In psychotherapy, an ambiguity toward bodily experiences becomes apparent that stems from the phenomenological consideration of the body as subject as opposed to the body as object: whereas the former means the experience of the body as constituting the first-person perspective from which an individual encounters the world, the latter refers to the objectified body that natural science examines on different levels (biochemical, neuronal, molecular, etc.) (11–13). Although this epistemological distinction might be neglected in psychotherapy, its consideration may lead to a gain in explanatory power from the possible phenomenological change between both perspectives involved. Likewise, Gallagher (14) differentiates body schema and body image as opposed stances of an individual. While the former describes the implicit, pre-reflexive role of the body as an inevitable precondition for action, the latter summarizes perceptions and evaluations of the body from an intentional point of view. For others, bodily experiences constitute a spectrum from body-related perceptions and emotions to the evaluation of the own body (15).

In case of pathology, Weizsäcker's model of the “Gestaltkreis” (16) offers an explanation of the invisibility of the psychological conflict of an individual as somatic symptoms replace it. Thus, it is the task of psychotherapy to clarify this shift by means of the complementary perspectives of client and therapist on the client's body. Whereas, the client has a private sphere no one else can enter, the therapist contributes his impression of the client's body, hence an outer perception, interpretation, and own reaction to it. The implied discrepancy between the perspectives is the precondition for the revelation of the unconscious in the intersubjective process (17, 18). In this sense, psychotherapy strives to find a functional explanation and solution in which thought, feeling, body experience, and expression can be intertwined (19). For example, “functional relaxation” understands breathing exercises as a semiotic process, in which the therapist contains the bodily signals of a client, interprets them as his own, and verbalizes the result (20, 21).

Due to its application in diverse and theoretically opposed psychotherapeutic streams, a common definition of BPT is not easy to state. Geuter (3) proposes to define BPT as a treatment employing psychological and bodily means equally. For him, the specificity of BPT lies in the continuous combination of both channels and the interaction of client and therapist on the body level. Loew et al. (22) differentiate body therapy and BPT. The former technique focuses on the experienced body to increase the person's physical and psychological well-being. BPT fulfills, in addition, the standards of a verbal psychotherapy following Strotzka (23), which demand a predefined and planned interactional process based on a theoretical foundation, which has clearly defined aims in the treatment of conditions of significant mental suffering or mental diagnoses. Furthermore, it implies verbal and non-verbal psychological means that can be taught and requires a stable emotional relationship between client and therapist. Likewise, for Wampold (24), psychotherapy is foremost an interpersonal treatment. For the purpose of this systematic review and meta-analysis, we hold on the definition of Loew et al. (22) which emphasizes the intersubjectivity of the body experience as the constituting element of BPT.

As mechanisms of change, Lahmann and Weise (25) presume a reduction of arousal and an amelioration of interoception for bodily signals resulting in an enhanced self-awareness and self-efficacy. In order to achieve a better understanding of the bio-psychosocial model, they propose to employ bodily techniques especially in patients presenting with somatic symptoms. Likewise, Röhricht (10) suggests a wide spectrum of application—from affective disorders, eating disorders, somatoform disorders to schizophrenia, i.e., he states that BPT “uniquely modulate[s] emotional processing, affect regulation, movement behavior, and bodily self-awareness.”

From a theoretical point of view, BPT fits well into the embodied mind paradigm (26) that nowadays prevails in cognitive science. According to this stance, human cognition cannot be explained only as a function of a brain but as an interplay of a whole body and its environment (14, 27). Shapiro (28) stresses that information processing begins in the sensorimotor periphery. From this inevitable evolutionary perspective, human experience and behavior should be understood by their genesis and embedment. Likewise, Petzold and Sieper (29) propose a hermeneutic point of view on BPT in which the informed body contains the mental history and state of a person. A historical consideration offers enriching ideas of implied mechanisms of change. In addition to the associative work psychoanalysis contributed to BPT, bioenergetics work with energetic cycles to enhance and finally reduce tension (2, 30) when solving a conflict. More recent approaches extend the technique of transference and countertransference on the body level of communication (31, 32). With regard to the other root of BPT, the gymnastic movement of the twentieth century stressed the healing power of focusing on detailed experience and rhythms of the body and repetition. Nowadays, an integrative body–mind–spirit model assumes, too, the interconnectedness of body and mind and include spirituality as an existential human domain (33). Such connections can be assessed and observed for, e.g., “grounding,” a mechanism commonly agreed in bodily psychotherapy and dance and movement therapy (34). It remains unclear whether such approaches address psychotherapy rather than aiming at multicomponent lifestyle modification programs (35). Röhricht (36) and Geuter (3) propose different grids to structure and integrate therapeutic streams under the umbrella of BPT. Along with this claim, arguments were brought up (37, 38) to introduce a summarizing work factor “body experience” in psychotherapy along with the five factors identified by Grawe (39).

