Cognitive‐behavioral and mindfulness‐based interventions for distress in patients with advanced cancer: A meta‐analysis

Various psychosocial interventions have been developed to reduce distress and improve quality of life (QoL) in patients with advanced cancer, many of which are traditional cognitive‐behavioral interventions (CBIs) or mindfulness‐based interventions (MBIs). The aims of this meta‐analysis were to determine and compare the overall effects of traditional CBIs and MBIs on distress and QoL in this population and to explore potential moderators of intervention efficacy.

Patients with advanced cancer experience many illness-related stressors that may heighten distress.These stressors include increased symptoms and functional decline, 1 growing dependence on caregivers, 2 complex medical decisions, 3 and the ultimate prospect of death. 4Other stressors include loss of work and social roles 5 and financial strain. 6Across the disease trajectory, nearly 30% of patients with advanced cancer have a diagnosable mood disorder. 7Higher distress in this population has been associated with reduced adherence to cancer treatment 8 as well as more frequent and longer hospitalizations. 9Higher distress has also been correlated with reduced quality of life (QoL) 10 and a desire for hastened death in palliative care settings. 11In cancer research, QoL has been defined as patients' appraisal and satisfaction with their functioning in multiple domains, including physical, emotional, social, and role functioning. 12rious psychosocial interventions have been developed to reduce distress and improve QoL in patients with advanced cancer, many of which are traditional cognitive-behavioral interventions (CBIs) or mindfulness-based interventions (MBIs). 13,14These interventions are grounded in different philosophies.Traditional CBIs are structured, present-time focused psychotherapies that are thought to reduce distress through restructuring maladaptive cognitions, increasing pleasurable activities, and enhancing problem solving. 15While traditional CBIs have been adapted for trials in advanced cancer, 16 some theorists have argued that this change-oriented approach may not be appropriate for distressing cognitions and emotions that are reasonable responses to cancer. 17,18I theorists believe that attempts to control one's experiences, cognitions, and emotions are central to the development of distress and posit mindful acceptance as an alternative to control attempts. 17,18Thus, MBIs are an alternative to traditional CBIs that may be a better fit for patients with advanced cancer, as MBIs view distress as normative and instead target the impact of difficult cognitions on behaviors and QoL. 17 Mindfulness, a key mechanism underlying MBIs' effects, has been commonly defined as awareness of the present moment with an attitude of non-judgment and acceptance. 19According to Monitor and Acceptance Theory, 20 enhanced awareness of the present moment without acceptance may increase negative affectivity; thus, modifying one's relationship to present-moment experiences through acceptance is necessary to reduce distress. 20[25] Randomized controlled trials (RCTs) of MBIs in advanced cancer have primarily examined intervention effects on distress and global QoL, with few trials including assessments of fatigue, pain, or other physical symptoms. 26ly three non-inferiority RCTs comparing traditional CBI and MBI have been conducted with cancer populations. 24,27,28One pilot and one larger RCT found that CBIs and MBIs were similarly efficacious in reducing distress in patients with primarily non-metastatic cancers. 27,28To date, only one RCT has compared an MBI to a CBI in patients with advanced cancer. 24This pilot trial found that patients in the ACT condition showed greater decreases in distress and better QoL than those in the CBI condition.
Prior reviews have not quantified and compared the impact of CBIs and MBIs on distress and QoL outcomes in advanced cancer populations.Most reviews of psychosocial interventions in advanced cancer are systematic or narrative reviews without meta-analyses. 29me reviews focus on specific intervention types, such as ACT 26 or meaning-centered interventions, 30 or specific populations, such as patients with metastatic breast cancer, 31 parents with end-stage cancer, 32 or palliative care populations more broadly. 33Other reviews focus on a specific outcome, such as depressive symptoms 34 or death anxiety, 35   (2) only cancer-specific distress outcomes reported (e.g., cancer-related posttraumatic stress); (3) no effect sizes reported or insufficient information provided to calculate an effect size; and (4) insufficient information provided to determine intervention eligibility.If a study provided insufficient information, the corresponding author was contacted for the missing information.
Regarding studies with overlapping samples, the article with the largest sample that included distress and/or QoL outcomes was included in the analyses.

