INTRODUCTION
What are the effects of addressing meaning in life with clients? In past millennia, talking about meaning in life was often restricted to clergy and philosophers, but in modern secular societies this task is more often taken up by psychological therapists, pioneered by Viktor Frankl and others (Vos, Reference Vos and Russo-Netzer2016a). Although some therapists have used strong metaphysical formulations in the past, meaning in life is nowadays used as a nonreligious term to describe a set of psychological experiences that can be empirically distinguished from such phenomena as happiness, meaninglessness, and ordinary daily life (Steger, Reference Steger and Wong2012; Vos, Reference Vos2017a). For instance, a large number of cognitive laboratory experiments, psychometric studies, and surveys have established that most people search for meaning, experience its presence, or use meaning-related coping styles (e.g., Batthyany & Russo-Netzer, Reference Batthyany and Russo-Netzer2014; Hicks & Routledge, Reference Hicks and Routledge2013; Greenberg et al., Reference Greenberg, Koole and Pyszczynski2013; Wong, Reference Wong2013; Reker, Reference Reker, Reker and Chamberlain2000). Meaning has been empirically defined as an individual's subjective sense of purpose, values, understanding of self and the world, self-worth, action-directed goals, self-regulation, and coping with existential challenges (Wong, 2012; George & Park, Reference George, Park, Batthyany and Russo-Netzer2014; MacKenzie & Baumeister, Reference MacKenzie, Baumeister, Batthyany and Russo-Netzer2014; Vos, Reference Vos and Russo-Netzer2016b). Individuals in more than 150 studies worldwide reported five types of meaning: material, social, self-oriented, higher types, and existential-philosophical types (Vos, Reference Vos and Russo-Netzer2016a).
Across a range of research, meaning has been shown to be important for clients. For instance, individuals with moderate to severe psychopathology described “living a meaningful and satisfying life” as the core of their psychological recovery in 75 studies (Andresen et al., Reference Andresen, Oades and Caputi2003; Reference Andresen, Oades and Caputi2011; Slade et al., Reference Slade, Leamy and Bacon2012). Perceiving life as meaningful is strongly correlated with a higher quality of life, lower levels of psychological stress (e.g., depression and anxiety), and better physical well-being (Ryff et al., Reference Ryff, Singer and Love2004; Reference Ryff, Love and Urry2006; Steger, Reference Steger and Wong2012; Brandstätter et al., Reference Brandstätter, Baumann and Borasio2012; Roepke et al., Reference Roepke, Jayawickreme and Riffle2014). Furthermore, perceived meaning functions as a source of resilience and meaning-centered coping styles reduce stress after traumatic life events (Folkman, Reference Folkman2008; Park & Folkman, Reference Park and Folkman1997; Park, Reference Park2010; Vos, Reference Vos and Russo-Netzer2016a). However, mental healthcare has traditionally overlooked meaning-centered concerns, such as experiencing a lack of meaning or purpose in life, or being unable to adjust life goals after such life-changing experiences as the loss of a loved one or a chronic disease (Vos, Reference Vos and Russo-Netzer2016a). In response, it has been argued that mental healthcare should be transformed from curing psychiatric symptoms to supporting clients to live a meaningful life (Seligman et al., Reference Seligman, Steen and Park2005).
Many authors suggest that meaning-centered concerns and their treatment are specifically relevant at transitional moments in life: adolescence, midlife, retirement (Battista & Almond, Reference Battista and Almond1973), bereavement (Neimeyer et al., Reference Neimeyer, Harris and Winokuer2011), or trauma (Schulenberg et al., Reference Schulenberg, Drescher, Baczwaski, Batthyany and Russo-Netzer2014). Researchers have particularly focused on individuals with a chronic or life-threatening disease, a majority of whom wonder how to live a meaningful life despite their illness (e.g., Henoch & Danielson, Reference Henoch and Danielson2009; Vehling et al., Reference Vehling, Lehmann and Oechsle2012; Vos et al., Reference Vos, Asperen and Oosterwijk2013). However, only a fifth report clinical depression or anxiety (e.g., Mitchell et al., Reference Mitchell, Chan and Bhatt2011), which explains the modest effect sizes of psychological treatments that focus on reducing psychopathology and not on meaning-centered concerns (Faller et al., Reference Faller, Schuler and Richard2013; Hart et al., Reference Hart, Hoyt and Diefenbach2012; van Straaten et al., Reference van Straaten, Geraedts and Verdonck-de Leeuw2010). Furthermore, cross-sectional studies demonstrate that patients cope better with their disease when they “create meaning out of chaos” (Bullington et al., Reference Bullington, Sjöström-Flanagan and Nordemar2005; King et al., Reference King, Hicks and Krull2006) and use meaning-centered coping styles (Park, Reference Park2010; Folkman, Reference Folkman2008).