There are few systematic reviews on BPT that mainly focus on distinct therapies such as mindfulness-based stress-reduction therapy (40, 41) and dance therapy (42). This seems to be unsatisfying as those—more active—treatment approaches assembled under the umbrella term BPT are preferable for a broad range of functional complaints including bodily symptoms and psychic suffering (43). Among these few reviews, Grossman et al. (40) draw an ambivalent picture of BPT. May (44), who included also studies with non-clinical samples and “gray” literature, reports positive results in half of effectiveness and efficacy studies and also stresses negative results. Loew et al. (22) report positive effects on symptoms, body perception, and social behavior in eight studies that employed integrated different methods of BPT. For functional relaxation, especially in the treatment of asthma, they found eight controlled, partly randomized studies that state therapeutic success in self-evaluation and physiological parameters. For concentrative movement therapy, Seidler (45) remarks that there are only a few studies; three quasi-experimental studies imply that this method increases well-being and self-consciousness. Röhricht (10) states an increase in evidence and describes a disorder specific benefit of BPT, namely, on schizophrenia. However, he criticizes retrospective studies, small samples, and a lack of RCTs. The, to our knowledge, most recent review, which captures a wide range of body therapies, states positive results for studies offering BPT (46).

To the best of our knowledge, none of the above-mentioned reviews conducted a meta-analysis on RCTs testing BPT. Accordingly, we identified and included all interventions that integrated body techniques into psychotherapy in the way outlined above. We addressed a broad range of health conditions, defined as “significant psychic suffering,” in order to determine the efficacy of BPT for a wide population of patients.

In this systematic review and meta-analysis, we included intervention trials that (1) addressed intersubjective body experience in the above-described manner, enframed in psychotherapeutic measures; (2) had a two-factorial design (two groups, measures pre- and post-treatment) with randomized group allocation; (3) publications had to indicate descriptive statistics such as mean and standard deviation or frequencies suitable for the determination of effect sizes; (4) had to be written in English or German and their abstract had to be available via scientific data bases; and (5) the included participants had to be adult (18+ years) and required a diagnosed psychiatric disorder or significant psychic suffering proven by reliable measures (e.g., cutoff values of validated psychometric scales, interviews or a thorough clinical evaluation). There were no restrictions regarding setting (outpatient or inpatient), format (individual or group), dose, and quality of activity applied in the control group.

Relevant articles were identified by searching PubMed (MEDLINE), PsycInfo, and PSYNDEX on November 4, 2018. We cross-referenced [(body OR bodily) psychotherap^*] with (clinical trial OR random^* OR empirical research OR evaluation studies). In addition, reference lists of the identified review articles and of the European Body Psychotherapy Association were scrutinized to include further studies. An updated hand search until April 30, 2021 did not detect recent relevant studies. Study selection was carried out by SR and HS, applying a short checklist for the abovementioned criteria. Doubtful cases were discussed between the authors and discrepancies solved by Delphi method, if necessary, under consultation of CL.

As an indicator of the studies' quality, we used the Cochrane Collaboration Depression, Anxiety, and Neurosis (CCDAN) quality assessment rating scale (47), which allows a detailed description of the intervention, of inclusion criteria, and of the study execution, resulting in a quantitative measure of the study quality (Supplementary Table 1). Independent ratings were simultaneously carried out by SR and HS.

