Religion, Spirituality, and Physical Health in Cancer Patients: A Meta-Analysis

Abstract

Background

Whereas religion/spirituality (R/S) is important in its own right for many cancer patients, a large body of research has examined whether R/S is also associated with better physical health outcomes. This literature has been characterized by heterogeneity in sample composition, measures of R/S, and measures of physical health. In an effort to synthesize previous findings, we conducted a meta-analysis of the relationship between R/S and patient-reported physical health in cancer patients.

Methods

A search of PubMed, PsycInfo, CINAHL, and Cochrane Library yielded 2,073 abstracts, which were independently evaluated by pairs of raters. Meta-analysis was conducted on 497 effect sizes from 101 unique samples encompassing over 32,000 adult cancer patients. R/S measures were categorized into affective, behavioral, cognitive, and ‘other’ dimensions. Physical health measures were categorized into physical well-being, functional well-being, and physical symptoms. Average estimated correlations (Fisher's z) were calculated using generalized estimating equations with robust variance estimation.

Results

Overall R/S was associated with overall physical health (z=.153, p<.001); this relationship was not moderated by sociodemographic or clinical variables. Affective R/S was associated with physical well-being (z=.167, p<.001), functional well-being (z=.343, p<.001), and physical symptoms (z=.282, p<.001). Cognitive R/S was associated with physical well-being (z=.079, p<.05) and functional well-being (z=.090, p<.01). ‘Other’ R/S was associated with functional well-being (z=.100, p<.05).

Conclusions

Results of the current meta-analysis suggest that greater R/S is associated with better patient-reported physical health. These results underscore the importance of attending to patients’ religious and spiritual needs as part of comprehensive cancer care.

Religion (e.g., religious affiliation, service attendance) and spirituality (e.g., connection to a source larger than oneself, feelings of transcendence) are important aspects of everyday life for many people; a recent poll found that 59% of people worldwide describe themselves as religious, irrespective of whether they regularly attend religious services.¹ Religion/spirituality (R/S) can be particularly important for individuals with a cancer diagnosis. The National Health Interview Survey found that 69% of cancer patients reported praying for their health compared to 45% of the general U.S. population.² R/S can help cancer patients find meaning in their illness,³ provide comfort in the face of existential fears,⁴ and receive support from a community of like-minded individuals.⁵ Much has been written about the importance of addressing spiritual needs as part of patient-centered cancer care.^6-8 Nevertheless, studies examining the effects of R/S on health outcomes in cancer patients have reported mixed results, likely due in part to small samples and heterogeneous measures of religion, spirituality, and physical health.^{9, 10} Varying definitions of R/S have also likely contributed to heterogeneous measures and, in turn, mixed results.

Greater clarity is needed to advance research and clinical applications regarding the role of R/S across the cancer continuum from diagnosis and active treatment to survivorship to end-of-life. To this end, we conducted a series of meta-analyses to quantitatively summarize associations between R/S and patient-reported mental, social, and physical health outcomes.^{11, 12} The current study focuses on physical health outcomes, including physical well-being, functional well-being, and self-reported physical symptoms.

R/S is theorized to affect physical health via two primary pathways: behavioral self-regulation and emotional self-regulation.^{13, 14} Regarding behavioral self-regulation, it has long been noted that religious affiliation and participation is associated with salutary health behaviors, such as avoidance of alcohol, tobacco, and drug use as well as fewer sexual partners.^{15, 16} Religious communities can also provide tangible support, such as transportation to medical appointments and provision of meals and other basic care needs. In contrast, religious fatalism is associated with worse compliance with screening and treatment recommendations,^{17, 18} while some religious sects proscribe health behaviors such as immunization and receipt of blood products. Regarding emotional self-regulation, religious fellowship can enhance positive emotion through social support, while R/S rituals and faith can provide hope, forgiveness, comfort, love, and other emotional benefits.¹³ A large literature has demonstrated the stress-buffering effects of R/S on physiological processes such as reduced cardiovascular reactivity, hypothalamic-pituitary-adrenal axis activation, and inflammation, which may improve health outcomes.¹⁹ In addition, because poor health habits can be viewed as behavioral attempts at emotional regulation (e.g., alcohol, tobacco, drug use, poor diet), the emotional benefits of R/S may favorably influence health behaviors.¹³ Consequently, in the current study we hypothesized that R/S would be associated with better physical health. Additional analyses examined whether dimensions of R/S were associated with specific physical health outcomes and whether sociodemographic or clinical variables (e.g., mean age of sample, gender composition, cancer type) moderated relationships between R/S and physical health. Because additional analyses were exploratory, no a priori hypotheses were made.

