Selection results

The flow chart of the study selection process, shown in Fig. 1, was based on PRISMA. We found 6,756 articles in the databases and 769 in the gray literature. After an independent two-step selection by the two reviewers, eight articles were finally included. We extracted the main characteristics of these eligible studies and summarized them (Table  1) according to the Cochrane Handbook [29].

Table 1 Characteristics of included studiesTypes of participants

In total, 721 participants from four countries, namely China (n = 316), the USA (n = 341), Sweden (n = 42), and Korea (n = 24), were included in this review. However, the 42 Swedish participants came from the same study, which was reported in two separate articles [31, 32]. While our strategy was not to select participants with one specific type of cancer, participants from seven of the eight studies were all people with breast cancer (n = 593), and another sample consisted of patients with a variety of types of cancer. Furthermore, only one study from China included men [33] (n = 41). Participants in all the other studies were female. Notably, participants were not co-diagnosed with other disease such as cancer with a mental health co-diagnosis.

Types of intervention

Table 1 shows the details of the interventions. The intervention methods of the included studies were primarily based on art therapy in group or individual sessions. The therapies lasted from 4 to 12 weeks, and almost all intervention instructors were qualified art therapists. However, in China, a system for qualification authentication for art therapists has not been established. In the Chinese studies, counseling psychologists or psychotherapists were recruited to conduct art therapy. The control group of some studies comprised a wait-list art therapy group who were receiving a regular care intervention. However, only one art therapy course in China consisted of two sessions per week [33], while the others included interventions once a week. In addition, Monti et al. [28, 34] used a type of art therapy called mindfulness-based art therapy, which combined art therapy with mindfulness therapy in intervention studies conducted in 2006 and 2013. The control group received a delayed intervention in 2006 and mindfulness-based stress reduction therapy in 2013.

Outcome measures

The main indicator was quality of life and its sub-dimensions (psychological health and physical health). These variables were assessed using the European Organization for Research and Treatment of Cancer Quality of Life Questionnaires, the Short-Form Health Survey, and the World Health Organization Quality of Life Questionnaires. Perceived emotional and somatic symptoms were primarily measured by the Symptoms Check List-90 (anxiety, depression, and fatigue), the Profile of Mood States (depression and fatigue), and the Personality Assessment Inventory (depression and anxiety). Table  1 contains detailed information about the included studies.

Risk-of-bias in the included studies

Figure 2 shows the risk-of-bias assessment for the eight included studies. All studies were RCTs, which should have been conducted with enough rigor to be able to apply the randomized assignment sequence method. However, some studies did not report the randomization generation method [28, 34, 35]. Although a random allocation square can conceal imperfections in the study and could break predetermined and predictable allocation sequences [36], only two of the eight studies took note of the concealment of a random allocation square [31, 32]. Blindness makes art therapy impossible for participants. Therefore, all included studies were evaluated to indicate a low risk of blindness among participants. While it was feasible to perform an outcome assessment of blindness, almost no studies mentioned blindness in the outcome assessment or data processing.

Fig. 2

The risk-of-bias assessment for each of the included studies

It is understandable that in intervention studies, a certain number of participants drop out. However, YuQiao et al. [37] recruited 230 participants in their study, and no participants dropped out. Such a perfect follow-up of participants is questionable. The Cochrane Handbook [29] states that if there are missing data, the authors must explain in detail the reason for the discrepancy and how they dealt with it when preparing a paper for publication. The majority of the studies included in our review did not perform well concerning attrition bias.

No studies tended to report data selectively. With regard to other biases, the detailed baseline information of most studies was not reported. However, the demographic data of the participants were compared. The outcome indicators of two studies [31, 38] were not compared at baseline. For example, Thyme et al. [31] stated that the outcomes of the two groups were approximately the same at baseline, but the P-value was not reported in the main text.

Quality grade of the included studies

With respect to the quality of the included RCTs in GRADE, only two studies [31, 32] performed well in random sequence generation and allocation concealment and were considered high-quality evidence. YuQiao et al.’s [37] evidence was graded as low quality due to the high risk of other bias, unclear selection, and detection bias. Monti et al.’s [28] study, which provided moderate-quality evidence, was not included in the meta-analysis because the control group (standardized mindfulness-based stress reduction) was different from those in the other studies (wait-list or usual care), and the data were incomplete. The RCT by Jang et al. [35] provided moderate-quality evidence given its unclear selection bias, even though they noted that their outcome measurement assessment was blinded. The remaining three papers were also considered to present moderate-quality evidence. Overall, the grade quality of the studies included in our review was rated as moderate.