To address the clients' meaning-centered needs, more than 28 different meaning-centered therapeutic approaches have evolved (see an overview in Vos [Reference Vos2017]). Some 21 include meaning among other therapeutic aims, such as acceptance and commitment therapy (Hayes et al., Reference Hayes, Strosahl and Wilson2012) and positive psychology (Seligman & Csikszentmihalyi, Reference Seligman and Csikszentmihalyi2000), and 7 for whom meaning is the predominant focus. This last group will be referred to as meaning-centered therapy (MCT), defined as a therapeutic approach addressing meaning in life as its main aim, with a systematic approach. Worldwide, 69 MCT institutes exist, associated with tens of thousands of therapists (Correia, Reference Correia2015). Preliminary metaanalyses of a small sample of six MCT trials showed large effect sizes on psychological well-being and quality of life, but this included only English publications with an existential-therapeutic focus (Vos et al., Reference Vos, Cooper and Correia2015). Thus, despite its popularity and positive preliminary findings, the precise effectiveness of MCT remains unclear. However, practitioners need to understand the potential beneficial effects of addressing meaning, with respect to self-selection biases and such nonspecific effects as time, attention, treatment expectations, and the difference in effectiveness between other therapeutic treatments (Chambless & Hollon, Reference Chambless and Hollon1998).
We have therefore systematically reviewed effectiveness studies on different types of MCT in multiple languages. As psychological treatments are usually complex and consist of multiple therapeutic techniques, eligible studies were semistructured and standardized via treatment manuals (Carroll & Nuro, Reference Carroll and Nuro2002). As placebo conditions and complete blinding are impossible in talking therapies, no type of control conditions were considered (Chambless & Hollon, Reference Chambless and Hollon1998). It has been recommended to focus reviews only on bona fide psychological treatments that address specific therapeutic aims via specific therapeutic methods (Wampold et al., Reference Wampold, Minami and Baskin2002). Therefore, this review only included MCT manuals primarily and were solely aimed at improving quality of life, such as meaning in life (primary outcome) and reducing such psychological stress as depression and anxiety (secondary outcome), via explicitly addressing meaning in life with a systematic method (mediator). Meaning-centered therapists hypothesize that MCT improves the client's meaning-making skills, which subsequently reduces their psychological stress (comprehensive reviews of the logical and conceptual model of MCT can be found elsewhere: Vos, Reference Vos and Russo-Netzer2016a; Reference Vos and Russo-Netzer2016b). This was tested via metaanalyses, examining the effects of MCT on quality of life and psychological stress, and testing the hypothesis that improvements in perceived meaning reduce psychological stress.
METHOD
Study Selection Rounds
The systematic literature review was conducted in nine consecutive rounds, as Figure 1 shows, in line with the PRISMA and MOOSE guidelines (Liberati et al., Reference Liberati, Altman and Tetzlaff2009; Stroup et al., Reference Stroup, Berlin and Morton2000). First, multiple databases were employed: PubMed, the Web of Knowledge, PsycINFO, Medline, Embase, scholar.google.com, and Scopus. Search terms and key authors were based on an exploratory review (Vos et al., Reference Vos, Cooper and Correia2015). These combined intervention terms included (therap*/counsel*/coach*/intervention/analys*/treat*/care/support/psychol*/psychiat*/outcome*/result*/ effect*/change/eval*/assess*/trial*) and the meaning-centered nature (meaning-in-life/meaning-of-life/search-for-meaning/noogenic/noetic/logo-ther*/logo-anal*/purpose-in-life/life-purpose/goals-in-life/life-goals/meaning-cent*/meaning-mak*/meaning-orient*/existential-analys*/logo-anal*), or key authors (Breitbart/Fabry/Fillion/Frankl/Henry/Langle/Lee/Lukas/Marshall/Neimeyer/Starck/Wong/Zuehlke). Given the large number of findings, we added PubMed Mesh terms ([counselling] OR [psychotherapy] OR [psychology]) and capped scholar.google.com results at 10,000 hits.
