Demographics of included studies.
Initial database searching results yielded 1,521 results in January 2023. After 544 duplicates were removed, 980 remained. We found 3 studies through hand searches. During review of titles and abstracts, an additional 925 were excluded because they did not include couple-based interventions (primary reason for exclusion) and/or they failed to meet all other inclusion criteria. Of the remaining 28, 16 primary studies were excluded due to 1) narrative review/systematic review (n=12), 2) a case study (n=1), 3) research without results (n=2) and 4) letter (n=1). Finally, 12 primary studies (n=12) met inclusion criteria and were in this systematic review and meta-analysis. See Figure 1.
The 12 primary studies that met inclusion criteria provided 13 between-group comparison (K=13) because one study had three group comparisons [15]. We compared groups that were similar except for the couple-based intervention. All 12 primary studies had been published between 2005 and 2021. A total of 938 participants were included across the 12 primary studies; 463 participants practiced in couple-based interventions and 465 participants were served as controls. Ten of twelve primary studies were conducted in the United States of America [1-2, 15, 18-23] as well as two in Canada [24-25]. See supplementary Table 2 and 3. Participants’ mean ages ranged from 40.95 to 68.11 years. See Table 1 for intervention description including total weeks of intervention, number of session/week and duration of session in minutes/session.
Table 1 Characteristics of Primary Studies (s=12)
Characteristics
|
s
|
Min
|
Q1
|
Mdn.
|
Q3
|
Max
|
Mean
|
SD
|
Mean age (years)
|
12
|
40.95
|
51.87
|
54.11
|
61.81
|
68.11
|
56.14
|
7.29
|
Total Sample size at analysis
|
|
|
|
|
|
|
|
|
- MBI group
|
12
|
8
|
16.50
|
25.00
|
56.50
|
96.00
|
35.62
|
25.92
|
Control group
|
12
|
7
|
12.00
|
25.00
|
51.50
|
113.00
|
35.77
|
31.34
|
Weeks of structured CBI
|
12
|
4.00
|
4.50
|
6.00
|
8.00
|
12.00
|
6.30
|
2.25
|
Days across intervention (length)
|
12
|
21.00
|
24.50
|
35.00
|
49.00
|
366.00
|
59.38
|
92.69
|
Structured MBI session/week
|
10
|
1.00
|
1.00
|
1.00
|
1.00
|
1.00
|
1.00
|
.00
|
Structured MBI min./session
|
9
|
34.00
|
50.00
|
60.00
|
67.50
|
90.00
|
59.88
|
16.01
|
Dose (length x duration)
|
9
|
1190.0
|
1260.0
|
1540.0
|
2520.0
|
3150.0
|
1808.33
|
756.35
|
Days after intervention measured
|
11
|
14.00
|
21.35
|
80.50
|
91.50
|
275.00
|
84.09
|
76.93
|
% Attrition, CBI group
|
12
|
.00
|
13.11
|
20.00
|
33.97
|
45.00
|
20.92
|
13.90
|
% Attrition, Control group
|
12
|
.00
|
8.33
|
17.18
|
31.86
|
41.67
|
18.60
|
14.01
|
s=number of studies providing data, Min=minimum, Q1=first quartile, Mdn=median, Q3=third quartile, Max=maximum, CBI=Couple-based intervention
Effects of couples-based interventions
Depression
The summary effect size across the 11 comparisons was g=.36 (95%CI=.026, .684, p=.034, I2=76%), indicating that couples-based interventions had a moderate effect in reducing depressive symptoms among patients with cancer. Of all 11 comparisons, four comparisons had significant positive effects improvement. See Figure 2.
The couples-based intervention group’s pre-post comparisons demonstrated significant reduction in depressive symptoms with an effect size of g=.74 (p=.022) for correlated groups (r=.8) and g=.72 (p=.014) for uncorrelated groups. The control group’s pre-post effect sizes showed no improvement in depressive symptoms for uncorrelated group (g=.51, r=.0) and improvement of depressive symptoms for correlated group (g=.49, r=.8). See Table 2.
