The literature search from various databases has identified 6847 publications up to July 26, 2019, of which 1146 were excluded as they were duplicates. Around 4400 titles and 1160 abstracts were excluded during the initial screening for the titles and abstracts, and 162 articles were considered for the retrieval of the full texts. Finally, after a detailed assessment, 130 studies were included. An additional eleven studies were obtained by cross-referencing making a final total of 141 studies at this stage. Of the total 141 articles that could be retrieved, thirteen articles were excluded due to a lack of a control group. And among the remaining 116 full-text articles, 71 were excluded due to the lack of statistical information and not receiving email replies from the authors. Additionally, another 45 articles were excluded as they have not reported the effect size for both wife and husband separately (n = 45). Finally, 12 articles remained that satisfied all the review criteria and were used for meta-analysis for the relationship satisfaction area, and 8 articles remained for the relationship communication area. Notably, 8 articles contain information for both RS and RC areas while 4 articles had information exclusively for the RS theme (Fig. 1).
Description of studies
The basic characteristics of the studies are shown in Table 1. According to the type of programs mentioned above, this review contained 76 evaluations of therapy programs, 35 education/communication skills programs, twenty (20) enrichment programs, and ten (10) counseling programs. However, during computation and reporting of effect sizes, among the 12 studies that remained for meta-analysis for RS, five (5) were education/communication skills programs, three (3) enrichment programs, and four (4) therapy programs. Among the eight studies that remained for meta-analysis for RC, five (5) were education/communication skills programs, two (2) were enrichment programs, and the last one being a therapy program. Usually, therapy and counseling programs happen in a clinical setting with a trained psychologist providing treatment. These programs can be based upon a variety of different treatment formats. In this study, the therapy program included were cognitive-behavioral couple therapy (CBCT), Integrative Behavioral Couple Therapy (IBCT) and Support-Focused Marital Therapy (SFMT). Education and communications skills programs tend to be didactic and support both distressed and non-distressed couples. Enrichment programs are usually limited to normal and healthy couples.
Most (n = 9) studies were conducted in high-income countries where seven (7) studies are from the USA, one (1) from Germany, and one (1) from Australia. Furthermore, only seven three were conducted in upper- and lower-middle-income countries: 2 from Korea, one (1) from China. The studies included in the meta-analysis had data from 4565 participants (2460 cases and / 2105 controls). All of the entered studies have used randomization methods, random assignment to intervention and control group, and had a control group. However, three (3) studies have reported their study design was a quasi-experimental study [17, 30, 32].
The educational method for Li (2015), Einhorn (2010), and Allen (2011) studies were a weekend workshop; for Alvaro (2001) study, it was held as a seminar, and Doss’s (2016) study ran online calling and chatting. Others used the trained psychologist or trainers for the operation of programs in defined sessions. Kong (2005) and Halford (2000) studies evaluated relationship satisfaction programs with five (5) sessions; Young-Ran (2012), Carson (2004), and Kroger (2017) were between 5 and 10 sessions (6, 8, and 9 sessions perceptively); Hrapczynski (2008) had ten (10) and Shapo (2001) had 12 sessions. For communication skill programs, Alvaro (2001), Allen (2011), and Kroger (2017) have included less than five sessions (4, 4, and 3 sessions perceptively), and Kong (2005), Hrapczynski (2008), Einhorn (2010), Young-Ran (2012), and Li (2015) have held more than five sessions (5, 10, 5, 6, and 6 sessions respectively). Most of the sessions in these studies have lasted for 10 h and above. From the 12 studies, five (5) studies assessed distressed couples, six (6) studies contained non-distressed couples, and one (1) study contained both distressed and non-distressed couples.
Risk of bias
The risk of bias was assessed for clinical trials. The bias such as the unclear risk of selection bias (due to lack of information on the method of randomization n = 11 and concealment, n = 12), performance bias (due to lack of information on blinding of participants and personnel, n = 12), and detection bias (blinding of outcome assessment, n = 12) were observed. The risk of bias for the included studies was low for reporting bias, attrition bias, and other sources of bias (Fig. 2).
Effects of interventions (meta-analysis results)
Of the total 12 interventions, the impact of five (5) education/communication skills programs, three (3) enrichment programs, and four (4) therapy programs were investigated on RS among couples.