Fig. 1. Flowchart of included and excluded studies.
Second, we hand-searched the journals Existential Analysis and International Forum for Logotherapy and the websites of MCT institutes (Correia, Reference Correia2015). Third, 10 experts searched and translated references in Arabic, Chinese, Dutch, Farsi, French, German, Indonesian, Italian, Korean, Portuguese, Russia, and Spanish. Fourth, all included authors were contacted to identify other studies. All 10 experts were fluent in their mother tongue and English, were professionally educated in MCT and/or existential therapy, at least at the master's level, had more than two years of therapeutic experience, and were trained to search and translate references. Fifth, additional studies were identified via reference lists. Sixth, all titles and abstracts were initially screened for eligibility. Seventh, studies were excluded though a thorough reading of abstracts. Eighth, studies were excluded on the basis of full-text manuscripts. Ninth, studies were excluded from their analyses, when they did not have enough data or did not measure quality of life or psychological stress.
Eligible studies had the following characteristics: (1) aiming to treat specific individual psychological problems (self-discovery, religious, and philosophical practices excluded); (2) primarily aiming to support clients to live a meaningful life without other primary therapy aims that could conflate study effects (e.g., acceptance and commitment therapy [ACT] was excluded, as its founders explicitly state that it has a range of aims and methods and does not merely focus on meaning (see Vos [Reference Vos and Russo-Netzer2016a] for a description of differences between ACT and MCT); (3) having a systematic approach (e.g., stepwise exploration of different meaning-centered topics); (4) being designed as a full therapeutic approach, consisting of multiple therapeutic techniques, and not only one specific technique, such as autobiographic writing or paradoxical intention (e.g., Chochinov et al., Reference Chochinov, Hack and Hassard2005; Bohlmeijer et al., Reference Bohlmeijer, Smit and Cuijpers2003); (5) using valid and reliable psychometric outcome instruments; (6) meeting Wampold et al.'s (Reference Wampold, Minami and Baskin2002) criteria for bona fide interventions; (7) having a retrievable and intelligible full-text manuscript; and(8) not having unrealistically large effect sizes (Hedges' g > 3) or a high risk of bias, such as nonselective reporting according to Cochrane's risk-of-bias criteria (Higgins & Green, Reference Higgins and Green2008) to prevent biased results (Rosenthal, Reference Rosenthal1991; Reference Rosenthal1995).
Given the large number of references, author J.V. conducted all rounds. His selection was consistent with author D.V., who independently searched and screened a random sample of 500 references and conducted rounds 3–5 (interrater κ = 0.92 and 0.82). Interrater reliability of independent bias assessment and coding of study characteristics were high (κ = 0.87, 0.85). Both raters have a clinically and scientifically relevant background in MCT, existential therapy, and empirical research.
Metaanalytic Steps
All studies had sufficient clinical homogeneity, as identified by the 10 experts. Of course, clinical homogeneity does not necessarily imply statistical homogeneous findings. Instead of cherrypicking and only presenting the most statistically homogenous and effective findings, or only the largest samples—as seems common practice in metaanalyses—it was decided to be transparent. This article will describe how studies were excluded and included and present findings at each metaanalytic step. Presenting all findings seems particularly relevant for new therapies with relatively few studies. The authors believe that transparency about the metaanalytic steps may yield relevant information about validity and generalizability, and guide the future direction of the therapeutic field, especially in the case of heterogeneity.
The studies were coded and metaanalyses conducted and presented in nine consecutive steps (see Table 1). In the first step, overall effects for all trials were calculated. Second, metaanalysis tested the hypothesis that there was a significant difference between effects measured immediately after the last therapy session and effects at follow-up between 4 and 12 months later. As differences were found, all next metaanalytic steps were separately conducted for immediate and follow-up measurements. Third, metaanalysis tested the hypothesis that different outcome instruments have different effects. To test this hypothesis, each outcome instrument in each study was coded in either the overall categories “quality of life” and “psychological stress,” following previous studies (Vos et al., Reference Vos, Asperen and Oosterwijk2013). If multiple instruments were used in one study to measure one category, average effect sizes were calculated. Whether an instrument belonged to a category was decided on the basis of content and nonsignificant small heterogeneity (p(Q) > 0.05 and I 2 < 50%). As large heterogeneity was found, more specific subcategories were created on the basis of their content and low heterogeneity (see Table 2).