| Data extraction
Variables were extracted and coded based on Lipsey and Wilson's (2001) example codebook. 40The first author (EK) performed all data extraction and the second author (ES) assessed the extracted data for 25% of the records.Extracted data included characteristics of records (first author, year, country of publication), participants (e.g., cancer type, race, mean age, % female, education level), studies (e.g., sample size, participant retention, distress eligibility criterion, control condition), and interventions (e.g., modality, duration, session length, individual vs. other, facilitator, CBI/MBI components).A priori decisions were made regarding the extraction of three-arm intervention trial data to reflect maximum differences between intervention and control conditions.Passive control data (e.g., usual care) were extracted rather than active control data due to potential overlap with CBI and MBI conditions (e.g., social support, topics discussed).
Additionally, data from a more intensive intervention arm (e.g., inperson over technology-based) were extracted, and MBI data were extracted rather than CBI data due to the smaller number of published MBIs.
To assess meta-analytic effects of traditional CBIs and MBIs on distress and QoL, the means and standard deviations for anxiety symptoms, depressive symptoms, general psychological distress, and QoL assessed at baseline and at the first post-intervention assessment were extracted.If complete case means and standard deviations were not available, intention-to-treat data or other reported statistical values were extracted to estimate the effect size (e.g., Fstatistic).Sensitivity analyses comparing the use of differing data types on meta-analytic results were non-significant.

| Quality assessment
Study quality was examined using Cochrane's Risk of Bias 2 Tool 41 ; studies were classified as having low risk, some concerns, or high risk under five domains of possible bias.The first author (EK) assessed study quality for all included records, and the second author (ES) assessed the study quality of 25% of the records.The inter-rater agreement for study quality was 100%.Figure 2 displays the risk of bias ratings for the 37 included studies.Overall, 6 RCTs had a low risk of bias, 25 had some concerns, and 6 had high risk of bias.Main concerns for bias were related to selection of the reported result (k = 21), missing outcome data (k = 19), deviations from the intended interventions (k = 5), and the randomization process (k = 4).All RCTs had some concerns in outcome assessment due to participants' knowledge of the intervention received (which is unavoidable for CBIs and MBIs).RCTs without an active control group were considered to be at higher risk of bias (k = 23).

| Statistical analyses
Data analyses were conducted using Comprehensive Meta-Analysis (Version 3.0) software. 42Meta-analyses were conducted for the overall and individual effects of traditional CBIs and MBIs relative to controls on distress and QoL outcomes at the first time point postintervention (aim 1) using random-effects models to account for the heterogeneity in intervention designs. 43Standardized mean differences were calculated correcting for small sample sizes (Hedges's g) 44 with 95% confidence intervals (CIs).Anxiety and depressive symptoms are highly correlated in advanced cancer. 45,46Additionally, some distress measures only provide information about general distress without differentiating between anxiety and depressive symptoms.Thus, measures of (1) anxiety symptoms, (2) depressive symptoms, and (3) general distress were conceptualized as one distress outcome to maximize the number of studies included in the meta-analysis.If multiple measures of a single outcome (e.g., anxiety and depressive subscales of the Hospital Anxiety and Depression Scale 47 ) were reported, effect sizes were averaged to obtain one effect size per outcome. 43Prior reviews on patients with cancer have also combined anxiety symptoms, depressive symptoms, and general distress to determine intervention effects. 31,48Supporting Information S2 contains meta-analytic results for separate outcomes of anxiety and depressive symptoms.Hedges's g was considered large around 0.8, medium around 0.5, and small around 0.2. 49Statistical significance was set at p < 0.05.A sensitivity analysis was performed, excluding each study in turn, before calculating the pooled effect size to assess the relative contribution of each study included in the meta-analysis.Furthermore, meta-analyses were also performed with the exclusion of low-quality studies (i.e., at high risk of bias).
Effect size heterogeneity was examined using Cochran's Q and I 2 statistics.A statistically significant Q value (p < 0.05) indicated heterogeneity across studies, 44,50 whereas the I 2 statistics quantified the proportion of variance in observed effects reflecting variance in true effects rather than sampling error. 51,52The effect sizes of traditional CBIs and MBIs were statistically compared using a meta-regression approach (aim 2).Meta-regressions, in which effect sizes are predicted using study-level characteristics, were used to examine moderators independently (aim 3).Moderators included intervention Publication bias was examined using Begg's funnel plots 53 with trim-and-fill adjustment, 54 and Egger's test for funnel plot asymmetry 55 was conducted if k ≥ 17 according to suggested guidelines. 43senthal's fail-safe N was calculated and compared to a suggested criterion (5k þ 10). 56I G U R E 2 Risk of bias.KRUEGER ET AL.

