Publications from 2011 to 2016 were systematically selected. A literature search was conducted in the following digital databases: PsycINFO, PubMed, Proquest, Cochrane Library, Embase, Applied Social Sciences Index and Abstracts (ASSIA), Cumulative Index to Nursing and Allied Health Literature, Social Sciences Citation Index (SSCI) and EBSCO, China National Knowledge Infrastructure (CNKI) and WANFANG. Search words were the following: cancer/tumor, family/spouse/partner, caregiver/carers, psychosocial intervention/treatment/therapy, clinical trial/explanatory trial/pragmatic trial/randomized controlled trial/RCT. Search terms were the following: (cancer/tumor) AND (caregiver/carers/family/spouse/partners) AND (psychosocial intervention OR treatment/therapy) AND (clinical trial/explanatory trial OR pragmatic trial OR randomized controlled trial OR RCT), and search terms in different databases were slightly different.

A pair of independent raters selected abstracts for full review based on inclusion/exclusion criteria. Primary researcher resolved any discrepancies and produced final list of studied for full-text review.

All papers selected for final inclusion met the following criteria: (i) employed a psychosocial intervention aimed specifically to target cancer caregivers needs without a primary focus on the cancer patient; (ii) psychosocial health indicators including quality of life, depression or anxiety; (iii) RCT method adopted; (iv) written in Chinese or English; (v) publication of literature between 2011 and 2016.

A pair of raters reviewed relevant studies and extracted data respectively. The discrepancies were resolved by primary researcher and final data were entered into a database management system. In consideration of the heterogeneity of interventions and data, this research adopts Crocharane as the risk of bias (ROB) assessment tool to evaluate the overall quality of the research. The assessment includes six perspectives: selection bias (selected by random sequence or allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (comprehensive reporting) and other bias. Then, we scored every work according to these six aspects, with 2 points for “yes,” 1 point for “unknown” and 0 point for “no.” The total score was divided into three types, which stand for different levels of overall ROB: 0–4 points are for high risks, 5–8 points medium risks, and 9–12 points low risks. The report of systematic reviews is standardized by PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses).

Descriptive and explanatory data analyses were presented for a general picture of psychosocial intervention for cancer caregivers from 2011 and 2016. Effective size (Cohen d) were analyzed for each study measuring the difference in outcomes of quality of life, depression and anxiety when available data permitted.

A total sample of 3,604 cancer caregivers enrolled at baseline and the number of caregivers in each study ranged from 20 to 484 (see Table 1). Among all studies, the overall participant attrition rate was 23.7% and study attrition rate varied from 0% to 65.8%. The reasons for attrition include the worsening or death of patients, the business of cancer caregivers. Of the 22 identified studies, three did not provide sufficient data to compute an effective size (Meyers et al., 2011; Porter et al., 2011; Leow et al., 2015).

The main caregivers was identified as a spouse or adult child, parent, brother and sister, friend or significant other. Most of the main caregivers are spouses (69.68%). There are two studies specifically focusing on spouses (McLean et al., 2013) or partners (Badger et al., 2011; van de Hurk et al., 2015), one study focuses on young parents (Barrera et al., 2014). The mean age of the caregivers was 55.34. Most of the caregivers were women, accounting for 65.02%. Two studies were specifically for female caregivers' intervention (Badger et al., 2011; Lai et al., 2012). In terms of cancer types, there were four studies specifically for lung cancer patients (Badger et al., 2011; Porter et al., 2011; DuBenske et al., 2014; Badr et al., 2015; van de Hurk et al., 2015), one for glioma (Boele et al., 2013), one for hematologic malignant tumor (Hendrix et al., 2013). Furthermore, there were 15 articles focused on specific cancer stage of patients: newly diagnosed cancer (Barrera et al., 2014); advanced lung cancer (DuBenske et al., 2014; Badr et al., 2015); high-grade glioma (Boele et al., 2013); hematologic malignant tumor patients (Hendrix et al., 2013). The rest of the research consists of patients with different types of cancer.

In terms of intervention types, there were 11 studies of families-patients paired skills training (Badger et al., 2011; Meyers et al., 2011; Porter et al., 2011; Belgacem et al., 2013; Clark et al., 2013; Fegg et al., 2013; Hendrix et al., 2013, 2016; McLean et al., 2013; Northouse et al., 2013). What's more, two studies evaluated family demands (Mitchell et al., 2013; Barrera et al., 2014) and seven studies conducted individual skills training to caregivers (Belgacem et al., 2013; Boele et al., 2013; Dionne-Odom et al., 2015; Leow et al., 2015; Sun et al., 2015; Shaw et al., 2016). Additionally, there was another two interventions studies focused on couple-based intervention of marital relationships (McLean et al., 2013) and group skills training (Fegg et al., 2013).