Table 1. Description of coding of variables and metaanalytic steps
Table 2. Overview of included instruments and subcategories
N/A = not applicable; ND = no data available.
Groups of outcome instruments: QoL = general instruments or subscales measuring quality of life; MiL = meaning in life; hope = hope, hopelessness and optimism; SE = self-efficacy; Soc = social relationships.
Specific categories of outcome instruments: AHS = Adult Hope Scale; AMIL = Adolescent Meaning in Life; BDI = Beck Depression Inventory; CDI = Children Depression Inventory; BHS = Beck Hopelessness Scale; BPRS = Brief Psychiatric Rating Scale; CBI = Core Bereavement Inventory; CES–D = Center for Epidemiological Studies Depression Scale; CSMLS = Chinese Sources of Meaning in Life Scales; DAS = Death Anxiety Scale; DASS = Depression and Anxiety Symptoms Scale; EWL = Eigenschaftsworterliste; ESK = Eksistenzskala; FACIT = Functional Assessment of Chronic Illness Therapy; GSES = Generalized Self-Efficacy Scale; HADS = Hospital Anxiety and Depression Scale; HAI = Hopelessness Assessment in Illness Questionnaire; HAS = Hamilton Anxiety Scale; HGRC = Hogan Grief Reaction Checklist; KASSL = Kieler Anderungssensitive Symptomliste; LOT = Life Orientation Test–Revisited; MAC = Mental Adjustment to Cancer Scale; MLQ = Meaning in Life Questionnaire; MQoL = McGill Quality of Life Questionnaire; MSAS = Memorial Symptom Assessment Scale; OHQ = Oxford Happiness Questionnaire; PSS = Perceived Stress Scale; PIL = Purpose in Life Scale; PMP = Personal Meaning Profile; PSWQ = Penn State Worry Questionnaire; POMS = Shortened Profile of Mood States; QoL = Quality Of Life scales; QoLC–E = Quality of Life Concerns at the End of Life; RGEI = Revised Greif Experience Inventory; RSES = Rosenberg Self-Esteem Scale; PTGI = Posttraumatic Growth Inventory; SAHD = Schedule of Attitudes Towards Hastened Death; SONG = Seeking of Noetic Goals; SPWB = Ryff's Scale of Psychological Well-Being; STAI = State–Trait Anxiety Inventory (only state used); SWB = Spiritual Well-Being Scale; TPF = Trier Personlichkeitsfragebogen; WHOQoL = World Health Organization Quality of Life Short Scale.
Fourth, metaanalysis tested the hypothesis that different study designs would cause different effect sizes. Studies were coded for being randomized and controlled, and for type of control condition (Table 1). A difference was found on the basis of large heterogeneity between studies and small heterogeneity within studies, and significant moderate/large contrasts (Cohen's d > 0.30, p < 0.05). The effect sizes were found to be inflated in noncontrolled trials. It was thus decided to focus on controlled trials only, to focus on estimating true effect sizes and not on artificially inflated effects. The next metaanalytic steps included both randomized and nonrandomized trials and all types of control conditions, as these did not differ in effects. When randomized and nonrandomized controlled and noncontrolled trials were compared, the effect sizes described changes in scores from baseline to posttreatment/follow-up measurement (“change effect”). This answered the question “How much change do individuals experience between the measurements before and after MCT?” When randomized and nonrandomized controlled trials were compared, the effect sizes described the effects of MCT compared with control groups (“relative effects”). This answered the question “What is the difference between the improvements in clients receiving MCT compared with clients in control conditions?”