To evaluate the effectiveness of the intervention, we calculated effect sizes [standardized mean differences (SMD), Hedges' g] based on the random-effects model between groups on the primary and secondary outcomes. In this way, we accounted for the plausible variability of effects among studies. In case an outcome of interest was reported by more than one scale within a study, we integrated those using z-transformation to avoid multiple testing (48). Per convention, the effect size was interpreted as small if 0.2 ≤ |ES| < 0.5, medium effect if 0.5 ≥ |ES| < 0.8, and large effect if |ES| ≥ 0.8 (49). An effect size higher than 0.5 was judged as clinically relevant (50). We estimated within group changes as well to analyze the absolute effect of an intervention and to exclude that between-group effects were not based on a deterioration of the symptomatology of the control group.

As primary outcomes, we chose measures of psychopathology and psychological distress. Whereas, the former depicts the main characteristic(s) of the psychic suffering, the latter describes the overall functional impairment due to mental or bodily symptoms. As secondary outcomes, measures of coping abilities, quality of life, body experience, and interpersonal difficulties were analyzed. Confidence intervals (95%) of effect sizes were estimated for significance tests using the Z-statistic (α = 0.05 when a difference between groups is predicted and α = 0.20 when the equivalence of groups is predicted) (48). Positive values of within-effect sizes represent an observed improvement concerning the regarded dimension, with the exception of quality of life and body experience. The time points of post-intervention measurement and, consequently, the intervals between pre-intervention measuring and follow-up differed. Tests of heterogeneity were carried out using the Q-Statistic (relevant heterogeneity was assumed for p ≤ 0.20). As a quantitative measure, I² indicates the percentage of variance that is attributable to study heterogeneity.

Subgroup analyses were carried out in case of significant heterogeneity for these dimensions: year of publication, sample size, diagnosis, bodily active control, and dose of therapy. Data management and statistical analyses were executed with the open-source online platform Covidence and the Review Manager 5.3 of the Cochrane Collaboration.

Concerning dropout, as a convention, sample size was fixed to the smallest available number of participants per dimension. For categorial data, odds ratios (and their SE) were calculated and transformed into SMD (and corresponding SE).

The literature search yielded 2,657 publications in three databases, whereof 2,180 remained after duplicates were removed. From these, 84 did meet the inclusion criteria. Another 32 records were identified by additional measures, whereof 3 records were not attainable in full text. One hundred thirteen publications were scrutinized in detail. Finally, 23 records from 18 studies were selected for meta-analysis (see Figure 1). Eight records could not be included, as they lacked important information for integration and analysis or offered physiological data only.

The characteristics of the included studies are summarized in Table 1. The whole sample consisted of 1,297 patients with a range of sample size over studies between 15 and 275 participants. Concerning type of intervention, three studies employed functional relaxation, one employed bio-energetics, five followed a manual on BPT (36), three followed a mindfulness-based protocol, two followed a common affect-focused root, three had background of psychomotor therapy, and one had an idiosyncratic approach (54). All but four studies applied BPT as a group therapy, often as a part of (multimodal) training programs (e.g., as adjunct therapy to psychotherapy and/or psychoeducation).

Changes in psychopathology as the most frequently examined outcome dimension were reported for 18 studies (Figure 2A). The majority of those studies (16, 88.9%) suggests that BPT—in comparison to the respective control condition—brings about an improvement in psychopathology, which was statistically significant in eight studies. No improvement at all was observed in two studies applying an equally dosed physical activity for the control group (54, 62). The total effect size is medium {g = 0.56 [95% CI (0.35, 0.77)], Z = 5.33, p < 0.001} with a moderate to high degree of heterogeneity [τ² = 0.11, χ2(17) = 44.97, p < 0.001, I² = 62%].

For psychological distress (Figure 2B), included studies demonstrated an—at least—small advantage of BPT as compared to control groups, which results in a moderate total effect [g = 0.52, 95% CI (0.27, 0.78), Z = 3.99, p < 0.001], again with moderate to high heterogeneity [τ² = 0.13, χ(11)2 = 34.06, p < 0.001, I² = 68%]. Moreover, the corresponding within-group effects on the primary outcomes are statistically significant (for details, contact authors).