| Study and participant characteristics
The electronic database search identified 4150 records.After excluding duplicates, 2889 records were extracted for title and abstract screening.Of these records, 229 were extracted for full-text screening.Among the full texts screened, 193 failed to meet the inclusion criteria.Overall, 36 records reporting on 37 RCTs were included in analyses, all of which were published, full-text journal articles (see Figure 1).
Studies were published between 1994 and 2022.Twenty-two of the 37 studies were conducted in the US, and 7 studies were conducted in Asia.The most common cancer diagnoses were mixed samples (k = 13) and breast (k = 10).Supporting Information S3 and S4 show study and intervention characteristics of the included studies, respectively.One intervention containing both CBI and MBI techniques was categorized as a MBI due to its ACT-based structure and emphasis throughout the intervention. 57  There was also a significant, minimal overall intervention effect on QoL post-intervention compared to controls (Hedges's g = 0.15, 95% CI [0.05, 0.25]).Similarly, a significant, minimal effect on QoL was found for traditional CBIs (Hedges's g = 0.13, 95% CI [0.02, 0.24]).However, MBIs produced a small, though non-significant, effect on QoL post-intervention (Hedges's g = 0.25, 95% CI [−0.00, 0.51]) that was numerically larger than the effect of CBIs.

| Sensitivity analyses
Four CBIs and 2 MBIs were found to be at high risk of bias.When excluded from analyses, small reductions in distress were found for all intervention types.However, while the overall intervention effect (Hedges's g = 0. -7 of 14

| Heterogeneity and moderation analyses
There was evidence of significant, medium heterogeneity between all included studies and MBIs for distress but not QoL (see Table 2).

Regarding aim 2, meta-regression results indicated that CBIs and
MBIs did not significantly differ in their impact on distress (b = 0.09, SE = 0.13, p = 0.50) or QoL (b = 0.09, SE = 0.12, p = 0.45) (see Table 3).However, these results should be interpreted with caution, as there was a limited number of studies in the meta-regression analyses, and lower statistical power likely limited the ability to detect significant differences between intervention types. 60Regarding aim 3, potential moderators were explored only for the combined, overall effects of CBIs and MBIs on distress or QoL (see Table 3

