Moderators of psycho-oncology therapy effectiveness: meta-analysis of therapy characteristics.

This article was downloaded by: [University Of Pittsburgh] On: 13 November 2014, At: 19:35 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Psychosocial Oncology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/wjpo20

Moderators of Psycho-Oncology Therapy Effectiveness: Meta-Analysis of Therapy Characteristics a

Heather A. Heron-Speirs LLB (Hons), MA (Psych) (Hons) , Shane T. a

Harvey BSocSci, MSocSci, PGDipClinPsych, PhD & Donald M. Baken BA (Hons), PGDipClinPsych, PhD

a

a

School of Psychology, Massey University , Palmerston North , New Zealand Accepted author version posted online: 25 Aug 2013.Published online: 31 Oct 2013.

To cite this article: Heather A. Heron-Speirs LLB (Hons), MA (Psych) (Hons) , Shane T. Harvey BSocSci, MSocSci, PGDipClinPsych, PhD & Donald M. Baken BA (Hons), PGDipClinPsych, PhD (2013) Moderators of Psycho-Oncology Therapy Effectiveness: Meta-Analysis of Therapy Characteristics, Journal of Psychosocial Oncology, 31:6, 617-641, DOI: 10.1080/07347332.2013.835022 To link to this article: http://dx.doi.org/10.1080/07347332.2013.835022

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &

Downloaded by [University Of Pittsburgh] at 19:36 13 November 2014

Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions

Journal of Psychosocial Oncology, 31:617–641, 2013 Copyright © Taylor & Francis Group, LLC ISSN: 0734-7332 print / 1540-7586 online DOI: 10.1080/07347332.2013.835022


Moderators of Psycho-Oncology Therapy Effectiveness: Meta-Analysis of Therapy Characteristics HEATHER A. HERON-SPEIRS, LLB (Hons), MA (Psych) (Hons), SHANE T. HARVEY, BSocSci, MSocSci, PGDipClinPsych, PhD, and DONALD M. BAKEN, BA (Hons), PGDipClinPsych, PhD School of Psychology, Massey University, Palmerston North, New Zealand

As part of a larger meta-analysis seeking moderators of the effectiveness of psycho-oncological interventions, this report focuses on intervention types and characteristics, including protocol components, means of delivery (mode, dose, and therapist variables), and mechanisms of effectiveness. The data set comprised 146 published and unpublished prospective controlled trials with outcomes of anxiety, depression, and distress. Analyses took into account two moderators from analysis of study design features. The authors conclude that each of the four main professional therapy types (education, relaxation, cognitive behavior therapy (CBT), and expressive-support) has effect and that it is more important to focus on participant variables, notably, elevated baseline distress. Therapy components delivered by nonprofessionals and interventions that affect the patient indirectly show potential. Recommendations for practice and research are made. KEYWORDS psycho-oncology, cancer, meta-analysis, patient, therapy, distress The search for moderators that predict better effect size for psycho-oncology intervention has produced few firm results to date. The field is enormously

The authors are grateful for funding from the Cancer Society of New Zealand and for the generous assistance of the many primary study researchers who responded to their requests for unpublished data. Address correspondence to Heather Heron-Speirs, LLB (Hons), MA (Psych) (Hons), c/Research Coordinator, Cancer Psychology Service, School of Psychology, Massey University, PO Box 11-222, Palmerston North 4442, New Zealand. E-mail: [email protected] 617


618

H. A. Heron-Speirs et al.

complex, layering a great range of patient medical characteristics and research approaches upon a full range of psychosocial intervention types and patient sociodemographics. Meta-analyses have produced an array of effect sizes, shedding little reliable light on issues relevant to clinicians, administrators, and researchers, and raising the likelihood that unrecognized variables from the complexity of the field are at play. This meta-analysis attempts to bring some order to the field, drawing into visibility some of those variables and taking them into account before investigating the efficacy of various intervention types. What could such variables be? We took the preliminary step of analyzing the impact of a range of primary study design features before undertaking the analysis of variables of substantive interest (Lipsey & Wilson, 2001) to establish whether the design features could be responsible for confounding and confusing substantive effects. If particular study design features produce heterogeneity in the data set, then the fact that different meta-analyses use different primary study sampling criteria could explain the inconsistencies in results. We conducted an effortful literature search and included much unpublished data and data from nonrandomized studies to build a data set of 146 prospective controlled trials dated 1975 to June 2008 where outcomes were anxiety, depression, and general distress. Our search domain and efforts, and our preliminary analyses, are detailed elsewhere (Heron-Speirs, Harvey, & Baken, 2012). That article also details coding and analysis issues and our terminology and explains our approach to the question of multiple comparisons (inflating alpha) and our decision to note significance levels greater than p < 0.05. It is important to read that article to understand this one. Our preliminary analyses of study design features found that, indeed, two features produced higher effects: studies recruiting patients with elevated baseline distress and/or with no history of distress, and studies with an untreated control group. Neither of these factors has been taken into account by previous meta-analyses before undertaking analyses of patient or therapy characteristics. We also found that the exclusion of nonrandomized studies was not empirically justified, and that the exclusion of studies for lack of blindness was inappropriate. This means that many meta-analyses in the field have excluded much valuable data from their literature search domains. Indeed, the data from more naturalistic designs may be the most valuable, because these designs can sometimes source participants who are not available for randomization for ethical or practical reasons in this doubly clinical field. Such patients may be the very ones most in need of therapeutic intervention and most likely to produce effect sizes that are relevant for accurately informing clinicians and administrators. Understanding the empirical dynamics of the design features of studies in our data set, we were able to proceed to substantive analyses on a firm foundation. We accounted for the two dichotomous study design features found to moderate effect sizes by stratifying our dataset according to a


Psycho-Oncology Meta-Analysis Therapy Characteristics

619

2 × 2 matrix. Analyzing each of the four strata separately, we examined the impact of a range of sociodemographic, psychological, and medical variables, as detailed in our second report, Heron-Speirs, Harvey, and Baken (2013). Consistent with our earlier finding that elevated baseline distress as a study design feature (i.e., recruitment procedure) predicted higher effect size, and consistent with hints in earlier meta-analyses (Meyer & Mark, 1995; Sheard & Maguire, 1999), and the finding of a recent meta-regression (Schneider et al., 2010), we found evidence of the importance of elevated baseline distress as a psychological variable that moderates therapy outcome. Further, there was evidence (statistically significant or notable trends) of higher effect size predicted by patient variables that could be expected to associate with higher cancer-related distress at baseline: people who are older, single, of lower income, male, or suffering a cancer other than early stage breast cancer (distinctive for its good prognosis). The finding that male gender predicts higher effect size agreed with an earlier finding by Rehse and Pukrop (2003). The finding in relation to early-stage breast cancer was consistent with a recent finding by Naaman, Radwan, Fergusson, and Johnson (2009) that metastatic breast cancer patients received benefit of moderate magnitude from therapy whereas early stage breast cancer patients received negligible benefit, which they speculated was due to lack of distress at baseline due to good prognosis. We suggested that all of the patient characteristics that we identified might be associated with elevated distress, for various reasons interacting with the cancer context, so that it is likely that baseline distress is the key patient variable. Thus the layers of complexity are beginning to clarify. Now we proceed to report our final set of analyses: an investigation into the effectiveness of various intervention types, including specific therapy components and delivery variables. Therapy type as a potential moderator of intervention effectiveness has received considerable meta-analytic attention. We have attempted to summarize effect size findings relative to the different therapies and populations that have been analyzed in Table 1. Direct comparison is problematic because categorizations and patient characteristics vary, but we note these points: First, all of the therapy types investigated by Devine and Westlake (1995) produced moderate to large effects (Cohen’s ds of 0.40 – 0.74) sustaining their conclusion that there was no statistically significant difference in the performance of different therapies, and therefore clinicians could choose between a wide range to suit individual circumstances. This is saying that therapy type is not an important moderator of intervention effectiveness. Second, there is a wide range between meta-analyses in the results produced by some therapy types. For example, cognitive behavior therapy (CBT) scores a high Hedges g of 0.81 in the analysis by Cwikel, Behar, and Rabson-Hare (2000), extraordinarily high gs of 1.99 and 1.20 in Osborn, Demoncada, and Feuerstein (2006), a negligible magnitude d of 0.13 in Tatrow and Montgomery (2006), and widely disparate ds of 0.11 and 0.85 in Naaman et al. (2009).