Effect of art therapyQuality of lifeOverall quality of life

Figure 3 shows a random-effect meta-analysis of the impacts on the overall quality of life after art therapy intervention. We found a significant effect of art therapy on quality of life, with a large effect size (SMD = 1.87; 95% CI = 0.47 to 3.28; p = 0.009). A high heterogeneity between studies was also found (x2 = 38.89; p < 0.00; I2 = 92%).

Fig. 3

Effect of art therapy on overall quality of life in people with cancer

Sub-dimension of quality of life: psychological health

Referring to the “psychological health” sub-dimension of quality of life, the results revealed no significant difference between the intervention and control groups (pooled effect SMD = 0.68; 95% CI = − 0.28 to 1.64; p = 0.16; Fig. 4). The heterogeneity of three studies (x2 = 2.73, p < 0.00, I2 = 94%) may have come from sampling errors or the use of different measurement instruments.

Fig. 4

Effect of art therapy on psychological health in people with cancer

Sub-dimension of quality of life: physical health

Referring to physical health, which is another sub-dimension of quality of life, we found that art therapy had no significant effect on the physical health of women with cancer (SMD = 0.16; 95% CI = − 0.04 to 0.36; p = 0.12; Fig. 5); heterogeneity between studies was found (x2 = 1.16; p < 0.00; I2 = 0%; Fig. 5).

Fig. 5

Effect of art therapy on physical health in people with cancer

Psychosomatic symptomsAnxiety

The overall effect size of intervention on anxiety was − 1.08 (95% CI = − 1.96 to − 0.19; p < 0.00; Fig. 6). This significant result indicated that art therapy could moderately reduce anxiety in patients with cancer, especially those with breast cancer. There was heterogeneity among studies (x2 = 45.27; p < 0.1; I2 = 91%). The sensitivity analysis indicated that after removing the studies by Jang et al. [35] and ShuFen et al. [33], the heterogeneity among studies reduced dramatically to 0%, but the significance of art therapy overall was unaffected (SMD = − 0.30; 95% CI = − 0.58 to − 0.01; p = 0.04), resulting in moderate-quality evidence and small effect size.

Fig. 6

Effect of art therapy on anxiety in people with cancer. (a) Subgroup analysis; (b) Sensitivity analysis

Depression

Four studies generated a pooled effect size (SMD = − 0.75; 95% CI = − 1.40 to − 0.11; p = 0.02; Fig. 7), whereby art therapy was found to reduce depression in women with cancer. Heterogeneity was observed among studies (x2 = 13.41; p < 0.00; I2 = 78%). The sensitivity analysis indicated that after the removal of Jang et al.’s (2016) study, the heterogeneity among studies reduced from 78 to 15%. This also affected the significance of art therapy overall (SMD = − 0.39; 95% CI = − 0.68 to − 0.11; p = 0.007), which resulted in moderate-quality evidence and small effect size.

Fig. 7

Effect of art therapy on depression in people with cancer. (a) Subgroup analysis; (b) Sensitivity analysis

Somatic symptoms

As Fig. 8a shows, art therapy had no significant effect on somatic symptoms (SMD = − 0.06; 95% CI = − 0.27 to 0.15; p = 0.58), and no heterogeneity was found between the three studies (x2 = 1.18; p = 0.55; I2 = 0%). Two studies that included women with cancer and fatigue were included in our analysis, which suggests that there was no significant effect of the intervention on fatigue (SMD = − 1.54; 95% CI = − 4.06 to 0.98; p = 0.23; Fig. 8b). Heterogeneity among studies was found (x2 = 13.89; p < 0.00; I2 = 93%). Three studies found that the intervention reduced pain (SMD = − 0.52 (Fig. 8c and 95% CI = − 4.06 to 0.98, p = 0.17), which did not achieve a significant effect with heterogeneity (x2 = 16.17; p < 0.00; I2 = 88%).

Fig. 8

Effect of art therapy on somatic symptoms, fatigue, and pain in people with cancer. (a) Somatic symptoms; (b) Fatigue; (c) Pain