderation of association between fatigue and avoidance.

Moderator	Point estimate	95% CI	z	t	k	F	df	Q
Age	0.0028	−0.0010 to 0.0066	1.43		67			2.03
Percent Female	0.0006	−0.0014 to 0.0025	0.57		70			0.32
Illness duration	−0.0032	−0.0096 to 0.0031	−1.00		47			0.99
Type of avoidance measure					71	0.66	2, 68
General avoidance	0.1801	0.0822 to 0.2781		3.67***	14
Fatigue-specific avoidance	0.0906	−0.0748 to 0.2560		1.09	8
Disease-specific avoidance	0.0487	−0.0628 to 0.1602		0.87	48
Chronic disease sample					59	1.68	4, 54
Cancer	0.2278	0.1536 to 0.3019		6.16***	24
Multiple sclerosis	0.0397	−0.0799 to 0.1594		0.67	14
Chronic fatigue syndrome	−0.0798	−0.2109 to 0.0512		−1.22	12
Autoimmune rheumatic diseasesa	−0.1347	−0.3111 to 0.0417		−1.53	5
Cardiovascular diseasesb	0.1106	−0.1065 to 0.3277		1.02	4

^*p <.05

^**p < .01

^***p < .001

^aAutoimmune rheumatic diseases include rheumatoid arthritis, systemic lupus erythematosus, connective tissue disease, spondyloarthropathy (seronegative spondylarthritis)

^bCardiovascular diseases include heart failure, heart disease, atrial fibrillation, and stroke.

Publication bias

A funnel plot was examined visually, appearing asymmetrical. A trim-and-fill method imputed “missing” studies into the analyses to adjust for asymmetry in the funnel plot. Following the trim-and-fill, twelve studies were imputed to the left of the mean (Figure 3). Although the trim-and-fill analysis resulted in a modified weighted mean effect size, the association between fatigue severity and avoidance remained statistically significant, r(83) = .18, p <.001, 95% CI [.13-.22], SE = .02. Both Egger’s regression test (z = .82, p = .41) and rank correlation test (Kendall’s tau = .01, p = .89) were non-significant, suggesting insufficient evidence of publication bias.

Figure 3.

Funnel plot of standard error following trim-and-fill method.

A p-curve examined if significant studies contained evidential value. A right-skewed distribution indicates true effects with more p-values on the lower end (.01), rather than the higher end (.04 or .05), to provide an indication of the degree of p-hacking and selective reporting of significant effects [57]. Notably, the p-curve analysis only examines statistically significant findings. As shown in Figure 4, the p-curve of included statistically significant studies includes mostly (84%) those at the ≤.01 significance value suggesting evidential value, Z = −18.97, p <.001. Although some evidence of publication bias (i.e., funnel plot) was observed, several empirical tests supported the evidential value of the current analysis (i.e., regression test, rank correlation test, and p-curve). Even after adjusting for probable publication bias (i.e., trim-and-fill), the association between fatigue severity and avoidance remained significant.

Figure 4.

P-curve distribution for p-values of statistically significant effects, p < .05.

Discussion

The current meta-analysis examined the association between fatigue severity and avoidance among individuals with chronic disease. The findings show a modest, yet significant, positive association between fatigue severity and avoidance. Given the current meta-analysis draws from correlational data, causality and directionality of this relation cannot be inferred. Tests of heterogeneity in the current analysis indicated there was a significant amount of between-study and within-study variability; however, moderator analyses of age, sex, illness duration, and type of disease and avoidance were all non-significant.

Although the association with fatigue severity appeared strongest with fatigue-specific avoidance, type of avoidance did not moderate the association. Notably, fatigue-specific avoidance was the least commonly measured domain among the included studies. In the 8 studies that examined fatigue-specific avoidance, four total measures were used: two that were validated measures modified for fatigue severity (Coping Strategies Questionnaire [128] and Tampa Scale for Kinesiophobia [36]), and two multi-item scales created that were not tested beyond initial development (Avoidance of Activity [129] and Fatigue Coping Behavior Scale [114]). These findings suggest there may be a paucity of fatigue-specific avoidance measures. Without widely available and utilized measurement of this domain, both researchers and clinicians remain unaware of an important fatigue management strategy. Future research should examine the relation between fatigue severity and fatigue-specific avoidance, including the development and validation of fatigue-specific avoidance scales that can be used across a range of samples and include multiple domains of avoidance (i.e., physical activity or cognitive avoidance).

Relatedly, over twenty different fatigue severity and impact measures were used across studies. Moderation of the type of fatigue measure used (i.e., severity or impact) was not explored in the current study, as most measures of fatigue include items of both severity and impact. For example, the Brief Fatigue Inventory asks participants to rate the level of fatigue right now on a scale of 0 (no fatigue) to 10 (as bad as you can imagine) and to rate how much fatigue has interfered across several life domains on a scale of 0 (does not interfere) to 10 (completely interferes). It also appears that many measures of fatigue do not utilize subscale scores, but rather total global indexes. Regardless of type, the measurement of fatigue, in chronic disease is common. A future direction in fatigue measurement may be re-evaluation of the measures being used to ensure that fatigue severity and fatigue impact are not being conflated. This would allow for distinction between the experience of intense fatigue versus the impairment caused by fatigue. A critical review that examined fatigue severity measures for chronic disease [130] found six scales that demonstrated good psychometric properties; nearly all of the six were included in studies examined herein. Given the prevalence of fatigue across disease populations, future research may benefit from the development and validation of fatigue severity measures that are suitable across disease samples. One existing measure that may be most appropriate is the Patient-Reported Outcomes Measurement Information System (PROMIS) fatigue item bank [131]. The PROMIS fatigue measure has been evaluated across multiple chronic disease populations, is responsive to symptom change across conditions, and is deemed appropriate for between-group comparisons [2]. Surprisingly, none of the included studies used the PROMIS fatigue item bank for the measurement of fatigue.