620


TABLE 1 Earlier Meta-Analyses, Summary of Therapy Type Effect Sizes Authors


Devine & Westlake (1995)

Sheard & Maguire (1999) Tatrow & Montgomery (2006) Zabalegui, Sanchez, Sanchez, & Juando (2005) Luebbert, Dahme, & Hasenbring (2001) Osborn, Demoncada, & Feuerstein (2006) Naaman, Radwan, Fergusson, & Johnson (2009)

Cwikel, Behar, & Rabson-Hare (2000) Rehse & Pukrop (2003)

Therapy Type

Anxiety

Depression

Distress

Education only

d = .74

d = .50

—

Muscle relaxation only Other relaxation / distraction only Muscle relaxation with guided imagery Multiple behavioral interventions with relaxation Education with other behavioral treatments Education with behavioral or non-behavioral counseling Expressive-supportive counseling Group psycho-education Relaxation CBT for breast cancer patients Social support groups

d = .60 d = .66

d = .40 —

— —

d = .62

d = .40

—

d = .59

—

—

d = .46

—

—

d = .52

—

—

—

d = .66

—

g = 1.59

g = .94

—

g = .21 —

g = .03 —

— d = .13

z score = .71

z score = .63

—

Relaxation training

d = .45

d = .54

—

CBT for survivors

g = 1.99

g = 1.20

—

Psychoeducation for breast cancer patients

d = .02

d = .45

—

Guided imagery and relaxation for breast cancer patients CBT for breast cancer patients Expressive-supportive therapies for breast cancer patients CBT All other treatments Educational treatments Social support Coping skills training (CBT) Psychotherapy

d = .40

d = .55

—

d = .11

d = .85

—

d = .43

d = 1.80

—

CBT = cognitive behavior therapy

‘psychological outcome’ g = .81 ‘psychological outcome’ g = .49 ‘emotional adjustment’ d = .96 ‘emotional adjustment’ d = .58 ‘emotional adjustment’ d = .48 ‘emotional adjustment’ d = .58



621

The implication is that unrecognized moderators are confounding results, as described above. Third, expressive-supportive therapies show potential with results that are consistently of moderate magnitude or better. The dose and means of delivery of a therapy might be expected to influence the benefit received. Primary studies commonly experiment with different delivery packages, comparing brief interventions with lengthy programs, simple brochures with multimedia educational packages, delivery by nurses or social workers with delivery by psychologists, and so on. For ethical reasons, or to address placebo effects, many studies provide some kind of attenuated treatment for the control group. However, there is very little in the psycho-oncology synthesis literature about delivery factors. Writing about group interventions, the review by Sherman et al. (2004) notes a tendency for long-term interventions to be provided for patients with advanced disease. The meta-analysis by Rehse and Pukrop (2003) found a greater effect for interventions lasting more than 12 weeks. The meta-analysis by Cwikel and Behar (1999) found that interventions delivered by social workers performed poorly. And the meta-analysis by Naaman et al. (2009), which drew only from breast cancer samples, found that interventions of more or less than 20 hours were effective, and therefore recommended shorter protocols. However, given the unknown impact of study design features and patient characteristics—notably baseline distress—in these studies, conclusions are of questionable validity. The present meta-analysis is designed to take into account the two study design characteristics found to moderate results, one of which is whether participants were screened at baseline for distress.

METHOD A 2 × 2 matrix of the study design features that we found moderated effect size (nature of the control group, i.e., “treated” or “untreated,” and whether or not recruits were selected for participation in the study after establishing distress from baseline screening) was used to stratify the data set in preparation for our substantive analyses (see Heron-Speirs et al., 2012, p. 58, Figure 1). “Treated” control groups included those that received treatment as usual, an attention placebo, or a treatment element placebo, whereas “untreated” control groups received nothing or were on wait list for the treatment. “Screening” included either or both (“simultaneous screening”) of “screening in” (screening potential recruits and accepting only those with elevated baseline distress) and “screening out” (screening potential recruits and rejecting those with a history of distress—note the distinction from criteria excluding patients with cognitive impairment or current psychotic illness or suicide risk). We found stronger effects from studies that used an untreated control group (i.e., the contrast between results from the control and the treated group was maximized), and from studies that screened at baseline.


622


Thus the studies that produced the strongest effect sizes screened and used untreated controls, whereas those that produced the weakest effect sizes did not screen at baseline and used a treated control. Where high-frequency data were available, analyses of patient characteristics were run only on untreated control studies (data from two quadrants of the matrix) thereby maximizing the variance available for the identification of substantive moderators. Data from the three outcomes (anxiety, depression, and distress) were combined (referred to as “omnibus outcome”) in small n analyses to increase power. Studies were coded according to therapy type and delivery features (mode, dose, and therapist characteristics). Therapy types were organized into four superordinate professional categories that were designed by content (rather than any delivery feature) and drawn from earlier works (Barsevick, Sweeney, & Haney, 2002; Cwikel & Behar, 1999; Meyer & Mark, 1995). They were: education/information (education); relaxationfocused behavioral therapies (relaxation); cognitive-behavioral therapies (CBT); and expressive-supportive or nondirective professional counseling/psychotherapy (expressive-support). Three additional small categories were: nonprofessionally led support or counseling (nonprofessional); ‘indirect’ interventions (those that deliver benefit to the patient by intervening with someone else—a medical professional or significant other—rather than the patient); and ‘other,’ largely comprising written emotional expression studies. These categories are further detailed in the appendix. Assessment time points were categorized as “early times” or “late times.” Early times was most frequently the earliest post intervention assessment before 6 months (alternatively, the assessment 3 to 6 months into openended/lengthy therapies, or the pre-medical-treatment assessment where assessment points were anchored on medical treatment progress); Late times was most frequently the earliest post intervention assessment after 6 months (alternatively, 6 to 12 months into open-ended/lengthy therapies, or the post-medical treatment assessment). A large proportion of therapy protocols in psycho-oncology are eclectic, including components from more than one of the super-ordinate types categorized above (e.g., a protocol may have components from education and CBT). Each study was coded under all of the therapy types that characterized them, and all of the specific therapy components. This meant that a given study could contribute outcome data to analyses under more than one therapy type or component. It is not appropriate to directly compare such statistically dependent data, and we also wished to preserve sample size, so these analyses were run independently of each other and the effect sizes tabled, rather than run simultaneously for direct comparison. Although we thought this the best approach given the nature and amount of data available, a number of cautions apply when considering these results: (1) there was no direct statistical comparison for heterogeneity and (2) effect sizes


623


are exposed to random distortion caused by other components of therapy packaged with the one of interest. Larger n in a given cell provides some protection against this, but because some therapy components are regularly coupled with others, an element of systematic variation can also confound conclusions. (3) Screened samples are exposed to uneven distributions of studies that screened in (higher effect sizes than unscreened studies) or screened out (higher than unscreened but not as high as screened in) or simultaneously screened both ways (highest). It is therefore safest to focus on consistent patterns across screening types or outcomes rather than on individual cell results. However, some analyses broke out the studies that screened out, minimizing the potential for distortion further.

