Effectiveness of family psychosocial intervention on mental health and family function of caregivers of children with cancer: a meta-analysis^†

A comprehensive search of published articles in international databases (PubMed, Web of Science, MEDLINE, Embase, Cochrane Library, and PsycARTICLES) and Chinese databases (CNKI, VIP, Wanfang, and CBM) was conducted, spanning from database inception until September, 2021. The search terms included the following subject headings and keywords: (family intervention OR family-based intervention OR family-centred intervention OR family psychosocial intervention OR family psychoeducation OR family therapy) AND (adolescent* OR teen* OR youth* OR child* OR infant* OR preschool OR minor*) AND (neoplasm* OR cancer* OR tumour* OR leukaemia). The strategy used to search the Web of Science is presented as an example in Table 1. A manual search of the related references was also performed to identify potentially relevant trials. The meta-analysis was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, and the protocol was registered with PROSPERO (CRD42022308181).

Articles that met the following inclusion criteria were selected for the subsequent analysis: (1) focused on caregivers of children with cancer; (2) included randomized participants in a family-based psychosocial intervention condition (interventions aiming to change the thoughts or behaviors of the family as a whole, such as family psychoeducation, family therapy, or family cognitive behavioral therapy) or a control condition (treatment as usual); (3) included both preintervention and postintervention assessments and reported at least 1 of the following outcome indicators: anxiety, depression, PTSD, and family function; and (4) the full-text article was available in English and Chinese. Meanwhile, articles were excluded from the analysis process if they (1) were duplicated studies, research protocols providing only an intervention model without data evaluation, case reports, conference abstracts, and reviews; (2) did not include data for meta-analysis; or (3) were family interventions targeting only disease-specific knowledge.

Two reviewers (Ren and Wang) independently reviewed the potential studies according to the inclusion/exclusion criteria. An initial eligibility screening based on the title and abstract was conducted, followed by an assessment of the full text of each study. Disagreements were resolved by consultation with a senior reviewer who approved the final list of included studies. Data were systematically extracted from each included study by the 2 reviewers independently using a standardized form to obtain the study characteristics and intervention characteristics. Any disagreements were resolved through group discussion until an agreement was reached. The following data were retrieved from each publication: name of the first author, year of publication, country, participants’ demographic characteristics, number of participants randomized to the intervention or control, intervention methods, duration and frequency of intervention, duration of followup, measurement tools, and outcome indicators.

Two reviewers independently evaluated the risk of bias in individual studies using the Cochrane risk of bias tool.¹³ The source of risk was assessed using the following 6 items: random sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting, and other biases. Each item was judged to be high risk, low risk, or unclear.

RevMan 5.4.1 (Cochrane Collaboration, Oxford, UK), a program of the Cochrane Library, was used for data analysis. The standardized mean difference (SMD) with 95% confidence interval (CI) was calculated to synthesize the pooled effects of the continuous data. Heterogeneity was tested using the I² statistic. A fixed effect model was used when the studies were homogenous (I² ≤ 50%, P > 0.1), and a random effects model was employed when study heterogeneity was found (I² > 50%, P ≤ 0.1). Sensitivity analysis or subgroup analysis was performed when appropriate. To assess publication bias, Egger’s and Begg’s tests were conducted using Stata 17.0 (Stata-Corp, College Station, TX, USA), and P < 0.05 was considered statistically significant.

The literature search identified a total of 1373 studies (14 in CNKI, 9 in Wanfang, 11 in VIP, 219 in CBM, 39 in WOS, 1 in Cochrane, 139 in PubMed, 121 in MEDLINE, 365 in Embase, 418 in PsyARTICLES, and 37 through manual search of the reference list). After duplicates (n = 15) were removed using NoteExpress software, 1358 studies were reviewed. After the titles and abstracts were screened, 1270 irrelevant articles were excluded. The full texts of 88 articles were assessed with the selection criteria, and a total of 11 RCT studies were selected for meta-analytic evaluation. The study selection process is summarized in the PRISMA flow diagram (Figure 1).

Figure 1.

Eleven studies^6,9,14-22 were included, and they were published between 1998 and 2021; of these, 4 were from the United States, ^16–18,20 2 were from China,^9,14 1 was from the Netherlands,¹⁵ 1 was from Iran,¹⁹ 1 was from Sweden,⁶ 1 was from Cyprus,²¹ and 1 was from Denmark.²² A total of 894 subjects were included in this review. Most included studies focused on the following psychosocial outcomes: anxiety (n = 8), depression (n = 5), and PTSD (n = 6). Only 2 studies reported on family function. The interventions ranged from 1 d¹⁶ to 24 weeks²² in duration. The interventions ranged in frequency from once every 2–3 weeks to twice a week. The detailed characteristics of the included studies are presented in Table 2.

Figure 2 summarizes the assessment results of the 7 domains of bias for individual studies. Of the 11 RCT studies included, only 2^16,22 were considered low risk for the generation of random sequences, and 4^6,15,17,18 were considered low risk for allocation concealment. No study was judged as having a low risk of bias for blinding of participants and personnel and blinding of outcome assessment because blinding family psychosocial intervention is difficult. All 11 studies were judged as having a low risk of bias for selective reporting bias and other biases, and only 1 study was judged as unclear for incomplete outcome data.

Figure 2.

Egger’s and Begg’s tests of anxiety, depression, and PTSD were not statistically significant (all P > 0.05), and Egger’s test of family function showed no statistical significance (P > 0.05), indicating that there was no publication bias (Table 3).

The main strength of this study is its exclusive focus on RCTs of family psychosocial interventions involving caregivers of children with cancer. However, several limitations should be noted. First, the differences in measurement tools, timing and duration of the interventions, intervention strategies, and follow-up period in the included literature may have resulted in bias in this analysis. Second, a certain degree of risk of bias was detected in the 11 included studies, possibly affecting the quality of the obtained evidence in this review. For instance, due to the complexity of family psychosocial intervention, blinding the participants and professionals involved in the interventions was difficult, leading to potential detection bias. Third, only published English and Chinese studies were included in this review, causing a potential language bias. Finally, the small number of included studies and small sample sizes of most included studies might limit the precision of evidence acquired via the pooled analysis.