| Publication bias
Trim and fill funnel plots, Rosenthal's classic fail-safe N, and Egger's regression tests indicated that publication bias was highly likely for both outcomes when examining overall intervention effects as well as individual effects of CBIs, but not MBIs (see Table 2 and  Results support the use of CBIs and MBIs for patients with advanced cancer, as these interventions produced small improvements in distress compared to control groups.The current metaanalysis found smaller effects of CBIs and MBIs on distress compared to a prior meta-analysis of psychotherapy's effects on distress outcomes in patients with advanced, incurable cancer compared to controls. 34The prior meta-analysis only included studies with a measure of depressive symptoms and examined anxiety and general distress secondarily.Conversely, in the present analyses, measures of anxiety symptoms, depressive symptoms, and general psychological distress were one outcome, and the greater number of studies may have contributed to smaller effects.
Furthermore, the prior meta-analysis included many interventions outside of traditional CBIs and MBIs. 34Most MBIs in the current meta-analysis were pilot studies designed to assess feasibility and acceptability rather than effect sizes.Additionally, few studies in our meta-analysis had a distress criterion for eligibility; thus, there may have been less room for improvement in distress outcomes across studies, leading to a smaller overall effect size.Furthermore, various cognitive and behavioral techniques and types of MBIs were tested across trials, which may have contributed to heterogeneity in effect sizes for distress.It is possible that patients with advanced cancer may respond better to specific intervention components based on their cognitive or physical symptom profile. 9,61sults also indicated that, overall, CBIs and MBIs had a minimal effect on QoL in patients with advanced cancer.A prior meta-analysis of three studies also found a negligible effect of psychotherapy compared to controls on QoL in patients with advanced, incurable cancer. 34This could be due to the overall decline in QoL that patients with advanced cancer experience as their illness progresses. 62rthermore, many QoL measures partially assess physical wellbeing, 58,63 and patients with advanced cancer report increased symptom burden and decreased physical well-being as their illness progresses. 64Thus, overall QoL scores may not have increased due to reductions in physical well-being, despite substantial improvement in other aspects of QoL. 16ile traditional CBIs focus on changing or controlling difficult thoughts and reducing distress, MBIs view distress as normative and instead focus on the impact of these difficult thoughts on behavior and QoL. 17 Thus, MBIs are theoretically more relevant for patients struggling with reasonable, distressing responses to cancerrelated stressors, such as debilitating symptoms, 65,66 role loss, 5 and their mortality. 4However, traditional CBIs and MBIs did not significantly differ in their effects on distress and QoL.Although findings are similar to non-inferiority trials of traditional CBIs and MBIs for patients with non-metastatic or mixed cancer stages, 27,28 our findings contrasted with the singular study comparing an MBI to a traditional CBI in patients with advanced cancer. 24Because this study was a small pilot trial, results should be cautiously interpreted.Albeit statistically non-significant, MBIs' numerically larger effects on distress and QoL in the current meta-analysis suggest a slight advantage of MBIs over CBIs that warrants further evaluation in a large-scale RCT.
According to Monitor and Acceptance Theory, enhanced awareness of one's experiences may actually increase distress; thus, modifying one's relationship to experiences through acceptance may be necessary to reduce distress and improve stress-related outcomes. 20,67While mindfulness techniques include acceptanceoriented instructions, evidence suggests that mindfulness practice typically increases the ability to monitor present experiences before, and at a faster rate than, the acceptance of these experiences. 20In the present study, CBIs' follow-up periods were, on average, twice the length of MBIs' follow-ups.Due to the brief durations of MBIs and short follow-up periods, patients may have been unable to fully cultivate acceptance of cancer-related thoughts and, thus, experienced less improvement in distress and QoL than hypothesized.
Longer term follow-ups may have shown greater impact on distress and QoL in patients assigned to MBIs compared to traditional CBIs.
Furthermore, compared to 24% of included CBIs, 43% of included MBIs had an active control comparator.This could also mask differential intervention effects, as more stringent controls would be expected to yield smaller effect sizes. 68though CBIs and MBIs delivered to individuals had a larger effect on QoL compared to those delivered to dyads and groups, intervention format did not significantly moderate effects on distress.Few meta-analyses have statistically compared effects of different intervention formats in terminally ill and cancer populations.A prior metaanalysis examining effects of psychotherapy for patients with mixed cancer stages also found a significant moderating effect of intervention format on QoL. 489 While an increased number of intervention sessions has been associated with greater reductions in distress, lengthier individual sessions have been associated with smaller reductions in distress outcomes, 33 perhaps due to increased fatigue.Thus, the heterogeneity of intervention session lengths and durations may have contributed to the non-significant effect of intervention dose on outcomes in the present study.Prior meta-analyses have also found that older patient age was associated with reduced intervention effects on distress and QoL. 33,69The restricted range of mean patient ages in the current study may have limited our ability to replicate this finding.

| Study limitations
Limitations of this meta-analysis should be noted.Samples primarily consisted of white middle-aged and older adults in the US.As older age is associated with lower levels of distress in cancer populations, [70][71][72] older patients may be less likely to experience large reductions in distress with psychological intervention.Alternatively, older adults may require modifications to fully benefit from psychosocial interventions, such as the use of simplified homework and mnemonic aids, shortened or stationary mindfulness practices, or a "LifeSkills" approach (e.g., reflecting on how one has navigated previous stressful life experiences). 73,74Additionally, statistical power to detect intervention effects was limited by the small sample sizes in the majority of trials, especially MBI trials.The current study was also subject to publication bias favoring statistically significant results for distress and QoL.Finally, results should be cautiously interpreted due to concerns about study quality, the limited number of studies reporting details for moderation analyses, and the heterogeneity of intervention components and effect sizes.