RESULTS Therapy Types PROFESSIONAL THERAPIES Analyses of the four main professional therapy types—education, relaxation, CBT, and expressive-support—were run separately, as noted above. Results are summarized in Table 2. It is important to note the categorization definitions specific to this analysis (screened out studies are broken out) and the effect size qualifiers in the note to the table. It can be seen that education and relaxation most effectively treated anxiety, although relaxation produced a surprising high result against depression for patients who were distressed at baseline (from only two studies, one of which screened both ways simultaneously). Lower confidence intervals from education studies regularly fell below zero (not shown on table), indicating that some patients are made more distressed by receiving more information, and for a lesser number that effect may endure into late times. For CBT, the large effect sizes for all outcomes from studies that screened for baseline distress actually derived from studies that simultaneously screened both ways, with distress screening alone producing only small to moderate effects. Some data suggested that a small effect from CBT endured into late times (beyond 6 months). Very strong early times effects were produced by expressive-supportive therapies against all three outcomes for patients screened in for baseline distress (and not simultaneously screened out: anxiety, g = 1.13; depression, 1.03; distress, 1.09; each from n of 2, p < 0.05) which caused the small to moderate magnitude CBT results for patients who had the same screening to pale by comparison. However, these results are supported by only two studies each and late times data were not available. The role of baseline distress screening in elevating effect size is seen repeatedly in relation to all three of the therapy types for which data is available—relaxation, CBT, and expressive-support. Noncomplex patients

624

0

0a 0

— —

a

d

—

— —

Anx

0a 0 0

a

— —

—

— — 0a —

Dep

0a 0 0

— —

— — 0 0a 0

Dis

a

a

0 c

c

ac

ac

—

0

— 0 0

a

0a —

0

a

a

— 0 —

a

Dep

—

Anx

Relaxation

— 0 0

0a —

0

a

—

—

a

Dis

0 0

a

— —

a

0a 0

a

0

—

a

— —

0 0

a

b a

a

Dep

b

Anx

CBT

— 0 0

a

— —

— 0 0

a

b

Dis

0

a

— —

0a

a

—

a

Anx

— 0 0

a

— —

a

a

— 0 —

a

Dep

— — — — 0 0

a

a

— 0 0

a

Dis

Expressive- Supportive

There are two results in each cell. The top result relates to early times; the lower to late times. CBT = Cognitive behavior therapy. Outcomes: Anx = anxiety; Dep = depression; Dis = distress. Patient groups: DisB = distressed at baseline, i.e., “screened in” and “screened in and out” categories combined; Non comp = psychologically non-complex cases, i.e., “screened out” for history of psychological distress. Unsc = received neither type of screening. Effect sizes: 0 = negative, zero, or negligible range (Hedges g < 0.15); = small range (g = .15 – .35); = moderate range (g = .36 – .65); = large range effect size (g > 0.66). Effect size qualifiers: a = result is larger than “negligible” (“0”) and is non-significant (p < 0.10) or is supported by n of only one or two studies. b These results are based on even, or near even, small frequencies of studies that screened in and that screened in and out, yielding a mean effect size that is strong. However, the studies that did not simultaneously screen out produced only a moderate 0.53 (p < 0.05, n = 2) against anxiety, a small 0.30 (p < 0.10, n = 3) against depression, and a moderate 0.45 (p < 0.05, n = 3) against general distress. c These results are influenced by a disproportionately high number of studies that combine relaxation with other therapies, probably resulting in the attenuation of early and the exaggeration of late, times effects. d Whilst this effect size is moderate, the “pure” education result is strong.

Unsc

Non comp

Treated control DisB

Unsc

Non comp

Untreated control DisB

Patient group

Education/Information

TABLE 2 Therapy Types, Summary



625

(those in studies that screened out potential recruits with distress history) tended to produce effects in the small range, although moderate magnitude early times effects were produced by relaxation against anxiety and general distress. Unscreened samples generally produced null, negligible or small effects. Notice the row of zeros against treated controls, showing no effect relative to placebo for this sample.


SPECIFIC THERAPY COMPONENTS More than 50 specific therapy components within the four professional therapy types were coded. Analyses were conducted by psychological outcome and results are displayed in Table 3, but the cautions specified earlier apply. Results were tabled only for those therapy components with a frequency of four or more studies in at least one cell, and using data from only those studies with untreated controls. Attention is drawn to the following results: For education, components that informed about coping strategies produced some small but significant results in relation to anxiety and general distress outcomes, for screened patients; For relaxation, progressive muscle relaxation (PMR) and unspecified training (probably PMR in most cases), medium magnitude significant effect sizes in relation to anxiety and distress for screened patients; and diaphragmatic breathing, medium in relation to distress for screened patients with a medium nonsignificant result in relation to anxiety; For CBT, cognitive restructuring, challenging negative thoughts, self-monitoring, problem identification, problem solving, pleasant activity scheduling, and assertiveness training produced significant results in relation to a range of outcomes but mostly for screened patients; For expressive-supportive therapies, existentially oriented therapies and those more generally oriented toward psychosocial and physical issues produced significant results of varying magnitude. Although existentially oriented therapy calculations mostly had small n, effect sizes were promising. DELIVERY VARIABLES A large range of variables relating to therapy delivery mode, dose, and the qualities of therapists were analyzed using only studies with untreated controls, once again. Direct comparison was possible for these analyses (independent data), but little heterogeneity was seen (Table 4). However, no or one-off homework produced better results for unscreened patients, with the exception noted in relation to CBT and relaxation where, for screened patients (though not unscreened) skill practice is necessary and a trend in favor of homework was shown (with homework, g = .72, p < 0.05, n = 11; without, .41, n = 4, p < 0.10; Q statistic p = .229). Trends favored disciplines of a more psychological nature for unscreened samples, and professional and

626

Component 0.78 (4) 0.16 (8)∗ 0.55 (3) 0.17 (6)† 0.09 (9) 0.16 (4) — 0.10 (5) — 0.13 (8) 0.09 (4) 0.17 (7)† 0.05 (3) − 0.06 (1) 0.08 (2) 0.16 (5)† 0.16 (5)† — 0.14 (5) 0.08 (2) 0.38 (1)∗ 0.23 (2)† 0.18 (6) 0.31 (6)∗

Unscr

Anxiety

0.30 (2) 0.31 (6)∗ — 0.24 (1) 0.54 (8)∗ 0.58 (4)∗ — 0.17 (4) 0.06 (3) 0.48 (4) 0.56 (5)† 0.55 (6)∗ — 0.85 (2) 0.50 (1)∗ 0.41 (8)∗ 0.29 (3) 0.39 (4)∗ 0.22 (5) 0.14 (2) − 0.06 (2) — 0.87 (2) 0.47 (6)∗

Scr

Unscr 0.20 (1) 0.14 (3) — − 0.01 (2) 0.02 (6) 0.09 (2) — 0.01 (4) — 0.02 (3) 0.13 (4) 0.02 (3) − 0.01 (1) 0.01 (1) — 0.13 (4) 0.19 (4) — 0.40 (3)† 0.01 (1) — — 0.51 (2)∗ 0.27 (3)