Methods

Definition of Terms

R/S was conceptualized a priori as composed of affective, behavioral, and cognitive dimensions.²⁰ Briefly, the affective dimension was defined as the subjective emotional experience of R/S, such as sense of transcendence, meaning, purpose, and connection to a source larger than oneself as well as struggle with or anger towards God. Affective R/S was conceptualized as one of many sources of emotional well-being; thus there is overlap between the two constructs but also a distinction as well (cf. Salsman et al.,¹¹). The behavioral dimension was defined as use of religious or spiritual practices or behaviors to manage stress and life events related to cancer and its treatment, such as meditation; prayer; pursuing connection with God; attending religious services; and strengthening connection with religious persons, activities, or groups. The cognitive dimension was defined as statements an individual believes to be true about R/S, such as causal attributions, spiritual post-traumatic growth, religious fatalism, and intrinsic religious or spiritual beliefs. Measures that did not fit well into these dimensions or encompassed multiple dimensions were included in an ‘other’ category (e.g., religious social support, religious affiliation). Physical health was defined a priori as physical well-being (i.e., ability to carry out activities of daily living, from basic self-care to more strenuous physical activities), functional well-being (i.e., perceived difficulties fulfilling roles at home, work, or in the community due to physical health), and self-reported physical symptoms (i.e., fatigue, pain, sleep, cognition, other physical symptoms).

Search Strategy

Standardized search strategies were developed and applied. Four electronic databases [PubMed, PsycInfo, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and The Cochrane Library] were systematically reviewed using controlled vocabulary terms for R/S (e.g., religio*, spiritual*), cancer (e.g., neoplasms, cancer, leukemia), and health outcomes (e.g., measure, scale, outcome*), which were specific to each database. Details regarding the search strategy are presented in Online Supplement I. The searches were inclusive of studies published in English from the earliest publication date available in each database through December 20, 2013. Unpublished studies were requested via professional listservs of the Society of Behavioral Medicine (Cancer Special Interest Group), American Psychosocial Oncology Society (Cancer Survivorship Special Interest Group), and the American Psychological Association, Divisions 36 (Religion and Spirituality) and 38 (Health Psychology).

Inclusion Criteria

Studies were included if they: (1) had an adult sample (≥18 years old) with a current or past diagnosis of cancer or undergoing procedures for a diagnosis of cancer at the time of study entry; (2) assessed an R/S variable; (3) assessed a physical, mental, or social health variable; and (4) reported an effect size measuring the bivariate association between R/S and health. Descriptive and measure development or validation studies were included if they otherwise met study inclusion criteria. Data from intervention studies were included except when the relationship between the R/S and health outcomes was potentially confounded with intervention effects. Excluded studies were: (1) qualitative assessments of R/S or health; (2) R/S intervention studies if the intervention was the only index of R/S; (3) needs assessments of R/S or health; and (4) caregiver or pediatric samples.

Screening Procedures

Individual rater pairs reviewed abstracts to determine which articles merited full review. All abstracts were reviewed independently and discrepancies were resolved by consensus. Studies meeting or possibly meeting (i.e., unable to determine based on the abstract) all criteria underwent full-text review. Four rater pairs independently reviewed the studies and entered data abstracted from the articles. R/S measures were categorized as affective, behavioral, cognitive, or other. Outcome measures were categorized as physical, social, mental, or other. Data from subscale scores, if available, were used in place of total scores to maximize analytic flexibility and improve interpretability of findings. Discrepancies were resolved by rater consensus. When articles provided insufficient data, attempts were made to contact authors for the required details.