| Clinical implications
At present, evidence equally supports the use of traditional CBIs and MBIs to reduce distress in patients with advanced cancer.Offering both options to patients allows for selection of the intervention consistent with their needs and preferences.Although intervention effects in this meta-analysis are considered small, the impact of the interventions is likely to be greater in patients with clinically meaningful distress.Furthermore, the absence of sociodemographic moderators of intervention effects suggests that CBIs and MBIs are equally efficacious for both men and women of various ages and levels of education, allowing providers to offer either treatment to a range of patients with various advanced cancers.

| Future research directions
There are several important future directions for research on CBIs and MBIs in advanced cancer.First, large-scale RCTs with longer follow-ups would allow for examination of maintenance or even improvement of effects, as patients would have more time to cultivate and integrate intervention skills into daily life.Given illness progression, allowing for flexibility in scheduling, offering virtual or home visits, and building rapport between research staff and study participants may improve retention for longer follow-ups in this population. 75,76Additionally, booster sessions may be tested to thereby excluding intervention trials targeting other distress outcomes.The current meta-analysis is the first to examine the overall effects of traditional CBIs and MBIs on general distress (i.e., anxiety symptoms, depressive symptoms, and general psychological distress) and QoL in patients with advanced cancer and to statistically compare the effects of these two types of interventions on outcomes.Comparing these interventions informs future research and has direct implications for the clinical care of distressed patients with advanced cancer.The aims of this meta-analysis are: (1) to determine the overall impact of traditional CBIs and MBIs on general distress and QoL postintervention in patients with advanced cancer; (2) to compare the effects of traditional CBIs and MBIs on general distress and QoL postintervention, and (3) to explore potential moderators of intervention efficacy, including both patient and intervention characteristics.

2 | METHODS 2 . 1 |
Literature searchand(2) full-text screening (see Figure1).There was strong agreement between reviewers for title and abstract as well as full-text screening (Cohen's kappas = 0.89 and 0.75, respectively).Conflicts at each stage were resolved during consensus meetings.Study inclusion criteria included: (1) a sample of adults (≥18 years) diagnosed with advanced cancer (as stated by the study authors or ≥70% stage III and/or stage IV); (2) RCT examining a CBI or MBI; (3) outcomes of anxiety symptoms, depressive symptoms, general psychological distress, or QoL assessed at baseline and post-intervention; and (4) available in English.Adapted from prior cancer research, 38 traditional CBIs were defined as interventions including at least one of the following components: cognitive restructuring, imaginal or in vivo exposure, coping skills training, behavioral activation, behavioral experiments, or stress and anxiety management.Relaxation training alone was not considered to be a traditional CBI.MBIs were defined as interventions emphasizing one's relationship to thoughts (e.g., acceptance of thoughts), including mindfulness practices in the majority of sessions or mindfulness homework throughout the intervention.Interventions promoting both CBI and MBI techniques (e.g., cognitive restructuring along with acceptance of other thoughts) or traditional CBIs emphasizing mindfulness in less than half the sessions were considered 'combination' therapies and were included in the analysis of the overall impact of CBIs and MBIs but excluded F I G U R E 1 PRISMA flow diagram.KRUEGER ET AL. when examining and comparing the individual effects of traditional CBIs and MBIs.Study exclusion criteria were: (1) yoga and other interventions focused on physical movement due to overlap with well-documented effects of exercise interventions 39 ;