− 0.26 (1) 0.14 (7) — 0.16 (1) 0.08 (5) 0.46 (3) 0.47 (1) −0.16 (4) − 0.15 (2) − 0.40 (2) 0.42 (8)∗ 0.28 (7) — 0.55 (3) 0.85 (3)† 0.41 (11)∗ 0.01 (2) 0.11 (4) 0.08 (5) 0.23 (2) − 0.17 (2) — 0.69 (2) 0.41 (5)

Depression Scr

Unscr 0.28 (6) 0.20 (5) 0.07 (5) 0.11 (5) 0.06 (6) 0.11 (2) — 0.09 (4) — 0.16 (5) −0.08 (4) − 0.05 (2) 0.17 (1) − 0.11 (1) −0.04 (3) 0.02 (6) 0.10 (2) − 0.14 (1) −0.04 (5) 0.11 (2) − 0.08 (2) 0.13 (4) 0.32 (3) 0.43 (4)

Distress

0.03 (1) 0.26 (7)∗ 0.25 (2) 0.40 (1) 0.51 (8)∗ 0.10 (3) 0.67 (1)† 0.22 (5) − 0.09 (1) 0.63 (4)† 0.31 (7)∗ 0.36 (7)∗ − 0.25 (1) 0.39 (4)∗ 0.93 (3)∗ 0.46 (10)∗ − 0.09 (1) 0.28 (3) 0.11 (5) 0.23 (2) −0.12 (3) — 0.89 (2)∗ 0.53 (7)∗

Scr

CBT = Cognitive behavior therapy; Scr = Screened; Unscr = Unscreened. Hedges g effect size point estimate (number of studies contributing to calculation) presented, using “‘windsorized, ‘early time’, untreated” control data. Studies for which status on screening, nature of control, or the variable of substantive interest was unclear, were excluded. † = statistically significant at p < 0.10, ∗ = statistically significant at p < 0.05; two tailed. Results supported by n < 2 shaded. Hence, the strong results are both unshaded and statistically significant

Education re cancer or cancer treatments Education re coping strategies Education re managing side effects Relaxation Stress management education / training Progressive muscle relaxation Relaxation education or training (unspecified type) Cue-controlled relaxation training (Guided) imagery Meditation Diaphragmatic breathing CBT Cognitive restructuring/reappraisal Challenging negative thoughts Positive self-talk/imagine success Self-monitoring Problem identification Problem solving Goal / expectation setting / plan making Pleasant activity scheduling Assertiveness / communication education or training Establishment or optimisation of use of social networks Body image counseling Sex therapy/education re sexual intimacy/sexuality ExpressiveRegarding existential, spiritual, grief, or death issues Supportive Regarding cancer experience, physical and psychosocial issues

Education

Type

Therapy

TABLE 3 Specific Therapy Components



627

TABLE 4 Therapy Delivery Mode, Dose, and Therapist Variables, Summary Variable and coded levels

Screened

Unscreened


Mode variables Therapy recipient Trend favors individual patient Homogeneity between 1. Individual patients, 2. individual patient, patient over patient plus significant Individual patients plus significant other, or other, or group of patients accompanied by a group of patients suggests significant other, 3. efficiencies are possible. Group of patients, and Indirect interventions show 4. Group of patients promise with their significant others. Significant others were most commonly spouses Delivery technology 1. In person, 2. By Insufficient data for Homogeneity highlights the telephone, 3. By some comparison efficiency and accessibility other interactive of noninteractive technology (personal technology where letter, e-mail or appropriate interactive web site), or 4. By noninteractive technology (i.e., by means that do not involve patient/therapist interaction, including printed material, video or audio recording, noninteractive website, or snoezelen environment) Therapy setting Homogeneity between Homogeneity between 1. Inpatient or residential available categories implies inpatient and outpatient care, 2. Outpatient that self-directed or settings (including hospital or telephone therapy at home hospice outpatient can be of similar value to clinics, professional outpatient therapy premises, and community facilities), 3. The patient’s home, or 4. “Split” i.e., the patient and therapist were in different settings, as for telephone delivery Dose variables Flexibility of number of Homogeneity implies sessions efficiency from a limited 1. Fixed number only, 2. format, but this result may Additional sessions be misleading available according to need, 3. Not applicable (e.g., self-directed)

Homogeneity implies efficiency from a limited format, but this result may be misleading (Continued on next page)

628


TABLE 4 Therapy Delivery Mode, Dose, and Therapist Variables, Summary (Continued) Variable and coded levels


Session hours with a therapist 1. ≤ 4 hours, 2. 5 – 11 hours, 3. ≥ 12 hours, 4. Not applicable (self-directed)

Weeks of therapy sessions 1. 1 – 3 weeks, 2. 4 – 9 weeks, 3. 10 – 22 weeks, 4. Not applicable (e.g., printed or recorded material was provided to patients) Number of therapy sessions 1. 1 – 4 sessions, 2. 5 – 7 sessions, 3. 8 – 13 sessions, 4. Not applicable (noninteractive) Nature of patient’s therapy worka 1. With homework/skill practice, 2. Without same Frequency of homework 1. One-off (e.g., the reading of a literature pack provided at the first or only session), 2, Regularly expected (e.g., with sessions), 3. Irregularly expected, 4. Optional, and, 5. None at all

Screened

Unscreened

Homogeneity found at immediate/short term but probably due to confounding by cancer site, or could result from appropriate hours being allocated for different therapy types. Effect confounded at medium term.

Homogeneity implies efficiencies could be obtained from lower doses or self-directed therapies, but possible confounding by cancer site was not investigated except for self-directed therapies. Effect evaporates by medium term

Weak immediate/short term Same result as for Session trend in favor of more hours with a therapist weeks. Effect confounded at medium term

Immediate/short term homogeneity suggests efficiencies from fewer sessions. Effect confounded at medium term

Same result as for Session hours with a therapist

Limited data, but homogeneity Limited data, but heterogeneity favors active suggests homework adds participation with no little to active participation homework Heterogeneous in favor of Homogeneity implies that no/one-off homework regular homework (beyond a one-off expectation) could be dispensed with, but not for relaxation therapies or CBT

Therapist variables Therapist discipline 1. Psychology, 2. Social workers do well (refer Psychiatry, 3. Social text below) work, 4. Counseling/trained therapist, 4. Nursing, 5. Multidisciplinary team, 6. Lay, 7. Mixed lay and professional team, 8. Not applicable (e.g., bibliotherapy)

Trend favors psychologists, psychiatrists and counselors/trained therapists

(Continued on next page)


629

TABLE 4 Therapy Delivery Mode, Dose, and Therapist Variables, Summary (Continued)


Variable and coded levels

Screened

Unscreened

Therapist experience level Trend favors professionals Homogeneous 1. Lay, 2. Students over students (including those with training), 3. Practitioners/professionals, 4. Mixed lay and professional, 5. Not applicable Therapist involvement 1. Minimal, i.e., Small trend favors one-on-one Homogeneity implies group one-on-one initial involvement involvement over contact for setting up one-on-one the research only; 2. Group, i.e., contact as part of group delivery and possibly initial one-on-one contact for set-up purposes as well; 3. One-on-one, i.e., individually delivered therapies or those with both group and individual components; and 4. Intense, i.e., therapist was available at group or individual sessions and also beyond the normal session frame, e.g., on crisis call. CBT = Cognitive behavior therapy. Windsorized untreated control data, omnibus outcome, at early times unless described as: immediate and short-term (up to one month after therapy); medium term (1 – 6 months); and combined late times (more than 6 months). ‘Heterogeneous,’ as opposed to ‘homogeneous,’ is used when statistically significant differences were found between variable categories, and the weaker term ‘trend’ when this was not established, but the results from one category appeared nonetheless quite different from others. a. This category originally coded for the active (interaction, discussion, exercises) or passive (listening, reading) nature of work, plus whether homework was required, but there was only one study in the untreated control data that used a passive protocol, so that study was excluded and the analysis effectively became a comparison of the effect of the presence or absence of a homework / practice requirement.

one-on-one delivery for screened samples. Social workers did particularly well with screened patients (g = .92), but the types of protocol they delivered were more structured and simple than the mixture of protocols administered by some other disciplines, and two of the five studies that produced this outstanding effect size were outliers that simultaneously screened both ways. Social workers also appeared to do particularly poorly with unscreened patients (g = -0.03) but n was only two.