The pooled MD was 0.28 (95% CI = 0.08 to 0.49, I2 = 79.3% p = 0.0001), 0.24 (95% CI = − 0.12 to 0.60, I2 = 78.7%, p = 0.855), 0.53 (95% CI = 0.35 to 0.71, I2 = 71.5%, p < 0.0001 0.000) respectively. As a result, the therapy programs showed the medium to large effect on CRS and educational programs showed small to medium effect. But the heterogeneity among studies was high. Enhancement programs showed no effect on CRS (Fig. 3A). In the area of RC, the enhancement programs showed small to large effect on CRC (pooled MD 1.31 (95% CI = 0.13 to 2.50, I2 = 94.7% p = 0.0001)) similar to the educational programs which also showed small to medium effect (pooled MD 0.32 (95% CI = 0.13 to 0.50, I2 = 74.5% p = 0.0001). However, the heterogeneity among these studies was high (Fig. 5A).
Subgroup analysis
The subgroup analysis conducted for following variables: gender differences (men, women), number of sessions (1: fewer 5 sessions, 2: between 5 and 10 sessions, 3: more than 10 sessions and workshop), total hours of interventions (1: fewer 10 h, 2: between 10 and 15 h, 3: more than 15 h), distress status (distressed and non-distressed), program type (therapy, enrichment, education/communication skills, and counseling). These variables were considered as moderators of marital satisfaction and communication effects based on previous literatures [10, 13, 14]. Subgroup analyses show that effect sizes are different for different subgroups of studies in current study.
In the RS area, the gender-based subgroup analysis showed the therapies had a more significant effect size on wives than husbands, and education was only effective for wives and enhancement programs with no effect for both couples (Fig. 3B). The subgroup analysis based on the number of sessions showed the between 5 and 10 sessions had a medium to large effect on CRS, and more than ten (10) sessions and workshops had small to medium effect but with high heterogeneity. Moreover, fewer than five (5) sessions did not affect couples’ relationship satisfaction (Fig. 3C). The subgroup analysis based on both the number of sessions and gender showed between 5 and 10 sessions had a larger effect size on husbands than wives. However, the workshop had no impact, and more than ten (10) sessions had the same effect on husband and wives (Fig. 3D). The subgroup analysis based on total hours of interventions showed fewer than 10 h of intervention had a medium to large effect on CRS while the more than 10 h had a small to medium effect (Fig. 4A). The subgroup analysis based on both hours and gender showed fewer than 10 h of intervention had a larger effect on wives’ relationship satisfaction than husbands. In contrast, between 10 and 15 h of intervention were effective only for husbands (Fig. 4B).
The subgroup analysis based on distressed and non-distressed couples showed interventions had small to medium effects on both distressed and non-distressed couples but with high heterogeneity (Fig. 4C). The subgroup analysis based on both distress/non-distressed and gender showed the interventions had a small to medium effect on both distressed wives and husbands but no effect on both non-distressed wives and husbands (Fig. 4D). As well, subgroup analysis based on distressed level, program type, and gender showed the only therapy programs were effective for both distressed wives and husbands and merely for non-distressed husband with small to larger effect without heterogeneity. Therapies and educational/communication skills were non-effective for both distressed and non-distressed wives and husbands. Sensitivity analysis did not apply to these results due to non-existing outlier data.
In the RC domain, the gender-based subgroup analysis showed that “education” was effective only for wives (Fig. 5B). The subgroup analysis based on the “number of sessions” showed that more than 5 sessions had a small to medium effect and less than 5 sessions had small to large effect on couples’ relationship communication (CRC) but with high heterogeneity (Fig. 5C). The subgroup analysis based on both the “number of sessions” and “gender” showed that less than 5 sessions had a small to medium effect on husbands with no effect on wives. Conversely, more than 5 sessions had a small to large effect on wives with no effect on husbands (Fig. 5D). The subgroup analysis based on “total hours of interventions” showed a small to large effect on CRC for interventions between 10 and 15 h (Fig. 6A). The subgroup analysis based on both “hours of intervention” and “gender” showed a small to larger effect on both husband and wives’ relationship communication for interventions between 10 and 15 h. In contrast, more than 10 h of intervention were effective only for wives (Fig. 6B).
The subgroup analysis based on “distressed” and “non-distressed” couples showed that interventions had “small to medium” effects on distressed couples and “small to large” effect among non-distressed couples but with high heterogeneity (Fig. 6C). The subgroup analysis based on both “distress/non-distressed” and “gender” showed that the interventions had “small to medium” effect among the distressed wives and “small to large” effect on non-distressed wives with no effect on distressed and non-distressed husbands (Fig. 6D).
Publication bias
Funnel plot 12 studies showed no evidence suggestive of publication bias, and also, the results of the Egger test were insignificant for publication bias (p = 0.460) (Fig. 7).