4 of 14 -
KRUEGER ET AL.format (individual vs. other), total intervention dose (number of sessions � session duration in minutes), control condition (active vs. passive), having a distress eligibility criterion at study entry, participant retention, patient age, gender (% female), and education level (% with at least a bachelor's degree or equivalent).
Thus, 21 studies were characterized as traditional CBIs, 14 studies were characterized as MBIs, and one article reporting on two samples was characterized as a combination therapy.The most commonly reported CBI components were cognitive restructuring, coping skills training, and stress and anxiety management.Half of the MBIs were ACT-based interventions (k = 7).Thirty-three studies reported distress outcomes post-intervention (depressive symptoms, k = 24; anxiety symptoms, k = 17; overall distress, k = 15), whereas 23 studies reported QoL outcomes post-intervention.The most common distress measure was the HADS 47 (k = 13), and the most common QoL measure was the Functional Assessment of Cancer Therapy-General 58 (FACT-G; k = 8).Supporting Information S4 contains the outcome measures used in each study.Across all RCTs, the average sample size was 78 participants (SD = 60.8;range = 16-232; k = 37) with a weighted mean age of 60.4 years (SD = 6.1; range = 50.0-71.4;k = 34).Six CBI and 2 MBI trials had a distress criterion for study entry.One MBI had a QoL criterion for study entry. 59The intervention format varied, including individual (k = 22), dyadic (k = 9), and group interventions (k = 5).Intervention dose varied across studies, ranging from one 60-min and one 30-min session in one intervention to as many as 35 120-min sessions in another.Intervention durations also varied, ranging from 1 to 35 weeks (M = 9; SD = 6).Most interventions included interactions with facilitators (k = 30).The control conditions across studies included treatment as usual (k = 14), attention controls (k = 9), waitlist (k = 7), and enhanced usual care (k = 5).One MBI trial had a CBI comparator.First post-intervention assessments were, on T A B L E 1 Descriptive statistics.
) due to the limited number of studies in intervention subgroups.None of the intervention and patient characteristics moderated intervention effects on distress and QoL post-intervention, with one exception: intervention format significantly moderated the intervention effect on QoL (b = −0.23,SE = 0.10, p = 0.02).Compared to interventions delivered to individual patients, CBIs and MBIs delivered to dyads or groups produced smaller effects on QoL post-intervention.After excluding low quality studies, no significant moderators were found for QoL or distress.
Supporting Information S5).Adjusting the effect size for the missing studies yielded minimal Hedges's gs for both overall intervention effects and CBIs' individual effects on distress and QoL postintervention.4 | DISCUSSION This is the first meta-analysis to examine and statistically compare the efficacy of CBIs and MBIs on distress and QoL in patients with advanced cancer.Compared to controls, CBIs and MBIs produced a small decrease in distress and a minimal increase in QoL postintervention.Furthermore, CBIs and MBIs did not significantly differ in effect sizes, suggesting that patients assigned to either traditional CBIs or MBIs experienced similar improvement in distress and QoL across intervention types.However, MBIs produced numerically larger effects on both distress and QoL outcomes, and lower statistical power likely limited the ability to detect statistically significant differences.Only intervention format (i.e., individual vs. other) significantly moderated the overall effect of CBIs and MBIs on QoL post-intervention, showing a larger effect for individually delivered interventions.No moderators of the overall intervention effect on distress were found.
determine their impact on intervention effects.Second, the effects of traditional CBIs and MBIs on distress and QoL warrant comparison in a large-scale RCT to further examine the potential, slight advantage of MBIs over CBIs found in this meta-analysis.Third, future trials should examine theory-driven mechanisms underlying CBIs' and MBIs' effects on distress and QoL to inform intervention refinement to maximize efficacy.For example, future CBI trials could measure changes in the content of thoughts as a mediator of intervention effects, whereas MBI trials could measure change in mindfulness as a mediator of intervention effects.Fourth, to inform precision mental health approaches, researchers should determine which intervention components are most effective in reducing distress outcomes and patient characteristics that moderate these effects.Focusing on patients with clinically meaningful distress and enrolling patients from various ethnocultural groups would ensure generalizability of findings to diverse patients with greater need for intervention.Lastly, the majority of included studies had some concerns for bias.Researchers should promote rigor and transparency by reporting trial methodology and results in accordance with the CONSORT (Consolidated Standards of Reporting Trials) 2010 guidelines 77 and registering trials before or at the onset of participant enrollment, consistent with the ICMJE (International Committee of Medical Journal Editors)policy.78 5 | CONCLUSIONSIn conclusion, results of this meta-analysis show that traditional CBIs and MBIs are both efficacious in producing small reductions in distress in patients with advanced cancer, although effects on QoL appear minimal given patient disease status.While MBIs may theoretically be a better fit than CBIs for patients experiencing normative distress related to advanced cancer, large-scale trials are needed to confirm this hypothesis.Interventions targeting patients individually, rather than in a dyadic or group format, may have a greater impact on QoL in advanced cancer.Further intervention comparisons and refinement are needed to optimize psychological aspects of advanced cancer care.
Note: Not all data were available for each study.Abbreviations: CBIs, cognitive-behavioral interventions; MBIs, mindfulness-based interventions.a Percentage of patients with at least a bachelor's degree or equivalent.b Sessions � minutes.c Weeks post-baseline.average3.2| Moderation analyses.
T A B L E 3 a Sessions � minutes.