Effect Size Measures

The meta-analysis was limited to studies that reported measures of bivariate association between R/S and physical health. Thus, the following types of effect size statistics were included: Pearson product-moment correlation coefficients, Spearman rank correlation coefficients, standardized mean differences between groups, and odds ratios based on dichotomizing two continuous variables. If an effect size estimate was not directly reported, an attempt was made to calculate it based on information provided in the article. Multivariable measures of association, such as regression coefficients or partial correlation coefficients, were excluded because they are not generally comparable to measures of bivariate association and present considerable analytic complications.²¹

Effect sizes were coded in a standard way such that a positive relationship between an R/S and health outcome reflected more R/S and better health. For purposes of meta-analysis, all reported effect sizes were converted to the Fisher z scale, a non-linear transformation of Pearson's correlation measure. The Fisher z scale was chosen because it normalizes and stabilizes the sampling variance of Pearson correlation coefficients that constitute the majority of effect size measures in the meta-analysis and because it has an unbounded range.²² Results are presented in z scale units, which were found in the current meta-analysis to be comparable in magnitude to reverse-transformed Pearson correlations.

Moderator Variables

Based on the existing literature base,^23-25 we anticipated relationships among R/S and health variables might vary as a function of several demographic and clinical characteristics. Accordingly, we coded articles to examine the degree to which these characteristics might moderate relationships among R/S and health indices. Demographic characteristics included: gender, age, race, and geographic origin of the sample. Clinical characteristics included: cancer type, cancer stage, and phase of the cancer continuum (i.e., diagnosis/treatment, post-treatment survivorship, end-of-life).

Meta-Analytic Procedures

Many studies meeting inclusion criteria provided effect size data for multiple measures of R/S, multiple measures of health outcomes, or both, all based on a common sample of participants.^† Consequently, a generalized estimating equation (GEE) approach and robust variance estimation were used to estimate average effect sizes and meta-regressions, which allowed for valid inferences even when the covariance structure of effect sizes drawn from common studies is unknown or misspecified.²⁶ For inference regarding single meta-regression coefficients (e.g., for testing the difference in magnitude of effect sizes between the affective and behavioral R/S dimensions), we employed robust t-tests involving small-sample corrections proposed by Tipton.²⁷ For inference regarding multiple meta-regression coefficients (e.g., for simultaneously testing differences between all four R/S dimensions), we used robust Wald test statistics that follow chi-square reference distributions when the number of independent studies is large.^‡28 Weights for the GEE analysis were determined based on a hierarchical model containing between-study and within-study random effects.

Results

Study Selection

As shown in Figure 1, a total of 2,073 unique abstracts were identified through electronic databases. Full text was retrieved for 293 studies, encompassing 2,648 unique effect sizes. For physical health outcomes, 883 effect sizes from 183 unique samples were identified. Of these, 128 effect sizes were excluded because they focused on a physical health outcome that: (1) had been diagnosed prior to the assessment of R/S (e.g., comorbidities, disease severity, recurrence); (2) was examined in only a small number of studies (e.g., cytokines, health behaviors); or (3) was context-dependent (e.g., end-of-life care, healthcare utilization). An additional 112 effect sizes were excluded due to missing data, and 146 were excluded because they were not bivariate correlations. Consequently, the current study examines 497 effect sizes from 101 unique samples. Characteristics of included studies are described in Online Supplement II.

Figure 1.

Flow chart of selection of studies

Study Participants

Included samples comprised a total of over 32,000 patients. Individual samples ranged in size from 11 to 8,000 patients, with a median of 153 and a mean of 321 (SD=909) patients per sample. Across samples, the mean patient age was 57.4 years (SD=7.5). The mean proportion of female patients was 65.3 (SD=28.0). Sixty-three percent of samples came from North America; of these, the mean percentage of Caucasian patients was 70.1 (SD=32.4). Across all included samples, 26% focused on the diagnosis and treatment phase of the cancer continuum, 23% on the post-treatment survivorship phase, 13% on the end-of-life phase, and 39% on multiple phases. Regarding cancer type, 22% of samples focused exclusively on breast cancer, 22% focused exclusively on another cancer (e.g., colorectal, thoracic), and 56% focused on a mix of cancer types. Regarding stage at diagnosis, 8% focused on patients with early-stage disease (i.e., in situ or local), 17% on advanced stage disease (i.e., regional or metastatic), 42% were mixed, and 34% did not report disease stage.