630


NONPROFESSIONAL INTERVENTIONS


There were few data available from interventions administered by nonprofessionals and when delivered independently of, rather than packaged with, professionally administered therapy components, results were generally null. However, the pair of studies by Weber et al. (Weber et al., 2004; Weber et al., 2007) showed that professionally instructed and supervised cancer survivors can achieve moderate-strong therapeutic effects against depression (gs of 0.96 and 0.59). Other studies showed that survivors can contribute to effective professional packages (e.g., where a survivor visit followed professional presurgical education; Ali & Khalil, 1989).

“INDIRECT” INTERVENTIONS There were few studies in this group so meta-analysis was not possible, but because they tended to produce moderate–strong effect sizes and this type of intervention has considerable potential utility, data relating to each study are presented in Table 5.

Mechanisms of Effectiveness Effect size data for three constructs that may be mechanisms of effectiveness were recorded: self-efficacy, perceived control, and self-esteem. However, once the two study design features were structured into analyses, sufficient data were available only in relation to the omnibus outcome and to selfefficacy, and then n > 3 were available only in relation to screened patients and relaxation (g = 0.52, n = 5), CBT (0.57, 7), and expressive-support (0.45, 4)(all ps < .05). Self-efficacy was defined broadly to include measures of perception of control and dispositional optimism/sense of coherence. Against this outcome, problem solving was the specific component that yielded the highest n (5) and produced an effect of moderate magnitude (screened, untreated controls, g = .64, p = .016).

DISCUSSION Therapy Types and Specific Components Consistent with the conclusion of the meta-analysis by Devine and Westlake (1995), our analyses with the data available did not allow us to identify any one therapy type as generally superior to others. The clinical implication is that therapy type should be selected appropriate to the individual patient. Distinctive patterns of outcome from each therapy type are now discussed.


631

TABLE 5 Indirect Interventions

Study Name and Confound Codes

ES (study n), Assessment Point, Outcome Construct

Descriptive Notes


Immediate Intervention Target: Significant other Bultz, Speca, Brasher, Geggie, & Page, 2000, Unclear/Untx

0.71† (32) im anxiety 0.84∗ (34) im depression

Goldberg & Wool, 1985, Unsc/Untx

0.67 (20) im anxiety

Rosenbaum, 2006, Unsc/Tx

0.43 (15) st depression

An education and support group for the significant others of breast cancer patients. Results for anxiety and depression fell only very slightly at mt, and were not reported for distress. Large baseline differences were not tested for significance and were not adjusted for, and could explain the results. Significant others of newly diagnosed lung cancer patients received 12 sessions of expressive-supportive therapy. Written emotional expression for the husbands of early stage breast cancer patients. Effect size rose from 0.2 at im.

Immediate intervention target: Doctor Rutter, Iconomou, & Quine, 1996, Sc.out/Tx Stewart et al., 2007, Unsc/Tx Kristeller, Rhodes, Cripe, & Sheets, 2005a, Unsc/Tx

0.23 (36) im anxiety 0.69∗ (36) im depression 0.15 (102) im distress 0.37† (111) st depression 0.37† (111) st distress

Doctors were trained in two 45-minute workshops to convey information re upcoming treatment with structure and style that aided understanding and recall. Intensive continuing education workshop in patient centered communication for family physicians, oncologists, and surgeons taking 6 hours rather than the usual 2. Oncologists were trained to make brief enquiries as to patients’ spiritual well-being. Training took 2–3 hours, and the one-off delivery to patients added an average of 2 minutes to the time spent with controls. Scores rose from negligible or slightly negative (n.s.) at im. to the st. measure at 3 weeks.

ES = Hedges g effect size point estimate; † = statistically significant at p < 0.10; ∗ = statistically significant at p < 0.05. Assessment point codes: im = immediately after intervention; st = short term follow up, i.e., up to one month after intervention, but not immediately after; mt = medium term, i.e., 1 – 6 months after intervention; ES = Hedges g effect size point estimate. Confound codes: screening status is presented first, then the nature of the control group after the slash: Unclear = screening status was unclear in the report; Unsc = unscreened; Sc.out = screened out for psychological history or complexity; Untx = untreated control comparison; Tx = treated control. a. This study was not actually coded “indirect” for the purposes of analysis, because it was decided that doctors did not receive an intervention (communication training) but, rather, delivered an intervention, having been trained in how to do so. It is a fine distinction, however, so the study is included here.

EDUCATION Education was shown to primarily treat anxiety, most effect sizes were small, and some confidence intervals fell below zero indicating that distress was


632


increased for some patients. However, there was relatively little data comparing with untreated controls, and no data for patients distressed at baseline. The therapy component that performed best and had sufficient n to merit attention was the teaching of coping strategies. Lack of data meant that the value of other specific education components such as informing patients about the disease, its medical treatment, and accompanying emotional issues cannot be commented upon though they may well have merit. It may be that the effect of education is sensitive to how freely available information about cancer is in a given society or to a given subpopulation. Two studies from developing nations where information was not freely available produced effect sizes of outlier proportions (Ali & Khalil, 1989; Corchado, 2006). Issues for research attention include the targeting of populations with particular needs for education, efficient but effective delivery (e.g., through indirect or nonpersonal intervention), and the discovery of moderators responsible for the distressing effect of education on some patients.

RELAXATION Relaxation therapies were also shown to primarily treat anxiety. Progressive muscle relaxation and diaphragmatic breathing performed best on available data. Few studies screened for baseline distress but from those that did some moderate magnitude or large effects were produced—including against depression, somewhat surprisingly. Studies that screened out patients with a history of distress produced moderate magnitude early times results against anxiety and distress (nothing against depression) which improved in the medium term (up to 6 months). Results suggest future research on distressed patients and patients from sociodemographic and medical groups who are likely to be more distressed in the context of cancer, and, perhaps, patients without distress history. Investigation into efficient administration to such groups, together with assessment over the long term, could be worthwhile.

CBT With one possible exception, the evidence showed that unscreened patients did not benefit from CBT, whereas patients without history of distress showed a small early benefit, perhaps increasing over time for the important depression result. Greater benefit was experienced by samples who were distressed at baseline, and simultaneous screening both ways further moderated results, producing the highest effects. The best-performing specific components were cognitive restructuring, challenging negative thoughts, self-monitoring, problem identification, problem solving, pleasant activity scheduling, and assertiveness training.


