Skip to main content

Advertisement

Table 6 Estimates of Direct and Indirect Effects With and/or Without Mediator-outcome Interaction

From: Implementation and reporting of causal mediation analysis in 2015: a systematic review in epidemiological studies

References Without exposure-mediator interaction (95 % confidence interval) With exposure-mediator interaction (95 % confidence interval) Discrepancy found with and without exposure-mediator interaction reason discussed
Banack et al. [26] Not reported Controlled direct effect with CVD
Risk ratio: 0.62 (0.49, 0.78)
Risk difference: −0.12 (−0.20, −0.04)
Controlled direct effect without CVD
Risk ratio: 1.30 (1.13, 1.49)
Risk difference: 0.03 (0.01, 0.05)
Total effect
Risk ratio: 1.24 (1.11, 1.39)
Risk difference: 0.03 (0.02, 0.05)
Not applicable
D’Amelio et al. [27] Effects of treatment on glucose level at 12 months mediated by OC at 6 months:
Natural direct effect: −0.033 (−0.186, 0.121)
Natural indirect effect: −0.050 (−0.178, 0.078)
Total effect: −0.082 (−0.174, 0.009)
Not reported Not applicable
Freeman et al. [28] Anxiety (Boot SE), P value
Direct effect: 0.21 (0.16), 0.19
Indirect effect: 0.18 (0.11), 0.09
Total effect: 0.39 (0.16), 0.01
Depression (Boot SE), P value
Direct effect: 0.24 (0.14), 0.09
Indirect effect: 0.15 (0.11), 0.18
Total effect: 0.39 (0.16), 0.01
BCSS—negative self (Boot SE), P value
Direct effect: 0.33 (0.18), 0.06
Indirect effect: 0.06 (0.08), 0.48
Total effect: 0.39 (0.16), 0.01
BCSS—positive self (Boot SE), P value
Direct effect: 0.40 (0.17), 0.02
Indirect effect: −0.01 (0.05), 0.92
Total effect: 0.39 (0.16), 0.01
BCSS—negative other (Boot SE), P value
Direct effect: 0.22 (0.16), 0.17
Indirect effect: 0.18 (0.11), 0.13
Total effect: 0.39 (0.16), 0.01
Not applicable Not applicable
Jackson et al. [29] Stroke
Direct effect: 1.13 (1.05,1.22)
Indirect effect: 1.005 (1.001,1.011)
Total effect: 1.14 (1.06,1.22)
Stroke
Direct effect: 1.13 (1.06,1.22)
Indirect effect: 1.005 (1.001,1.011)
Total effect: 1.14 (1.06,1.22)
No
Not applicable
Kositsawat et al. [30] Not reported Not applicable Not applicable
Louwies et al. [31] Systolic blood pressure
Direct effect: 2.93 (CIs not reported)
Indirect effect: −0.42 (−1.35 to 0.17)
Total effect: not reported
Diastolic blood pressure
Direct effect: 3.15 (CIs not reported)
Indirect effect: −0.59 (−1.44 to 0.07)
Total effect: not reported
Not applicable Not applicable
Lu et al. [32] Overweight (for
metabolic mediators)
Blood pressure
Natural direct effect: 1.16 (1.09–1.24)
Natural indirect effect: 1.06 (1.03–1.08)
Total effect: 1.22 (1.14–1.30)
Obesity (for
metabolic mediators)
Blood pressure
Natural direct effect: 1.28 (1.15–1.43)
Natural indirect effect: 1.13 (1.07–1.19)
Total effect: 1.42 (1.25–1.60)
Overweight (for
metabolic mediators)
Blood pressure
Natural direct effect: 1.16 (1.09–1.24)
Natural indirect effect: 1.05 (1.02–1.08)
Total effect: 1.22 (1.14–1.30)
Obesity (for
metabolic mediators)
Blood pressure
Natural direct effect: 1.28 (1.15–1.42)
Natural indirect effect: 1.10 (1.03–1.17)
Total effect: 1.43 (1.25–1.62)
No
Not applicable
Mendola et al. [33] Not reported Peri- or intraventricular hemorrhage:
Controlled direct effect: 3.2 (1.4–7.7)
Total effect: 2.9 (2.4–3.4)
Not applicable
Messerlian et al. [34] Uterine: <35 weeks
Controlled direct effect: 2.43 (0.85, 6.93)
Total effect: 2.27 (1.32, 3.89)
Not applicable Not applicable
Raghavan et al. [35] Metabolic syndrome score:
Direct effect: not reported
Indirect effect: 1.20 (1.07, 1.33)
Total effect: not reported
Not applicable Not applicable
Rao et al. [36] Smoking: low vs. high
Controlled direct effect: 1.5 (1.4, 1.5)
Total effect: 1.6 (1.4, 1.9)
Not applicable Not applicable
Song et al. [37] HOMA-IR:
Effect not mediated: 1.32 (0.95, 1.88)
Mediated effect: 1.22 (1.02, 1.49)
SHBG:
Effect not mediated: 1.97 (1.25, 3.10)
Mediated effect: 1.16 (1.03, 1.33)
E-selectin:
Effect not mediated: 1.56 (1.10, 2.21)
Mediated effect: 1.12 (1.02, 1.23)
Systolic blood pressure:
Effect not mediated: 1.61 (1.25, 2.16)
Mediated effect: 1.03 (1.01, 1.07)
Not reporteda Not applicable
Xie et al. [38] Not reported Not applicable Not applicable
  1. BCSS brief core schema scales; CI confidence interval; CVD cardiovascular disease; HOMA-IR homeostatic model assessment for insulin resistance; OC osteocalcin; SE standard errors; SHBG sex hormone-binding globulin
  2. aDespite allowing for interaction, only models assuming no interaction were adopted due to no significant interaction between any of the exposures and mediators was observed