Data on coping abilities, quality of life, body experience, and interpersonal difficulties were available for a restricted number of studies only (see Figures 3A–D). On coping, five studies showed a moderate to high homogeneous total effect size for BPT, which was statistically significant [g = 0.68, 95% CI (0.43, 0.92), Z = 5.42, p < 0.001]. As the effects on primary outcomes, this effect relies on a more pronounced amelioration of the intervention groups. At the same time, an effect of BPT on quality of life could not be demonstrated [g = −0.09, 95% CI (−0.26, 0.08), Z = 1.06, p = 0.29, n. s.] in seven included studies, although five of these studies revealed an at least slight improvement (represented by a numerically negative effect) in this dimension. In four studies reporting data for body experience, an improvement for BPT could not be demonstrated [g = 0, 95% CI (−0.41, 0.41), Z = 0, p = 1.00, n. s.]. As to interpersonal conflicts, the four analyzed studies suggested a small effect size, indicating a benefit of BPT compared to CG bordering statistical significance [g = 0.32, 95% CI (0, 0.65), Z = 1.95, p = 0.05, n. s.].

Post-hoc, we performed subgroup analyses on within-group effects in primary outcomes (see Figures 4A,B). Categorial moderators divided studies naturally into two groups, whereas median splits were applied for continuous moderators based on all studies. Due to the exploratory character of these analyses, we did not interpret tests of significance. It should be noted that we did not investigate interdependence of moderators.

Studies published before 2013 had a slightly higher effect than those published later on both primary outcomes. Concerning sample size, results were inconsistent across primary outcomes. With regard to diagnosis, 12 studies allowed a distinction between effects for patients with affective and somatoform disorders as compared to those with schizophrenia, with smaller and more heterogenous effects for the latter. In studies with an active control group, the advantage of BPT diminishes, as compared with inactive CGs (e.g., waiting list). As to dose, a smaller number of intervention hours corresponded to a larger effect.

To estimate long-term stability of the effects, we integrated catamnestic data. On average, follow-up was conducted 7 months (range, 3–13) post-intervention. For psychopathology, the long-term total effect-size slightly decreased to a small to medium effect [g = 0.41, 95% CI (0.14; 0.67), Z = 2.99, p < 0.01] with 10 included studies [n (BPT) = 384, n (CG) = 369]. However, for psychological distress, the effect remained quite stable with g = 0.47 [CI 95% (0.22; 0.72), Z = 3.75, p < 0.001] induced from eight studies [n (BPT) = 319, n (CG) = 306], with both effects statistically significant.

In order to account for possible publication bias, we generated funnel plots on primary outcomes (see Supplementary Figures 1A,B). As for psychopathology, studies gather symmetrically around the estimated common effect. However, as the supposed triangle is missing its top, there is a lack of studies with bigger sample sizes. The distribution of studies for psychological distress shows an asymmetrical form, suggesting a loss of smaller studies with more pronounced effects.

The effect size on primary outcomes corresponds to an NNT of 4 (24) that is comparable to the one found by Koch et al. (42) for dance and movement therapy (g = 0.34–0.46) on symptomatology. In comparison, for mindfulness-based interventions (41), an effect of 0.73 was found on depressive symptoms. All effects integrated in our study remain smaller than the commonly assumed effect of psychotherapy (0.80) when compared to a control without any treatment. A possible explanation with regard to heterogeneity is that, indeed, different effects, respectively, treatments and comparisons, are summarized. This counts as well for study designs—additive designs investigating BPT as a part of an already established multimodal treatment will not be likely to reveal large effect sizes. Heterogeneity affects within-group CG effects as well. Accordingly, the extensive variability of the chosen CGs is underlined by the corresponding subgroup analyses.

Subgroup analyses could be performed for common factors reported by all or at least most included studies. Study size influences both primary outcomes differently, as the effect decreases with increasing study size only on psychopathology, not on psychological distress. This might be seen as an argument against publication bias.

Concerning diagnosis, the diminished effect in schizophrenia seems to be common for different types of psychotherapy (70–72), which reflects, on the one hand, the general difficulty to treat this disorder. On the other hand, it underlines the need to develop diagnosis-specific therapeutic techniques, as an effect of BPT on positive symptoms could not have been shown so far (64, 67). The result that, compared to men, women might benefit from BPT with regard to body expressiveness (73), hints to the need for individualized therapies.