633


Our results suggest that there is no point in treating patients with CBT unless they have elevated distress, or are from a population particularly vulnerable to it in the cancer context. The dynamic involving patients who are distressed at baseline and without history of distress deserves investigation with assessment at long term. It is speculated that though distress motivates patients for change and provides psychometric “floor space” to register improvement, patients without distress in their backgrounds are also likely to have the internal and external resources to enable them to fully exploit this highly structured therapy. It may be that these personal resources also act to resist erosion of effect over time. Future research could inform differential service provision for patients with, versus without, a history of distress. EXPRESSIVE SUPPORT Expressive-supportive therapies produced some spectacular results against all three outcomes at early times for patients distressed at baseline, but from only two studies each and with no late times follow-up data available. Existentially oriented therapy did better than that oriented toward the biopsychosocial effects of the disease generally. Unscreened patients gained a small and durable benefit, but it evaporated when compared with placebo/treated control. This evidence suggests that expressive-supportive therapy is of little value to patients who are not experiencing elevated distress, nor belong to a population particularly vulnerable to it in the context of cancer. The limited data available for distressed patients suggests the therapy has strong potential as a first line intervention, and more research with such samples and of long term effects would be valuable. NONPROFESSIONAL

AND INDIRECT INTERVENTIONS

It is not surprising that nonprofessional interventions do not do well in these analyses given that they typically have very broad, practical, and preventative aims and that their clientele are often those comfortable with reaching out for assistance—often middle-class women with breast cancer. Such services doubtless play a role in curtailing the need for professional help. Despite generally very small effects against our psychological outcomes, there were individual studies that showed the potential for survivors to contribute effectively to professionally overseen interventions, as detailed above. Indirect interventions showed exciting potential for making impact with efficiency by improving the communication skills of medical staff, and for benefitting patients who cannot or will not come to therapy by treating a significant other. Medical specialists are in a unique position of intimate access and credibility, and the doctor–patient relationship should be exploited as fully as possible to enhance patients’ mental health. Invariably

634


indirect interventions result in higher patient satisfaction with their doctors also, which should improve treatment compliance, feeding back into improved mental well-being. The potential of indirect therapies for producing broad and efficient benefits deserves much greater attention.


Delivery Variables A good deal of homogeneity resulted from comparisons around therapy delivery variables (mode-, dose-, and therapist-related) suggesting that the ‘hows’ of therapy delivery are not the most important factors in producing effect size—or, at least, not factors that require much improvement in the way that they are already managed. It also suggests some scope for efficiencies where therapy types allow. Homogeneity was apparent in the effectiveness of more and fewer sessions, but for screened patients assessment at up to one month after therapy favored intervention spread over more weeks (nonsignificant trend). An earlier finding (Rehse & Pukrop, 2003) of significantly greater effect for therapies of more than 12 weeks duration was not replicated. Self-directed work emerged comparatively strongly suggesting another opportunity for efficiency where therapy type allows, but perhaps not for screened patients. Heterogeneity favoring no or one-off homework was found for unscreened patients. It may be that this group, which lacks proven distress, is not motivated to comply with demanding homework expectations. For screened patients, in relation to CBT and relaxation, a trend in favor of homework was shown which is consistent with a meta-analysis sampling general psychological clients (Kazantzis, Deane, & Ronan, 2000). Regarding therapist variables, for unscreened samples there was a nonsignificant trend in favor of those disciplines that specialize in work of a more psychological nature, which may hint that such specialized skill is necessary to elicit results where baseline distress has not been established. For screened patients there were trends toward more professional and one-on-one involvement by therapists. Results for social workers working with screened patients and structured protocols were strong, contrary to an earlier finding (Cwikel & Behar, 1999), suggesting an avenue for efficiencies. Note that although findings of homogeneity in relation to most delivery variables suggest the possibility of efficiencies, the dataset itself is biased toward “easy cases” in terms of the medical, psychological, and sociodemographic status of patients. For example, patients with serious medical comorbidities were often excluded, notwithstanding that such cases are common in the clinical setting, and so were patients whose presentation was complicated by psychosis, dementia or suicidality. Our results therefore indicate only a starting point for relatively simple cases.


635

Self-Efficacy It was not surprising that the self-efficacy of screened patients was lifted moderately by relaxation therapies and CBT, consistent with the possibility that this construct may be a mechanism of therapy effectiveness. However, the similar result for expressive-supportive therapies—which do not teach mastery overtly—is interesting.


CONCLUSION This meta-analysis set out to help clarify moderators of intervention effectiveness in a complex field. Our two earlier articles explicate findings of primary study design features and patient sociodemographic and medical characteristics that moderate effect size. This article boils down to a general finding that the ‘hows’ of intervention—therapy types and means of delivery— are probably the least important consideration in producing strong effects. Although there appear to be gains available in terms of effectiveness and efficiency by developing therapy content and means of delivery, the single most important clinical factor that has emerged from this study is the importance of baseline distress, and the significance of that factor for populations that are particularly vulnerable in the context of cancer. Here are our recommendations:

Practice 1. Risk and distress screening. Patient characteristics that have predicted or trended towards higher effect sizes—older (and, probably, younger than age 40), poorer, male, single, cancer site/stage other than early-stage breast—should be included in a risk screen together with psychological history and present distress symptoms. Initial risk and regular distress screening should be built into national practice guidelines and standard medical procedures. Particularly sensitive time points for distress screening may be shortly after diagnosis, toward the end of medical treatment, at recurrence, and at distant disease spread. It is important that patients at risk not be expected to identify their own distress and seek help, nor that medical staff be expected to detect distress simply in the course of their duties. 2. Service access through distress. Elevated distress should be the entry criteria for therapies requiring a significant therapist delivery. Interventions that fulfill general functional needs (e.g., education) are an exception to be provided to all. However, there are efficiencies that are possible in delivery of those therapies (refer below).


636


3. Effective and efficient delivery mechanisms. Less well paid professionals such as social workers can very effectively deliver manualizable therapies (relaxation and CBT) to distressed patients, and particularly those who do not have psychologically complex backgrounds. The services of more highly paid professionals (psychologists and psychiatrists) can be reserved for more complex cases and also to give attention to tuning systems of general medical care to reduce and detect distress. It may be best to do without homework where it is not an integral part of the therapy (as it often is for relaxation and CBT), and beyond a one-off introductory dose. Brief therapies can be highly effective delivered to vulnerable subpopulations. Longer therapies will no doubt be required when cases are complex. It seems likely that much educational material and perhaps some relaxation therapies may be as effectively provided to unscreened patient populations by cheap non-interactive means (e.g., DVD or pamphlet) as by interactive means, but obviously it is important that patients are encouraged to engage with the material (e.g., given a brief introduction by a nurse with an invitation to ask questions). Psychological quality in content and delivery form should be ensured.

Research 1. Sample selection. With an exception made to provide for the functional needs of all patients with appropriate educational intervention, research samples should always be either prescreened for distress or selected on the basis of reason to believe that a particular population generally suffers heightened distress in the cancer context. Sampling the right people needs to take priority over study design or therapy variables. This is challenging but in some cases payment for time and/or transport will be all that is required to enable poorer or minority people to participate. Another option may be to take the therapy to the patient when he or she is already at hospital for medical treatment or consultation or to deliver the therapy in a do-at-home form. A simple mechanism such as the label used to refer to a therapy type may make a difference to the participation of men (we suggest terms like “information,” “coaching.” and “buddying” instead of “counseling” and “support”), and the presentation of therapies may make all the difference to uptake by some minority ethnicities also. Although some sensible method of control is necessary to head off maturation and selection effects, appropriate sampling should not be sacrificed in favor of a particular method of allocating participants to groups (randomization). It is also more important to reach more vulnerable populations (e.g., those suffering less common cancers) than to attain large sample size. Lower sample size reduces the power to achieve statistically significant results, but this should be



2.