Table 2. Effect size of Couple-Based Intervention vs Control groups
Comparison
|
Yoga group
|
k
|
ES
|
p(ES)
|
95% CI
|
SE
|
I2
|
Q
|
p(Q)
|
Depression
|
Couple-based intervention vs. Control groups
|
11
|
.36
|
.034
|
.026, .684
|
.168
|
76.66
|
42.84
|
<.001
|
Single group
Couple-based intervention
|
pre- vs. post (r=0.0)
|
9
|
.72
|
.022
|
.105, 1.342
|
.316
|
90.56
|
84.76
|
<.001
|
pre- vs. post (r=0.8)
|
9
|
.74
|
.014
|
.148, 1.327
|
.301
|
98.00
|
401.98
|
<.001
|
Control group
|
pre- vs. post (r=0.0)
|
9
|
.51
|
.113
|
-.121, .499
|
.322
|
92.06
|
100.76
|
<.001
|
pre- vs. post (r=0.8)
|
9
|
.49
|
.111
|
-.114, 1.103
|
.310
|
98.35
|
484.61
|
<.001
|
Anxiety
|
Couple-based intervention vs Control group
|
8
|
.29
|
<.001
|
.143, .445
|
.077
|
0.00
|
4.66
|
.701
|
Single group
|
Couple-based intervention
|
|
|
|
|
|
|
|
|
pre- vs. post (r=0.0)
|
8
|
1.27
|
.003
|
.421, 2.114
|
.432
|
92.92
|
98.87
|
<.001
|
pre- vs. post (r=0.8)
|
8
|
1.31
|
.002
|
.485, 2.128
|
.419
|
98.56
|
496.05
|
<.001
|
Control group
|
|
|
|
|
|
|
|
|
pre- vs. post (r=0.0)
|
8
|
.44
|
.202
|
-.235, 1.109
|
.343
|
92.13
|
88.97
|
<.001
|
pre- vs. post (r=0.8)
|
8
|
.44
|
.195
|
-.233, 1.093
|
.336
|
98.39
|
436.97
|
<.001
|
Anxiety
The summary effect size across 8 comparisons was g=.29 (95%CI=.143, 445, p<.001, I2=0%), indicating that couples-based interventions had a small effect on improving anxiety symptoms among patients with cancer. Also, of all 8 comparisons, two comparisons had positive effects improvement. See Figure 3.
Additionally, the couple-based interventions group’s pre-post comparisons demonstrated significant reduction in anxiety symptoms with effect size of g=1.31 (p=.002) for correlated group (r=.8) and g=1.27 (p=.002) for uncorrelated group. The control group’s pre-post effect sizes showed no improvement in anxiety symptoms for uncorrelated group (g=.44, r=.0) and improvement of depressive symptoms for correlated group (g=.44, r=.8). See Table 3.
Table 3 Categorical Moderator Results for Depression Comparing Couple-based intervention versus Control Groups
Moderator
|
k
|
ES
|
SE
|
Var.