Fifth, the hypothesis was tested that sample characteristics influence effect sizes, such as participants' inclusion criteria, age, and gender (Table 1), via contrasts, moderator, and metaregression analyses. The variables included were based on conceptual reviews and preliminary metaanalyses of MCT (Vos et al., Reference Vos, Cooper and Correia2015; Vos, Reference Vos and Russo-Netzer2016a; Reference Vos and Russo-Netzer2016b). Sixth, similar analyses were employed to test the hypothesis that the study results were influenced by treatment characteristics, such as type of MCT, group or individual format (Table 1, based on Vos et al., Reference Vos, Cooper and Correia2015). Seventh, similar analyses were used to test whether studies differed in effect size due to describing different types of therapeutic skills in the treatment manuals, which answered the question of whether therapeutic skills lead to larger effects. The treatment manuals were coded according to the presence of 39 core meaning-centered therapist skills that have been described elsewhere (Vos, Reference Vos and Russo-Netzer2016a; Reference Vos2017), such as: providing didactics about meaning in life, stimulating clients to set and experiment with achievable goals in daily life, focusing on self-worth, and doing mindfulness exercises (Table 1).
In step eight, similar analyses tested the hypothesis that effects were influenced by sample size, year of publication, or precision of effects (Dechartres et al., Reference Dechartres, Altman and Trinquart2014; Capellini et al., Reference Capellini, Smythe and Silva2012). Additional reanalyses of steps 1–6 were conducted, first only in studies with the 25% largest sample sizes, second only in studies published since 2000, and third only in studies with the 25% most precise effects. The effects were regarded similar to the original findings if the significance and magnitude of the effects were similar (significant/nonsignificant; small/moderate/large) and effect sizes did not differ by more than 10%.
In step nine, we tested the unique assumption of this therapeutic approach that MCT improves the client's perceived meaning, which subsequently reduces their psychological stress. The effects of MCT on improving meaning was already tested in previous steps, as part of the effects on quality of life. Metaanalyses tested the hypothesis that the changes in meaning in life predicted the changes in psychological stress. We assumed that the results from this statistical test would suggest mediation, that is, the improvement of perceived meaning explains the effects of MCT on stress (mediation). Thus, we did not only assume a correlation between these change scores but also a causal relationship: the reason for assuming mediation is that all treatment manuals described therapeutic techniques that explicitly and systematically addressed meaning with the aim of reducing stress, while the manuals in control conditions did not.
STATISTICAL PROCEDURES
We entered means, standard deviations, N, p value (or other statistics if not available, e.g., t value or F value), according to Introduction to Meta-Analysis (Borenstein et al., Reference Borenstein, Hedges and Higgins2009). When correlations among pre-, post-, and follow-up assessments were not reported, formulas were used (Morris & DeShon, Reference Morris and DeShon2002; Dunlap et al., Reference Dunlap, Cortina and Vaslow1996) or the average correlation of 0.70 was inserted. In case of multiple control groups, psychological treatments with the largest effect size were selected.
Effects were calculated with Hedges' g and its 95% confidence interval (95% CI). This is a variation on Cohen's d, corrected for biases due to small sample sizes and regarded a robust technique in the social sciences (Hedges, Reference Hedges and Olkin1985). This may be conservatively interpreted with Cohen's (Reference Cohen1988) convention of small (0.2), medium (0.5), and large (0.8) effect sizes. Random effects were calculated, as studies differed in terms of population and MCT type, and random effects adequately mirrored heterogeneity in behavioral studies with non-inflated alpha (α) levels (Hunter & Schmidt, Reference Hunter and Schmidt2000). Spurious outliers were identified in each metaanalytic step and discarded by using a trimming technique that excluded studies where the 95% CI was lower than the aggregated confidence interval of all studies (Borenstein et al., Reference Borenstein, Hedges and Higgins2009). Publication bias was tested in each metaanalytic step by visual inspection of funnel plots and calculation of Egger intercepts, using a trim-and-fill procedure, which provides an estimate of effect size after publication bias has been taken into account (Duval & Tweedie, Reference Duval and Tweedie2000) (n.b.: publication bias will only be reported in this article in case of significant bias). A-priori power analyses estimated that five or more studies are required to detect moderately large effect sizes, similar as what was done in previous studies (Vos et al., Reference Vos, Asperen and Oosterwijk2013), with a power over 0.80 (Valentine et al., Reference Valentine, Pigott and Rothstein2010; Borenstein et al., Reference Borenstein, Hedges and Higgins2009). If fewer than five studies reported on an outcome, this outcome was not presented.
RESULTS
Description of Included Trials
We found 52,220 citations (see Table 1), most of which were excluded due to irrelevance, duplication, or non-quantitative/non-trial designs. Some 32 trials were excluded for high risk of bias, 6 for irrelevant outcome measures, and 4 for lacking useful data. A total of 60 trials were included in the final review, covering 3,713 participants. This comprised 26 randomized controlled trials (N = 1,975), 15 nonrandom controlled trials (N = 709) and 19 nonrandomized noncontrolled trials with pre/post measures only (N = 1,029).