With regard to the mode of intervention, it consisted of family connect intervention, including the assessment of caregivers' needs, family relationship maintenance, self-care of caregivers and so on (Shaw et al., 2016); FOCUS programme for promotion of family relationship and positive attitude, efficiency of solving problems, reduction of the uncertainty and symptom management (Northouse et al., 2013); intervention models based on self-determination theory (SDT), teaching self-care of caregivers, the patients' symptom management, relationship maintenance strategies and cognitive behavioral strategies to cope with depression and anxiety (Badr et al., 2015); using CBT mode to increase caregivers' care capacity (Boele et al., 2013; Clark et al., 2013); paired intervention of emotional focus therapy (EFT) (McLean et al., 2013), facilitation of communication and mutual support between patients and family members (Ledderer et al., 2013); education of symptom management and emotional stress reduction intervention (Porter et al., 2011; Lai et al., 2012; Sherwood et al., 2012; Belgacem et al., 2013; Hendrix et al., 2013, 2016; Leow et al., 2015; Sun et al., 2015). While symptom management referred to providing information about prevention of infection, control of pain, maintenance of nutrition, stress reduction intervention covered deep breathing, muscle relaxation and imagination of pleasure; comprehensive health enhancement support system (CHESS) model, which was a system for elevating the skills of family members of cancer patients based on a network, and it contained all kinds of information about cancers, served as a platform for communication and support between different family members and also provided feedback to family members (DuBenske et al., 2014); existential behavior therapy (EBT), including meditation, search for meaning, self-care and stress management, seeking personal value again and farewell (Fegg et al., 2013); telephone interpersonal counseling (TIP-C), which was based on the theory of interpersonal therapy plus cancer education (Badger et al., 2011); supportive talks and residential rehabilitation course (Ledderer et al., 2013) needs assessment tools like NAT-C and PAT (Mitchell et al., 2013; Barrera et al., 2014); problem behavior coping model (COPE) dealing with problems like body and mind control of patients, resources and management (Meyers et al., 2011); grief counseling (Dionne-Odom et al., 2015).

First of all, intervention format of included studies could be divided into three aspects: families-patients paired group intervention, caregivers' individual intervention and group intervention. Among them, 11 studies offered interventions to both patients and their family caregiver, 12 interventions were delivered solely to the caregiver and one provided group intervention. In terms of time and frequency of the intervention, most of them were comparatively regular, about once a week or biweekly, and the time span ranged from one and a half months to 2 years. In addition, relatively short interventions lasted 1–2 h or 2–3 h were also provided (Hendrix et al., 2013, 2016).

Regarding the practitioner of the interventions, eight were done by nurses (Porter et al., 2011; Lai et al., 2012; Sherwood et al., 2012; Belgacem et al., 2013; Hendrix et al., 2013, 2016; Ledderer et al., 2013; Northouse et al., 2013); six were conducted by professional psychologists, oncologists, behavioral therapists and trained health educators (Meyers et al., 2011; Boele et al., 2013; Fegg et al., 2013; McLean et al., 2013; Badr et al., 2015; Shaw et al., 2016); four were performed by mufti-disciplinary teams including professionals like psychotherapists, specialists and physical therapists (Clark et al., 2013; Mitchell et al., 2013; Barrera et al., 2014; Sun et al., 2015), one were led by social workers or nurses (Badger et al., 2011) and one was operated by family members through online distance learning (DuBenske et al., 2014).

In terms of the subject of research, eight studies measured quality of life (Meyers et al., 2011; Boele et al., 2013; Clark et al., 2013; Fegg et al., 2013; Ledderer et al., 2013; Northouse et al., 2013; Sun et al., 2015; Shaw et al., 2016), two studies examined depression (Sherwood et al., 2012; McLean et al., 2013), and another two studies took anxiety as the sole outcome variable (Porter et al., 2011; Barrera et al., 2014). In addition, three measured quality of life, depression and anxiety simultaneously (Fegg et al., 2013; Hendrix et al., 2013; Mitchell et al., 2013). Four examined both anxiety and depression (Lai et al., 2012; DuBenske et al., 2014; Badr et al., 2015; Hendrix et al., 2016), three took both quality of life and depression as the outcomes (Badger et al., 2011; Dionne-Odom et al., 2015; Leow et al., 2015).

Moreover, there was a lack of consistency in the measurement tools used to assess QoL, depression and anxiety. The measurement tool of QoL varied from the Caregiver Quality of Life Index-Cancer (n = 3); City of Hope QoL (n = 2); SF-12v2 (n = 2); SF-36 Health Survey (n = 1); the General Functional Assessment of Cancer Therapy Scale (FACT) (n = 1); Health Related Quality of Life (HRQOL) (n = 1); WHOQOL-BREF (n = 1); and Satisfaction with Life Scale (SWLS) (n = 1). In 10 of the 14 studies, a statistically significant improvement in general QoL was identified. In the remaining four studies, no significant change in QoL was observed. Of the nine studies where effective sizes could be calculated, three studies had effective size suggesting a small to medium effect of interventions on QoL outcomes (ranging from 0.27–0.62) (Cohen, 1988). In addition, six studies measured follow-up effect of intervention, and the time point varied from 4 weeks to 12 months. Four studies demonstrated a nil effect of treatment on QoL with Cohen d ranging from 0.07 to 0.18. The remaining two studies had effect sizes suggesting a small (0.21–0.25) and medium (0.43–0.52) effect of treatment on QoL respectively.