As to study design, comparisons with CGs that followed a physical activity (e.g., Pilates, massage) led to smaller between-group effects of BPT on primary outcomes than comparisons with inactive controls. Because of the small number of studies, we did not differentiate between exercise only and body therapy (techniques that focus on the experienced body to increase the person's physical and psychological well-being) as control conditions. A medium effect of exercise could be found on depressive symptoms (74); yoga might even have a small advantage compared to exercise and be—combined with pharmacotherapy—as helpful as psychotherapy (75). Interestingly, the absolute size of within-group effects varies strongly between primary outcomes when applying this moderator, which calls for further studies to clarify its role. A dependence of effect on activity of control group was found for mindfulness-based interventions (41) and poses a known threat to the proof of specific effectiveness of psychotherapy (76, 77). The ongoing debate about the impossibility of constructing an inert placebo stresses the importance of establishing common work factors (78) as might be embodiment. Further on, the interpersonal work on body experience, if fitted to specific disorders, could be an additive that qualifies a psychotherapy as opposed to body therapies without this component.

Surprisingly, BPT interventions with smaller dose (hours of therapy) produced larger effects than those with higher dose, an observation that cannot be confirmed by other meta-analyses (76, 79). As a confounder, the difference in periods between pre- and post-treatment measurements has to be taken into account. While low-dose interventions typically lasted 5–12 weeks, high-dose interventions took 8–60 weeks. Consequently, post-treatment measuring of the former might benefit from the natural course of mental disorders. For example, an untreated depressive episode usually lasts ~2–3 months (80), so that remission might fall together with post-treatment measuring of low-dose interventions. On the other hand, the diminished effect of high-dose and longer lasting interventions could mirror the delayed occurrence of an aversive reaction when applying BPT that might be a necessary step in a healing process (81).

Secondary outcomes were tested to analyze potential mediators of BPT. The large, homogeneous effect on coping abilities should not be granted as a general effect because of the small number of studies that it relies on. Nevertheless, this finding might be seen as a positive indicator that BPT targets emotion regulation. While other meta-analytic data on coping is rare, an evaluation on psychosocial group interventions that addressed coping reported no amelioration (82). Consequently, there is a need of integrating process variables in meta-analytic research, including a differentiation of positive and negative coping styles (83, 84).

Considering quality of life, both groups benefited equally, which contradicts meta-analyses on other types of psychotherapy (85). When comparing different diagnoses, it is compelling that in the present review, studies concerning depression and anxiety disorders typically report a more pronounced amelioration than those with schizophrenia.

Neither did we find a difference concerning body experience, where comparable evidence is rare as well. In this regard, the distinction between body schema and body image by Gallagher (14) might be helpful. Thus, on a theoretical level, it might not be adequate to combine scales that measure each just one aspect, as they may follow different rationales and/or time courses, for example. However, for interpersonal problems, we found a slight advantage of BPT that bordered significance and was comparable to the effects of other psychotherapies (42, 86).

General concerns regarding meta-analysis potentially limit the usefulness of the applied meta-analytic approach when informing patient care (87). In particular, a risk of bias in this meta-analysis remains. Indeed, the respective funnel plots did not indicate absence of any bias for each outcome. These findings underline the need for high-quality studies with bigger sample sizes to estimate the effect more precisely. As to study identification, the search term might be criticized because it favors studies of a certain therapy tradition by employing typical vocabulary. Consequently, it cannot be ruled out that types of therapy that are similar in content but use a different wording are neglected. In addition, plausible moderators should be entered in a multiple regression to address possible interactions. While the overall number should be reduced, it might be interesting to differentiate between dose and length and consider the severity of disorder. Finally, the scrutinized evidence was too small to statistically test potential process variables and moderators.

The evidence identified in this study suggests that BPT often was applied as adjunct therapy rather than an alternative to single psychotherapy. Accordingly, BPT often was examined in (multimodal) treatment programs for following an additive research design. This issue might result in some benefit for external validity. Moreover, the most commonly used group setting may improve efficacy and economical implementation of these approaches.