Studies were conducted in the Middle East (k = 18), North America (k = 16), South-East Asia (k = 14), Europe (k = 6), South Africa (k = 3), and South America (k = 2). Samples included physical illness (k = 26), psychiatric diagnoses (k = 8), transitional moments in life (k = 12), caregivers (k = 7), substance misuse (k = 4), marital issues (k = 2), and prisoners (k = 1). The mean age was 42.4 (SD = 16.2) years, ranging from 15 to 79. Some 31% were male (M = 31.7%, SD = 20.2%), 32.6% (SD = 19.7%) held their highest degree in higher education, 52.5% (SD = 17.6%) in secondary school/college, and 14.9% (SD = 4.9%) in primary school or had no education.
The control conditions included care as usual (k = 23), alternative treatment (k = 15), and waiting list (k = 3). As most care as usual included an alternative treatment (k = 16), these groups were reformulated as treatment. Alternative treatments (k = 23 + 15 = 38) included support groups (k = 19), cognitive behavioral therapy (k = 4), psychoeducation/bibliotherapy (k = 3), relaxation/mindfulness (k = 3), or a best-practice integration of treatments (k = 9). The mean number of MCT sessions was 8.65 (SD = 3.3). Some 32 trials included individual treatments, 26 groups, and 2 coupled treatments. Treatments included logotherapy (k = 15), general MCT (k = 29), meaning-centered psychotherapy (k = 9), meaning-making interventions (k = 3), existential analysis (k = 2), and meaning therapy (k = 1). Tables 3 and 4 describe these treatments, their differences and their overlaps.
Table 3. Characteristics of the trials included in the metaanalyses
MCT = meaning-centered therapy; N = number; N/A = not applicable; ND = no data available.
Types of control condition: AI = active intervention; CAU = care as usual; WL = waiting list.
Groups of outcome instruments: QoL = general instruments or subscales measuring quality of life; MiL = meaning in life; Hope = hope, hopelessness, and optimism; SE = Self-Efficacy; Soc = social relationships.
Table 4. Description of differences and overlap of the different types of meaning-centered therapy included in the metaanalyses
Metaanalytic Results
Step 1. Overall effect sizes were large (g = 1.62, SE = 0.32). This result was discarded due to its very large heterogeneity (I 2 = 95%).
Step 2. Significant differences were found between immediate and follow-up effects (d > 0.46, p < 0.001). Therefore, all the next metaanalytic steps were conducted separately for immediate and follow-up effects.
Step 3. Large heterogeneity was found, both immediate and at follow-up (respectively, I 2 = 92, 94%). Therefore, the outcome instruments were recategorized. Some 49 studies included quality-of-life instruments and 49 psychological stress instruments (i.e., 21 studies measured both meaning and stress and 39 described either meaning or stress; see Table 5). Metaanalyses showed large improvements from baseline measurement to immediate posttreatment and follow-up on quality of life (Hedges' g = 1.13, SE = 0.12; g = 0.99, SE = 0.20) and psychological stress (g = 1.21, SE = 0.10; g = 0.67, SE = 0.20). As analyses showed large between-study heterogeneity (I 2 > 90%) and positive nonsignificant publication bias, the outcome instruments were recategorized into more specific instrument categories. The subcategories for quality of life that had the least heterogeneity and smaller nonsignificant publication biases were: general quality of life; meaning in life; hope, hopelessness, and optimism; self-efficacy; and social relationships. The least heterogeneous subcategories for psychological stress were depression, anxiety, and existential anxiety (Table 2). Regardless of how the instruments were categorized, heterogeneity remained large (I 2 > 45%), suggesting that other moderators influenced heterogeneity. Therefore, the findings in this step are not further described (see Table 5).