Meta-Analysis

GEE analysis was conducted to determine the average effect size between R/S and physical health, pooling across all dimensions of R/S. Results are displayed in Table 1. R/S was significantly associated with physical health (z = .153, p<.001). Additional analyses were conducted to determine whether average effect size between R/S and physical health differed by R/S dimension. As shown in Table 1, dimensions significantly associated with physical health were affective, cognitive, and ‘other’ R/S (p values <.05) but not behavioral R/S (p=.57).

Table 1.

Estimated Associations between Religion/Spirituality and Overall Physical Health

Religion/Spirituality Dimension	Estimate (SE)	t value	Number of Unique Samples	Number of Effect Sizes
All	.153 (.019)	8.24***	101	497
Affective	.263 (.023)	11.31***	55	223
Behavioral	.010 (.017)	0.58	29	96
Cognitive	.065 (.016)	3.94***	22	90
Other	.079 (.030)	2.62*	23	88

Note: Estimates are z-scale effect sizes. SE = standard error.

^*p<.05

^***p<.001

Analyses were also conducted to examine whether average effect size between overall R/S and physical health differed by physical health sub-domain, including physical well-being, functional well-being, and self-reported physical symptoms (i.e., fatigue, pain, sleep, cognition, and other physical symptoms). No differences in effect sizes were found among cognition, pain, sleep, fatigue, and other physical symptoms (p=.52) so these outcomes were collapsed into a single category labeled ‘physical symptoms.’ Physical well-being and functional well-being were found to be significantly different from one another (p<.01). Consequently, analyses were conducted to estimate effect sizes between R/S dimensions and physical well-being, functional well-being, and physical symptoms. As shown in Table 2, affective R/S was significantly associated with all three categories of physical health, cognitive R/S was associated with physical well-being and functional well-being, and other R/S was associated with functional well-being (p values <.05).

Table 2.

Estimated Associations Between Religion/Spirituality and Physical Health Sub-Domains

Religion/Spirituality Dimension	Physical Well-Being			Functional Well-Being			Physical Symptoms
	Estimate (SE)	Number of Unique Samples	Number of Effect Sizes	Estimate (SE)	Number of Unique Samples	Number of Effect Sizes	Estimate (SE)	Number of Unique Samples	Number of Effect Sizes
All	.098 (.026)***	50	119	.202 (.025)***	54	136	.154 (.024)***	61	242
Affective	.167 (.038)***	33	58	.343 (.036)***	28	61	.282 (.034)***	28	104
Behavioral	−.016 (.022)	16	24	.040 (.023)	14	26	.005 (.020)	18	46
Cognitive	.079 (.033)*	8	20	.090 (.026)**	14	31	.044 (.021)	13	39
Other	.032 (.039)	13	17	.100 (.041)*	14	18	.089 (.049)	13	53

Note: Estimates are z-scale effect sizes. SE = standard error.

^*p<.05

^**p<.01

^***p<.001

Relationships among R/S dimensions and physical health were further decomposed post hoc by categorizing R/S measures within each dimension into sub-dimensions (see Online Supplement II). Within the affective dimension, sub-dimensions were spiritual well-being (49 samples), spiritual distress (8 samples), R/S experiences (1 sample), and self-transcendence (1 sample). Within the behavioral dimension, sub-dimensions were R/S coping (21 samples), private R/S activities (7 samples), public R/S activities (4 samples), and composite R/S activities (2 samples). Within the cognitive dimension, sub-dimensions were R/S beliefs (8 samples), spiritual growth (7 samples), image of God (2 samples), locus of control (2 samples), R/S problem-solving (2 samples), religious orientation (2 samples), causal attributions (1 sample), and R/S importance (1 sample). Within the ‘other’ dimension, sub-dimensions were composite indices of R/S (15 samples), other (6 samples), religious affiliation (2 samples), and R/S social support (2 samples). Sub-dimensions assessed in less than 3 independent samples were excluded from analyses. Relationships among the remaining sub-dimensions and physical health are shown in Table 3. Sub-dimensions significantly associated with physical health were spiritual well-being and spiritual distress in affective R/S and spiritual growth in cognitive R/S (p values <.05).