With respect to anxiety, the measurement tools varied from Profile of Mood State (POMS-B) (n = 3); Hospital Anxiety and Depression Scale (HADS) (n = 2); State Trait Anxiety Inventory (STAI) (n = 2); Brief Symptom Inventory (BSI) (n = 1); Patient Reported Outcomes Measurement Information System (PROMIS) (n = 1). In seven of the nine studies, a statistically significant improvement in anxiety was identified. In the remaining two studies, no significant change in anxiety was observed. Of the six studies where effective sizes could be calculated, two studies had effective size suggesting nil to small effect of interventions on anxiety (ranging from 0.11 to 0.41) and another two studies had high effect of interventions on anxiety with estimated d ranging from 1.12 to 1.3 (Cohen, 1988). Moreover, two studies measured follow-up effect of intervention, and the time point varied from 1 to 8 months. Both studies demonstrated a nil to small effect of treatment on anxiety with Cohen d ranging from 0.04 to 0.44.

As for depression, measurement tools consisted of Center for Epidemiology Studies-Depression Scale (CES-D) (n = 4); Hospital Anxiety and Depression Scale (HADS) (n = 2); Profile of Mood State (POMS-B) (n = 2); Brief Symptom Inventory (BSI) (n = 1); Beck Depression Inventory (BDI) (n = 1); Patient Reported Outcomes Measurement Information System (PROMIS) (n = 1); and self-developed depression scale (n = 1). In eight of the 13 studies, a statistically significant improvement in depression was identified. In the remaining five studies, no significant change in anxiety was observed. Of the seven studies where effective sizes could be calculated, two studies had effective size suggesting small effect of interventions on anxiety (ranging from 0.27 to 0.39) and another study assessed depression demonstrating a large effective size (d = 1.8) (Cohen, 1988). Additionally, four studies measured follow-up effect of intervention, and the time point varied from 6 weeks to 12 months. Four studies demonstrated a nil to small effect of treatment on anxiety with Cohen d ranging from 0.04 to 0.47 (Cohen, 1988).

As was shown in Table 2, among the 21 interventions, 17 had shown effects in psychological health on at least one dimension. group training was effective in improving quality of life, easing anxiety and depression (Fegg et al., 2013). Furthermore, paired-intervention of self-care skills to cancer caregivers and symptom management to patients (Badr et al., 2015) and individual training of stress coping skills (Lai et al., 2012; DuBenske et al., 2014; Leow et al., 2015) Assessment of the families' demand (Mitchell et al., 2013), education of symptom management, stress coping and problem-solving (Meyers et al., 2011; Belgacem et al., 2013; Boele et al., 2013; Northouse et al., 2013) were specifically effective in improving quality of life.

With regard to the practitioners of the interventions, six of the interventions conducted by psychological consultants, researchers, health educators and oncologists were effective (Badger et al., 2011; Meyers et al., 2011; Boele et al., 2013; Fegg et al., 2013; McLean et al., 2013; Badr et al., 2015), four of the interventions led by nurses were effective (Porter et al., 2011; Belgacem et al., 2013; Northouse et al., 2013; Barrera et al., 2014), one of the interventions carried out by social workers or nurses was effective (Barrera et al., 2014). In addition, four of the interventions operated by the multidisciplinary teams composed of oncologists, radiotherapists, psychological consultants, physiotherapists and physical therapists, physicians and nurses and social workers were effective (Clark et al., 2013; Mitchell et al., 2013; Barrera et al., 2014; Sun et al., 2015). Moreover, self-help e-learning by caregivers was also proved to be effective (DuBenske et al., 2014).

Among current interventions, 91% applied RCT, two of which were quasi-experimental design (DuBenske et al., 2014; Sun et al., 2015). In terms of the measures of comparative interventions, there were three kinds of compared groups: 13 interventions adopted the normal mode, four adopted controlling, four were used for information provision and one is made pending (Dionne-Odom et al., 2015). Furthermore, from 2 weeks after the baseline assessment to 1 year after the baseline assessment, 11 interventions got two follow-up visits, and five got three to four; one included only pre and post-tests, and one got five follow-up visits. Additionally, eight interventions included post-tests while the other 14 interventions included none and only collect data from the follow-up visits.

Crocharane, the ROB assessment tool, was used to evaluate the overall ROB. The result showed that only one intervention was assessed with high ROB (Ledderer et al., 2013), while 12 interventions had middle ROB and nine had low ROB (see Table 2). The one with high ROB would not be discussed further.

This review took only three indexes into consideration, which are caregivers' quality of life, degree of depression and degree of anxiety. Thus, there could be selection bias. In addition, quality of life is a broad concept without consistent definition, and no one definition was used in current systematic review, which limits the generalizability of the results of this review. At the same time, the cancer types and stages as well as intervention mode and length varied widely in this review, making it difficult to compare the effectiveness of interventions across included studies. Furthermore, this study only reviewed research in English and Chinese, which could also limit the generalizability of the results. Reviewing only the quantitative research could also increase selection bias. Furthermore, most research was conducted in developed countries such as European countries and the United States. Due to the differences in both culture and policy, whether or how the conclusions could be extended in other culture contexts needs further confirmation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.