Step 4. Controlled trials were significantly less effective than noncontrolled trials (d = 0.51, p < 0.01), but there were no significant differences between randomized and nonrandomized controlled trials (d = 0.11, p > 0.05). Thus, the effect sizes were significantly higher in noncontrolled studies: these effects are caused by the study characteristics and not by the MCT. To compensate for the effects of this artificial inflation and develop a more valid estimate of the true effect sizes, all further steps were only conducted in controlled trials that had low heterogeneity and high homogeneity. Compared with control groups, MCT had large effect sizes, immediately posttreatment and at follow-up, on quality of life (g = 1.02, SE = 0.06; g = 1.06, SE = 0.12) and psychological stress (g = 0.94, SE = 0.07, p < 0.01; g = 0.84, SE = 0.10). As the immediate effects on quality of life were heterogeneous, subcategories of outcome instruments were analyzed: the effects were larger on general quality of life (g = 1.37, SE = 0.12) than on meaning in life (g = 1.18, SE = 0.08); self-efficacy (g = 0.89, SE = 0.14); social well-being (g = 0.81, SE = 13); and hope, hopelessness, and optimism (g = 0.80, SE = 0.13). The effects on psychological stress were nonheterogeneous and therefore not further specified (see Table 5).
Steps 5–8. In controlled trials, moderation and metaregression analyses were not significant (p > 0.05) with respect to sample, treatment, therapy skills, and study characteristics. In the combination of all controlled and uncontrolled trials, nine MCT therapist skills were moderately strong moderators: studies had larger effect sizes when the treatment manuals did not include religious/spiritual formulations, were structured, explicitly stimulated clients to set and experiment with achievable goals in daily life, used mindfulness exercises, explicitly discussed one type of meaning per session, addressed self-worth, discussed existential limitations, mentioned the coherence of time, and focused on creating a positive therapeutic relationship (respectively, d = 0.47, 0.39, 0.36, 0.33, 0.32, 0.27, 0.26, 0.24, 0.23; all p < 0.01). Analyses of steps 1–6 with the 25% largest samples, published after 2000, and the 25% most precise effects showed similar effects as the main analyses and are therefore not further described.
Step 9. Changes in meaning and psychological stress were statistically heterogeneous (p(Q) = 0.67, I 2 > 75%). Changes in meaning in life predicted changes in psychological stress with strong negative effects (β = –0.56, VAF (Variance Accounted For) = 31.4%, SE = 0.11, p < 0.001; intercept β = 0.13, SE = 0.10, p = 0.09; N = 20) (see Table 6).
Table 5. Pre–post effects of meaning-centered therapies on quality of life and psychological stress in the combined group of all trials (k = 60): overall effects and effects for significant subgroups
Metaanalytic steps are explained in Table 2. All subgroups/moderators are significant with p < 0.05 and Cohen's d >.50; SE = standard error; 96^ CI = 95% confidence interval; I 2 test for heterogeneity in %; all Q values for heterogeneity are not significant except where indicated.
*(p < 0.05).
**Heterogeneity can be explained by significant differences between different outcome instruments (all moderators: Cohen's d ≥.50, p(d) < 0.01); therefore, the effect sizes for different groups of outcome instruments are presented in this table.
***Number of studies too small to interpret findings, based on a-priori power calculations.
Table 6. Effects of meaning-centered therapies on quality of life and psychological stress compared with control groups in randomized and nonrandomized controlled trials: overall effects and effects for significant subgroups
Metaanalytic steps are explained in Table 2.
All moderators are significant with p < 0.05 and Cohen's d > 0.50.
SE = standard error; 95% CI = 95% confidence interval; I 2 test for heterogeneity in %; all Q values for heterogeneity are not significant except when indicated.
*p < 0.05.
**Heterogeneity can be explained by significant differences between different outcome instruments (all moderators: Cohen's d ≥ 0.50, p(d) < 0.01).
Therefore, the effects sizes for different groups of outcome instruments are presented in this table.
***Sample size too small to interpret findings, based on a-priori power calculations
DISCUSSION
Our findings indicate that MCT largely improves the client's quality of life and reduces their level of psychological stress. Compared with control groups, MCT had larger effects on all outcome instruments. MCT was primarily effective in improving the general quality of life and meaning in life, secondarily in reducing psychological stress, and also in improving social relationships, self-efficacy, and hope/hopelessness/optimism. All positive effects were maintained between 4 and 12 months after the last therapy session.