Table 3.

Estimated Associations Between Religion/Spirituality Sub-Dimensions and Physical Health

Dimension	Sub-Dimension	Estimate (SE)	Number of Unique Samples	Number of Effect Sizes
Affective
	Spiritual well-being	.281 (.030)***	49	197
	Spiritual distress	.192 (.047)**	8	20
Behavioral
	Private R/S activities	.017 (.031)	7	20
	Public R/S activities	.084 (.059)	4	9
	R/S coping	−.010 (.031)	21	66
Cognitive
	R/S beliefs	.045 (.027)	8	26
	Spiritual growth	.104 (.029)*	7	17
Other
	Composite R/S	.076 (.042)	15	53
	Other	.131 (.078)	6	17

Note: Estimates are z-scale effect sizes, adjusted for number of independent samples. SE = standard error.

^*p<.05

^**p<.01

^***p<.001

Moderator Analyses

Sociodemographic (i.e., average age of the sample, gender composition of the sample, geographic origin of sample) and clinical variables (cancer type, cancer stage, phase in the cancer continuum) were examined as moderators of the magnitude of effect size between overall R/S and overall physical health. No moderators were statistically significant (p values >.08).

Reporting Bias

Threat of reporting bias was assessed through visual inspection of funnel plots of the effect sizes for each R/S dimension (see Figure 2). The plots did not display noticeable asymmetry that would indicate reporting bias, with the possible exception of the affective R/S dimension. A cluster-robust variant of Egger's test did not detect reporting bias (p=.08). The marked heterogeneity of effects within each R/S dimension was apparent from the plots. Of the effect sizes excluded due to missing data, 30 effect sizes from 10 samples were noted to be statistically insignificant. In a sensitivity analysis, imputing a value of zero for each of these effect sizes and including them with the reported effect sizes did not alter the substantive results.

Figure 2.

Funnel plots of effect sizes versus standard errors for each R/S dimension

Discussion

The current study, one of a series of meta-analyses focused on R/S in cancer patients, estimated the magnitude of associations between R/S and patient-reported physical health. Data were drawn from 101 unique samples encompassing over 32,000 patients. Although previous research examining these associations has produced mixed results, the current meta-analysis found statistically significant relationships between overall R/S and physical health. While it has long been known that R/S is important to many cancer patients, findings from the current study are noteworthy because they suggest that greater R/S is indeed associated with better physical health in this population. No sociodemographic or clinical variables moderated this association, suggesting these salutary relationships were not unique to a particular subgroup of patients.

Examination of dimensions of R/S indicated that affective R/S was most strongly related to physical health, with cognitive and ‘other’ R/S related to physical health to a lesser extent. Affective R/S was significantly associated with physical well-being, functional well-being, and physical symptoms. Both spiritual well-being and spiritual distress, affective sub-dimensions, were associated with overall physical health. These findings are intriguing in light of observational and experimental studies suggesting that positive mood results in reduced symptom reporting.^29-32 Previous research regarding affect and symptom reporting has focused on general measures of positive mood, rather than the specific emotions characteristic of spiritual well-being, such as a sense of transcendence, meaning, purpose, and spiritual connection. Due to the small number of studies examining objective health outcomes such as inflammation, disease recurrence, or survival, the current meta-analysis was unable to assess whether the association between affective R/S and physical health generalizes beyond self-reported health. Nevertheless, spiritual distress, or feelings of abandonment by God and/or one's religious community, is associated with increased depression and decreased adherence to medication and medical advice among cancer patients, suggesting a plausible pathway by which such an association could occur.^{16, 33, 34} Future research should move beyond self-report assessment to more sophisticated measures of putative biological and behavioral mechanisms of the relationship between R/S and physical health.