3.

4.

5.

6.

7.

8.

9.

637

compensated for by higher effect sizes. Furthermore, effect sizes can and should be published regardless of their statistical significance and these data can be combined with those from other small studies by meta-analysis to test for significance (Kline, 2004). Preliminary research directed solely at how to effectively recruit an appropriate sample may well be required in relation to vulnerable populations. Patients without history of psychological distress. It may be that such noncomplex patients can take better advantage of some therapies— particularly more demanding cognitive or behavioral therapies—or therapy generally, and retain effects longer. Data distinguishing them should be broken out and analyzed for differences in effect size and durability. Data needed. Longer term data (6 months and one year) are needed generally. For the sake of enabling studies to be incorporated in future meta-analyses, effect sizes or means and standard deviations at each assessment point should be provided. Illness-specific measures should be used. Except where a particular combination of therapy types is indicated theoretically, therapies should be delivered individually to enable the effects of particular therapies to be understood. Refer to the tables in this article for data gaps relative to specific therapy types. Relaxation issues. Relaxation is an anxiety therapy, and yet there were results that indicated that it could be effective against depression where patients were distressed at baseline, inviting research. Theoretical mechanisms. Theorized mechanisms of effectiveness should drive therapy design, including its fit with sample characteristics, and be explicated, measured, and built into analysis. Self-efficacy has shown relevance in this study, and the quality of patients’ social support should be much more often assessed and factored into result analyses. The nature and value of the social support inherent in many delivery modes may be interesting to investigate. Indirect and professionally supervised non-professional interventions. The considerable potential of these types of interventions to efficiently and effectively reach patients in large numbers or who would otherwise not be accessible is exciting and deserves attention. Existential issues. This is a relatively poorly researched area in AngloWestern countries but expressive-supportive therapies relating to existential, spiritual, grief, or death issues showed considerable potential, as did an oncologist-delivered intervention (Kristeller, Rhodes, Cripe, & Sheets, 2005). Another area deserving of attention. Middle and lower income countries. The need for research is great in these countries and effect sizes from countries outside the Organization for Economic Co-operation and Development (who mostly used distress screening and untreated control comparisons) were strong. Reviews and meta-analyses. To avoid the distortion of conclusions, future syntheses of this body of literature need to take into account the two


638


study variables found to structure the data set, namely, recruit screening (at least for distress, and preferably for psychological history) and the nature of the control comparison. All studies with comparable controls should be admitted, not just randomized trials, unless there is empirical evidence to justify the distinction. As noted in our earlier report, large randomized stratified studies often use unscreened middle-class earlystage breast cancer patients, presumably to attain sample size, but this sampling predicts low effect size, and these “quality studies” thus jeopardize the reputation of psycho-oncology for therapy efficacy. Participant characteristics should be taken into account in designing syntheses and in drawing conclusions. 10. Foreign language studies. There is a need to incorporate the body of trial data reported in foreign languages into the meta-analysis of moderators. Some allowances for cultural difference would be necessary, but outcomes would be powerfully persuasive.

CLOSING REFLECTION When the proposal for funding this research was originally submitted, it was not imagined that such substantial recommendations for practice and research would result. However, the approach that was taken has revealed that the confusion that has beset the field has deep roots and therefore significant implications. The poor effects shown by many studies—primary and review studies—appear to be the product of blind adherence to biomedical ideals in research, i.e., the notion that “the large randomized double-blind trial” is the model of quality, which persist despite Cochran’s advice that there is no “gold standard” study design (Cochrane Collaboration, 2006, para. 6.11) and of how ill this design fits the doubly clinical field of psycho-oncology. It is hoped that, in future, researchers and reviewers will feel justified and emboldened by the evidence provided in this study to think in more creative and effective scientific ways about how best to approach work with those of this special population who need intervention.

REFERENCES Ali, N. S., & Khalil, H. Z. (1989). Effect of psychoeducational intervention on anxiety among Egyptian bladder cancer patients. Cancer Nursing, 12(4), 236–242. Barsevick, A. M., Sweeney, C., & Haney, E. (2002). A systematic qualitative analysis of psychoeducational interventions for depression in patients with cancer. Oncology Nursing Forum, 29(1), 73–86. Bultz, B. D., Speca, M., Brasher, P. M., Geggie, P. H., & Page, S. A. (2000). A randomized controlled trial of brief psychoeducational support group for partners of early stage breast cancer patients. Psycho-oncology, 9, 303–313.



639

Cochrane Collaboration. (2006, September 2006). Cochrane handbook for systematic reviews of interventions, version 4.2.6. 4.2.6. Retrieved from http:// www.cochrane.org/resources/handbook/hbook.htm Corchado, J. L. (2006). The effects of preparatory sensory information on breast conservative and mastectomy cancer patients: Mood states and self-concept (Unpublished 3244765). Chester, PA: Widener University School of Nursing. Cwikel, J., & Behar, L. (1999). Social work with adult cancer patients: A vote count review of intervention research. Social Work in Health Care, 29(2), 39–67. Cwikel, J., Behar, L., & Rabson-Hare, J. (2000). A comparison of a vote count and a meta-analysis review of intervention research with adult cancer patients. Research on Social Work Practice, 10, 139–158. Devine, E. C., & Westlake, S. K. (1995). The effects of psychoeducational care provided to adults with cancer: A meta-analysis of 116 studies. Oncology Nursing Forum, 22, 1369–1381. Goldberg, R. J., & Wool, M. S. (1985). Psychotherapy for the spouses of lung cancer patients: Assessment of an intervention. Psychotherapy and Psychosomatics, 43(3), 141–150. Heron-Speirs, H. A., Baken, D. M., & Harvey, S. T. (2013). Moderators of psychooncology therapy effectiveness: Meta-analysis of socio-demographic and medical patient characteristics. Clinical Psychology: Science and Practice, 19, 402–416. Heron-Speirs, H. A., Harvey, S. T., & Baken, D. M. (2012). Moderators of psychooncology therapy effectiveness: Addressing design variable confounds in metaanalysis. Clinical Psychology: Science and Practice, 19, 49–77. Kazantzis, N., Deane, F. P., & Ronan, K. R. (2000). Homework assignments in cognitive and behavioral therapy: A meta-analysis. Clinical Psychology: Science and Practice, 7(2), 189–202. Kline, R. B. (2004). What’s wrong with statistical tests - and where we go from here. In Beyond significance testing (pp. 61–91). Washington, DC: American Psychological Association. Kristeller, J. L., Rhodes, M., Cripe, L. D., & Sheets, V. (2005). Oncologist Assisted Spiritual Intervention Study (OASIS): Patient acceptability and initial evidence of effects. International Journal of Psychiatry in Medicine, 35(4), 329–347. Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage. Luebbert, K., Dahme, B., & Hasenbring, M. (2001). The effectiveness of relaxation training in reducing treatment-related symptoms and improving emotional adjustment in acute non-surgical cancer treatment: A meta-analytical review. Psycho-Oncology, 10(6), 490–502. Meyer, T. J., & Mark, M. M. (1995). Effects of psychosocial interventions with adult cancer patients: A meta-analysis of randomized experiments. Health Psychology, 14(2), 101–108. Naaman, S., Radwan, K., Fergusson, D., & Johnson, S. (2009). Status of psychological trials in breast cancer patients: A report of three meta-analyses. Psychiatry: Interpersonal and Biological Processes, 72(1), 50–69. Osborn, R. L., Demoncada, A. C., & Feuerstein, M. (2006). Psychosocial interventions for depression, anxiety, and quality of life in cancer survivors: Meta-analyses. International Journal of Psychiatry in Medicine, 36(1), 13–34.