To put our findings into perspective: these effects seem slightly larger than the moderate-to-large effects of excluded therapies that address meaning as one of many therapeutic aims and methods, such as broad positive psychology interventions, acceptance and commitment therapy, structured life review, and autobiographic writing (Chochinov et al., Reference Chochinov, Hack and Hassard2005; Seligman et al., Reference Seligman, Steen and Park2005; Sin & Lyubomirsky, Reference Sin and Lyubomirsky2009; Davis et al., Reference Davis, Morina and Powers2015; Ost, Reference Ost2014; Bohlmeijer et al., Reference Bohlmeijer, Smit and Cuijpers2003; Ando et al., Reference Ando, Morita and Akechi2010). Thus, our metaanalysis indicates that clients benefit from explicitly and systematically addressing meaning in life, either as a standalone MCT or as part of a complex treatment. The finding that the effects of MCT are slightly larger than the multiple-aims/multiple-methods interventions may suggest that therapies are more effective when they focus primarily and systematically on meaning.
MCT is based on the hypothesis that the client's level of psychological stress decreases because it helps them to experience life as more meaningful. Indeed, the treatment manuals described therapeutic techniques that directly addressed meaning in life, MCT clients experienced significant improvements in terms of meaning in life, and that these improvements correlated positively with decreased stress—it mediated stress. This is in concord with studies indicating that the meaning-centered coping skills taught in MCT predict better long-term well-being (Folkman, Reference Folkman2008; Park, Reference Park2010; Steger, Reference Steger and Wong2012).
These conclusions are based on the most robust studies and a categorization process selected via nine metaanalytic steps. Overall, effects were statistically larger in the short term as opposed to the long term, but the absolute differences were small and the effects large. The differences between instrument categories can be explained by the fact that quality of life and psychological stress are related but different phenomenological experiences. MCT also focuses primarily on improving quality of life, due to which the effects on quality of life are larger than those on psychological stress (Vos et al., Reference Vos, Asperen and Oosterwijk2013). The publications reported herein included trials from all of the continents (save Antarctica) with different client populations, although with an overrepresentation of physical illnesses (which did not significantly influence effect sizes). For example, more trials should be conducted in populations that were absent from the current analyses, such as Africa, South America, Australia, and Russia. Control groups mainly comprised psychological treatments. No differences were found between different types of treatments (e.g., care-as-usual practices or new standardized treatments). Most control groups are considered best practice or the gold standard in the field, which supports the ecological validity of the metaanalytic findings, as these are likely valid estimations of true effect sizes (Cuijpers et al., Reference Cuijpers, van Straaten and van Oppen2008). Moderator analyses showed that the heterogeneity of the included samples did not lead to statistical heterogeneous findings (the type of MCT was only slightly different in different samples, but the core clinical model was the same, so that there was relatively little clinical heterogeneity). This seems to confirm the hypothesis that MCT is equally effective across different populations.
The only significant moderators were eight therapeutic skills. The skills predicted larger effect sizes in the analyses of all noncontrolled, controlled, nonrandomized, and randomized studies taken together. These skills did not influence effect size in controlled trials, which can be explained by the fact that almost all controlled trials used these therapeutic skills, leaving little variation between the studies. In the light of this, MCT therapists may consider using nonreligious/spiritual formulations, structure, mindfulness, practical goal-setting exercises, focusing on one type of meaning per session, and addressing self-worth, existential limitations, coherence of time, and establishing positive therapeutic relationships.
In the light of the current findings, MCT can be seen as a bona fide intervention from which many clients appear to benefit. This warrants making MCT more widely available. Although effects were statistically similar in clients with primary mental or physical health concerns, from a clinical perspective MCT seems particularly relevant to individuals in transitional moments in life or with a physical disease, as many of them report meaning-centered concerns (Vos, Reference Vos and Russo-Netzer2016b). For example, the relevance for individuals with a chronic or life-threatening physical disease is underlined by metaanalyses of the 10 MCT trials, showing large effect sizes for physical well-being (Vos, Reference Vos and Russo-Netzer2016b). Although more cost-effectiveness studies are required, one randomized controlled trial indicated its cost-effectiveness (van der Spek et al., Reference van der Spek, Vos and van Uden-Kraan2014), and our metaanalysis found large effect sizes with a small number of sessions. MCT is therefore strongly recommended for inclusion in healthcare guidelines.
ACKNOWLEDGMENTS
The authors are grateful to the large consortium of colleagues who helped in a range of ways: by searching the literature in multiple languages, by sharing data for metaanalyses, and by proofreading.