Cognitive R/S was also significantly associated with better physical health, including both physical and functional well-being but not physical symptoms. Regarding cognitive R/S sub-dimensions, spiritual growth but not R/S beliefs was significantly associated with physical health. Measures of spiritual growth, such as the spiritual change subscale of the Posttraumatic Growth Inventory,³⁵ focus on the extent to which cancer is perceived to have positively impacted patients’ spiritual and religious lives. Patients who experience R/S with a new depth and richness after cancer may benefit from increased access to the emotional and behavioral self-regulatory resources of R/S that are theorized to promote physical health. It should be noted that spiritual growth is often measured retrospectively; some studies have found that perceptions of growth result from patients’ denigration of their own pre-cancer attributes rather than true change.³⁶ Nevertheless, positive reappraisal of cancer via perceived spiritual growth, even if illusory, may help patients maintain a connection to a loving God despite the disappointments and stresses inherent in the cancer experience.

R/S measures categorized as ‘other’ also showed a significant association with better physical health, specifically greater functional well-being. This dimension was primarily comprised of measures of general or non-specific religiousness or spirituality that encompassed more than one of the affective, behavioral, and cognitive dimensions. Examination of ‘other’ sub-dimensions, including composite measures of R/S versus ‘other’ measures that were not easily categorized elsewhere (e.g., spiritual health, spiritual perspective), indicated non-significant relationships with physical health. Mixed findings regarding ‘other’ R/S, taken together with the non-significant association between behavioral R/S and physical health, suggests that associations between R/S and physical health are not robust across all R/S measures. This pattern of results underscores the importance of using a well-defined R/S taxonomy to guide future research questions.

To our knowledge, the current study is the first meta-analysis of R/S and patient-reported physical health in cancer patients. Strengths include an innovative and important research question, a large pool of samples from which data were drawn, and rigorous statistical methodology. In addition, examination of dimensions of R/S was based on an a priori taxonomy. Limitations should also be noted, however. Studies included in the meta-analysis tended to collect data from small samples of convenience, with nearly half of the studies including less than 150 participants. Moreover, methods sections were often incomplete in regards to measures, procedures, and sample descriptions. In addition, R/S measures varied in quality; while many effect sizes resulted from psychometrically-sound measures such as the Functional Assessment of Cancer Therapy – Spirituality (FACIT-Sp),³⁷ the Religious COPE,³⁸ or the Post-Traumatic Growth Inventory,³⁵ other studies used measures that were more obscure or investigator-derived. In addition, the current meta-analysis was comprised of bivariate correlations drawn mostly from cross-sectional studies, which prevents making inferences regarding the strength or directionality of the causal relationship between R/S and physical health. It could be argued that physical health may contribute to enhanced R/S. For example, patients in better health may be better able to experience greater meaning and peace (i.e., affective R/S), attend religious services (behavioral R/S), and perceive God as benevolent (i.e., cognitive R/S) relative to patients in poor health. The small number of longitudinal studies in the meta-analysis precluded analysis of study design as a moderator of effect size. Nevertheless, findings of physical benefits from interventions designed to enhance R/S indicate that R/S does contribute to better health.^39-41 Additional research is needed to parse the contributions of R/S to physical health and vice versa.

In summary, the current meta-analysis contributes to the large body of literature on R/S in cancer by confirming that R/S is associated with better patient-reported physical health. This finding underscores the need for timely and culturally-sensitive provision of religious and spiritual support to patients at all stages of the cancer continuum, from diagnosis to end-of-life care. Although few spiritually-based interventions have been tested in cancer patients, available data suggest that they may improve quality of life and physical recovery. ^39-41 Future studies should focus on understanding the biopsychosocial mechanisms of this relationship as well as rigorous evaluation of additional interventions in cancer patients to address R/S needs, enhance resilience, and promote better physical health. Such interventions have the potential to enhance quality of life after cancer diagnosis, regardless of disease state.

Precis.

To synthesize existing literature on the relationship between religion/spirituality and patient-reported physical health in cancer, a meta-analysis was conducted of 491 effect sizes from 101 unique samples comprising 32,000 adult cancer patients. Results indicated that religion/spirituality was significantly related to better physical health outcomes.