|
95%CI
|
Z
|
p(Z)
|
Qbet
|
p(Qbet)
|
Source characteristics
|
Funding
|
|
|
|
|
|
|
|
.311
|
.577
|
Unfunded
|
2
|
.153
|
.402
|
.162
|
-.635, .941
|
.380
|
.704
|
|
|
Funded
|
9
|
.403
|
.201
|
.040
|
.009, .797
|
2.007
|
.045
|
|
|
Method characteristics
|
Intention-to-treat
|
|
|
|
|
|
|
|
.411
|
.522
|
No
|
8
|
.286
|
.204
|
.042
|
-.114, .687
|
1.402
|
.161
|
|
|
Yes
|
3
|
.541
|
.340
|
.116
|
-.126, 1.208
|
1.589
|
.112
|
|
|
Fidelity of intervention
|
|
|
|
|
|
|
|
4.048
|
.044
|
No
|
7
|
.170
|
.161
|
.026
|
-.145, .484
|
1.058
|
.290
|
|
|
Yes
|
4
|
.702
|
.210
|
.044
|
.290, 1.114
|
3.340
|
.001
|
|
|
Intervention characteristics
|
Group discussion
|
|
|
|
|
|
|
|
.603
|
.438
|
No
|
4
|
.455
|
.254
|
.064
|
-.042, .952
|
1.794
|
.073
|
|
|
Yes
|
6
|
.195
|
.219
|
.048
|
-.235, .625
|
.888
|
.375
|
|
|
Counseling
|
|
|
|
|
|
|
|
.022
|
.883
|
No
|
7
|
.286
|
.216
|
.047
|
-.138, .709
|
1.323
|
.186
|
|
|
Yes
|
3
|
.348
|
.365
|
.133
|
-.368, 1.064
|
.954
|
.340
|
|
|
Participants
|
|
|
|
|
|
|
|
|
|
Breast cancer
|
|
|
|
|
|
|
|
2.171
|
.141
|
No
|
8
|
.238
|
.186
|
.035
|
-.126, .603
|
1.283
|
.199
|
|
|
Yes
|
2
|
.760
|
.301
|
.091
|
.170, 1.350
|
2.525
|
.012
|
|
|
k=number of comparisons, Q=heterogeneity statistics, SE=standard error, Var.=variance, NR=not reported
Subgroup analyses
Because moderate heterogeneity was identified across studies (I2=76%) in depression outcomes, a subgroup analysis was conducted to explore the source of heterogeneity. There were four moderators that influenced the effect size. Fidelity of couple-based intervention had a greater effect (g=.702, 95%CI .290, 1.114, p=.001) than non-fidelity of intervention (g=.170, 95%CI -.145, .484, p=.290). An increase in the length of intervention in total days decreased the effect size of couple-based intervention on depressive symptoms in patients with cancer (β=-.023, Q=3.96, p=.046). The higher the number of weeks across the intervention, the lower the effect size in reducing depressive symptoms in patients with cancer (β=-.164, Q=3.96, p=.046). Finally, conducting outcome assessments at an increasing number of days post-intervention yielded lower effect size of reducing depressive symptoms in patients with cancer (β=-.003, Q=8.06, p=.004). See Table 4.
Table 4 Continuous Moderators of the Effects of Couple-based intervention on Depression
Moderator
|
k
|
Slope
|
SE
|
Tau2
|
Qmodel
|
p
|
Study characteristic
|
|
|
|
|
|
|
Publication year
|
11
|
.019
|
.024
|
.124
|
.63
|
.427
|
Sample characteristic
|
|
|
|
|
|
|
Age (mean)
|
11
|
.012
|
.019
|
.139
|
.37
|
.541
|
Method characteristic
|
|
|
|
|
|
|
%Attrition
|
11
|
.009
|
.010
|
.118
|
.84
|
.360
|
Intervention characteristics
|
|
|
|
|
|
|
Intervention length (total day)
|
11
|
-.023
|
.011
|
.076
|
3.96
|
.046
|
Number of weeks across intervention
|
11
|
-.164
|
.082
|
.074
|
3.96
|
.046
|
Number of session/wks.
|
|
|
|
|
|
|
Duration of CBI min./session
|
8
|
-.012
|
.009
|
.083
|
1.62
|
.203
|
Dose (Length x Duration)
|
8
|
-.003
|
.001
|
.079
|
1.82
|
.178
|
Days After intervention measured
|
11
|
-.003
|
.001
|
.042
|
8.06
|
.004
|
k=number of comparisons, Q=heterogeneity statistics, CBI=Couple-based intervention
Publication bias
For depression outcomes, the funnel plot was visually asymmetrical suggesting that publication bias was likely; primary studies with large sample sizes with negative effects were missing. The Egger’s test was significant with an intercept of -1.64 (95%CI: -5.43, 2.15, p=.176). Begg and Mazumdar’s test was also significant (p=.500). These results indicate the no potential presence of publication bias.
For anxiety outcomes, the funnel plot was visually symmetrical suggesting that publication bias was not likely; primary studies with large sample sizes with negative effects were missing. The Egger’s test was significant with an intercept of -1.01 (95%CI: -2.92, .90, p=.121). Begg and Mazumdar’s test was also significant (p=.193). These results indicate the no potential presence of publication bias.