Given tests of heterogeneity showed a significant amount of between-study and within-study variability, several moderator analyses tested the impact of several patient and clinical demographics on the association between fatigue severity and avoidance. All the variables examined, however, did not significantly impact the strength of association between fatigue severity and avoidance. It is possible that the inclusion criteria for the current study may have had an impact on these analyses. Per the definition of chronicity, it is possible that the length of illness was non-significant due to range restriction. The moderator analyses of age and gender were also non-significant. Though limited, these findings align with other work that examined the role of cognitive and behavioral avoidance on fatigue in patients with chronic fatigue syndrome [132]. Regarding the non-significant effect of age, range restriction among included studies may have impacted findings.

The current study also examined the impact of type of chronic disease. Moderator analyses found that the strength of the association between fatigue severity and avoidance was not impacted by chronic disease type. This finding suggests that different disease groups (cancer, multiple sclerosis, chronic fatigue syndrome, autoimmune rheumatic diseases, and cardiovascular diseases) may not differ from each other regarding the strength of the fatigue-avoidance relation. Other research has found that fatigue severity in chronic disease is largely explained by transdiagnostic factors, over and above chronic disease type [14]. As such, transdiagnostic approaches may broaden the reach of interventions that apply widely to chronic disease types. Future research should continue to examine how chronic disease type impacts the relation between fatigue severity and avoidance. The current study examined all cancers as one disease category. Future work may benefit from examining the relation between fatigue and avoidance among different categories of cancer.

Clinical Significance

Findings support a significant, albeit small, association between fatigue severity and avoidance. As both fatigue and avoidance impact multiple domains of functioning, psychological interventions targeting avoidance more broadly may improve fatigue, as well as other areas of functioning (e.g., symptoms of depression or anxiety). Some prior research has shown support for cognitive-behavioral therapy (CBT) and exercise-related interventions (e.g., graded exercise therapy) for treatment of fatigue. These interventions may be targeting avoidance. In CBT, patients often learn to identify avoidance and approach feared situations. In exercise-based interventions, patients are taught to approach, rather than avoid, physical activity. Relatedly, preliminary support exists for third-wave CBT approaches, such as acceptance and commitment therapy (ACT), for the treatment of fatigue [133–135]. ACT targets experiential avoidance, or the avoidance of private experiences (e.g., thoughts). The effectiveness of these CBT and graded exercise therapy in particular for managing fatigue severity is contentious in chronic fatigue syndrome [136]. More research on the treatment-related mechanisms of change could shed light on the effect avoidance has on fatigue outcomes, as well as other therapeutic processes. Finally, recent findings regarding the COVID-19 pandemic have found that 20% of individuals who recovered from the illness experienced elevated fatigue post-diagnosis [137]. Given the increased likelihood of individuals experiencing post-COVID-19 fatigue worldwide, identifying malleable key processes that could be targeted with early intervention is critical.

Limitations

Several limitations should be considered. The study protocol of this meta-analysis was not pre-registered. Although great attempts were made to include all possible studies meeting inclusion criteria, many articles identified for inclusion did not report the Pearson correlation between fatigue severity and avoidance directly, though data were requested. Though the search strategy for chronic disease inclusion was broad, not all categories were reflected in the included samples. Future research would benefit from the examination of fatigue severity and avoidance in more diverse chronic disease samples. Relatedly, the current study only examined chronic physical health conditions. As such, results cannot generalize to chronic psychological conditions or to non-chronically ill individuals. Future research should continue to examine the relationship between fatigue and avoidance in more populations, such as individuals with depressive disorders and healthy individuals. Inclusion criteria also included measures of both fatigue severity and impact. Measures of fatigue impact appear to use the impact on daily activities from fatigue (i.e., such as impairment in motivation and functioning) as a proxy for fatigue severity. Thus, it is possible that some of the measurements used may assess different aspects of fatigue. This limitation should be considered when interpreting the results of this meta-analysis and these measures more broadly. Relatedly, inclusion criteria did not specify a threshold for what is considered “elevated” or “clinically meaningful” to capture the full range of fatigue severity. Lastly, as the meta-analysis draws from correlational data; directionality of this relation cannot be inferred.

Conclusion

Fatigue is a common and impairing symptom experienced among individuals with chronic disease. It is related to adverse outcomes, such as reduced health-related quality of life, psychological distress, and impairment in functioning. Avoidant responses to fatigue may contribute to adverse outcomes. Avoidance is examined as a transdiagnostic process and a risk and/or maintenance factor for significant adverse physical health outcomes. The current study aimed to synthesize the strength of the association across chronic disease samples. A small positive association was found between fatigue severity and avoidance, such that those with greater fatigue severity reported more engagement in avoidance, and those with more engagement in avoidance reported greater fatigue severity. Future research should focus on the development and validation of fatigue-specific avoidance measures as well as fatigue severity measures that can be used across chronic disease samples.