640


Rehse, B., & Pukrop, R. (2003). Effects of psychosocial interventions on quality of life in adult cancer patients: meta analysis of 37 published controlled outcome studies. Patient Education and Counseling, 50, 179–189. Rosenbaum, D. I. (2006). What’s good for the gander is good for the goose: Helping cancer patients to cope by treating their spouses. Binghamton, NY: State University of New York at Binghamton, New York. Retrieved from http://proquest.umi.com/pqdweb?did=1221724391&Fmt=7&clientId=13636& RQT=309&VName=PQD Rutter, D., Iconomou, G., & Quine, L. (1996). Doctor-patient communication and outcome in cancer patients: An intervention. Psychology & Health, 12(1), 57–71. Schneider, S., Moyer, A., Knapp-Oliver, S., Sohl, S., Cannella, D., & Targhetta, V. (2010). Pre-intervention distress moderates the efficacy of psychosocial treatment for cancer patients: A meta-analysis. Journal of Behavioral Medicine, 33(1), 1–14. Sheard, T., & Maguire, P. (1999). The effect of psychological interventions on anxiety and depression in cancer patients: Results of two meta-analyses. British Journal of Cancer, 80(11), 1770–1780. Sherman, A. C., Mosier, J., Leszcz, M., Burlingame, G. M., Ulman, K. H., Cleary, T., . . . Strauss, B. (2004). Group interventions for patients with cancer and HIV disease: Part III. Moderating variables and mechanisms of action. International Journal of Group Psychotherapy, 54(3), 347–387. Stewart, M., Brown, J. B., Hammerton, J., Donner, A., Gavin, A., Holliday, R. L., . . . Freeman, T. (2007). Improving communication between doctors and breast cancer patients. Annals of Family Medicine, 5(5), 387–394. Tatrow, K., & Montgomery, G. H. (2006). Cognitive behavioural therapy techniques for distress and pain in breast cancer patients: A meta-analysis. Journal of Behavioral Medicine, 29(1), 17–27. Weber, B. A., Roberts, B. L., Resnick, M., Deimling, G., Zauszniewski, J. A., Musil, C., . . . Yarandi, H. N. (2004). The effect of dyadic intervention on self-efficacy, social support, and depression for men with prostate cancer. Psycho-Oncology, 13(1), 47–60. Weber, B. A., Roberts, B. L., Yarandi, H., Mills, T. L., Chumbler, N. R., & Wajsman, Z. (2007). The impact of dyadic social support on self-efficacy and depression after radical prostatectomy. Journal of Aging and Health, 19(4), 630–645. Zabalegui, A., Sanchez, S., Sanchez, P. D., & Juando, C. (2005). Nursing and cancer support groups. Journal of Advanced Nursing, 51(4), 369–381.


641


APPENDIX: DETAIL OF THERAPY TYPE CATEGORIZATION 1. Education/information provided by a professional regarding cancer, cancer treatments, facilities (including orientation tour), or adjunctive services, nutrition, exercise, coping strategies, or symptom management; includes bibliotherapy or information provided by some technological means, but does not include active rehearsal of new behaviors 2. Relaxation focused cognitive-behavioral treatment, i.e., counseling/training in the use of coping strategies that focus on relaxation or stress management, including progressive muscle relaxation, mindfulness-based stress reduction, guided imagery, meditation, hypnotherapy, diaphragmatic breathing, autogenic training, systematic desensitization, biofeedback, electromyography, distraction, music. 3. Cognitive behavioral therapy, broadly focused, i.e., counseling/training in the use of coping strategies that focus on cognitive reappraisal or behavior modification or reinforcement, such as cognitive restructuring/reappraisal, challenging negative thoughts, positive self-talk, self-monitoring of thoughts or skills taught, problem identification, problem solving, contingency management, goal/expectation setting, activity pacing, behavioral activation, pleasant activity scheduling, assertiveness/communication/relational skills training, disability management, emotional control and anger management, fighting disease, cathartic, active interpretation / reconstruction, and may include role play or modeling. Problem solving included a wide range of topics, e.g., loneliness and isolation, morale and self-management, sexuality and contact, body self-esteem and general mood, communication, body self-image and social adjustment, existential plight, social alienation and self-identity, emotionality and personal control, dysphoria, and depression, 4. Expressive-supportive therapy (nondirective professional counseling/psychotherapy), i.e., interactive verbal interventions, including nondirective, psychodynamic, existential, emotionally supportive/expressive/reflective regarding the disease, its treatment, prognosis, and recovery, disability or death, general or crisis intervention; no specific behavioral or coping skills are taught; includes social support by professionals, but excludes therapist reconstruction. 5. Nonprofessionally led support or counseling, e.g., survivor testimony, self-help groups, telephone counseling, the teaching of coping skills by lay persons. 6. Indirect intervention, i.e., the immediate target of intervention is someone other than the cancer patient (e.g., communications training or counseling directed at medical staff or spouse without the patient present) but with the intention of benefiting the patient. (Measures were taken on the patient.) 7. Other (which transpired to comprise mostly written emotional disclosure studies) Note that if a study was categorized “indirect,” it could not also be coded under any other category, e.g., the counseling of a spouse without the presence of the patient (but outcome measures are taken on the patient) would be coded “indirect” but not also “expressive-supportive.” This is because this category was identified by its distinctive delivery mechanism rather than its content so it was not comparable with other studies.

Cognitive predictors and moderators of winter depression treatment outcomes in cognitive-behavioral therapy vs. light therapy.

Client characteristics as moderators of the relation between the therapeutic alliance and outcome in cognitive therapy for depression.

Effectiveness of prednisolone during phenytoin therapy.

Identifying moderators of the adherence-outcome relation in cognitive therapy for depression.

Treatment moderators of cognitive behavior therapy to reduce aggressive behavior: a meta-analysis.

Moderators of treatment response to trauma-focused cognitive behavioral therapy among youth in Zambia.

Delivery Systems Can Determine Therapy Cost, and Effectiveness, More Than Type of Therapy.

The effectiveness of family therapy: a review of outcome research.

Optimal duration of androgen deprivation therapy following radiation therapy in intermediate- or high-risk nonmetastatic prostate cancer: A systematic review and metaanalysis.

Effectiveness of antiepileptic therapy in patients with PCDH19 mutations.

Effectiveness of phosphodiesterase-5 inhibitor therapy for portopulmonary hypertension.

Effectiveness of therapy after stroke is time dependent.

The Effectiveness of Neural Therapy in Patients With Bell's Palsy.

Cost-effectiveness of Telaprevir combination therapy for chronic hepatitis C.

Effectiveness of High-Intensity Laser Therapy in Subacromial Impingement Syndrome.

A systematic literature review on the effectiveness of eurythmy therapy.

The effectiveness of Multisystemic Therapy (MST): a meta-analysis.

Safety and effectiveness of endovascular therapy for claudication in octogenarians.

Effectiveness of mobilization therapy and exercises in mechanical neck pain.

The effectiveness and safety of electroconvulsive therapy (ECT).

Subacromial impingement syndrome--effectiveness of physiotherapy and manual therapy.

[Effectiveness of neuro-electric therapy in drug resistant endogenous psychoses].

Effectiveness of Vitamin D Therapy in Orthopaedic Trauma Patients.

Cost-effectiveness of proton beam therapy for intraocular melanoma.