TABLE 4.

Multiple mediation testing of adherence measures on the effect of social support on weight change at year 1^*

Potential mediators	Effect of social support on mediator (a)	Effect of mediator on % weight change (b)	Effect of social support on % weight change without mediation (c)	Effect of social support on % weight change with mediation (c’)	Indirect effect	95% Bootstrap Confidence Interval
Social Support: Family and friends join physical activity
Weekly minutes of physical activity	57.2 (13.7) <0.001	−0.016 (0.004) <0.001	−2.10 (0.84) 0.012	−1.06 (0.75) 0.155	−0.85 (0.31)	−1.56 – −0.33
Total meal replacements consumed	36.8 (23.1) 0.113	−0.004 (0.002) 0.103			−0.17 (0.15)	−0.54 – 0.05
Treatment sessions attended	1.89 (0.92) 0.041	−0.209 (0.058) <0.001			−0.37 (0.21)	−0.84 –0.001
Social Support: Family and friends shop and prepare food
Weekly minutes of physical activity	−17.1 (12.9) 0.185	−0.016 (0.004) <0.001	0.51 (0.76) 0.791	−0.18 (0.67) 0.782	0.28 (0.24)	−0.13 – 0.83
Total meal replacements consumed	−30.6 (21.2) 0.149	−0.004 (0.002) 0.103			0.14 (0.13)	−0.06 – 0.44
Treatment sessions attended	−0.31 (0.85) 0.711	−0.209 (0.058) <0.001			0.06 (0.18)	−0.28 – 0.43
Social Support: Family and friends attend medical visits
Weekly minutes of physical activity	22.4 (14.9) 0.134	−0.016 (0.004) <0.001	−0.78 (0.91) 0.412	−0.62 (0.77) 0.419	−0.35 (0.22)	−0.82 – 0.04
Total meal replacements consumed	17.1 (24.6) 0.488	−0.004 (0.002) 0.103			−0.08 (0.11)	−0.33 – 0.11
Treatment sessions attended	0.29 (0.98) 0.765	−0.209 (0.058) <0.001			−0.06 (0.19)	−0.46 – 0.29

^*Regression coefficient for all paths based on models adjusting for baseline age, education, income, insulin use, weight, sex, and clinic site. All p-values and CIs are corrected for multiple testing.

Regression coefficient (standard error)

p-value