Table 4: Assessment of the reduction of the intervention effect on twelve-week change in RDQ* associated with mediator(s).

Total intervention
effect
Remaining effect after mediator(s)Effect reduced by mediator(s)% Reduction due to mediator(s)
Reduction (95% CI)

(I) Main cohort

Yoga versus self-care
 Self-efficacy only (A) (−1.64, −0.30)35.7%
 Improved sleep only (B) (−1.09, −0.03)18.3%
 Hours of back exercise only (C) (−0.91, 0.32)9.2%
 All A + B + C in the model (−2.44, −0.54)56.4%
 Best overall mediator scenario1 (−2.47, −0.58)61.3%
Stretching versus self-care
 Self-efficacy only (−1.13, −0.02)23.3%
 Hours of back exercise only (−1.16, 0.72)14.3%
 A + B in the model (−1.58, 0.38)29.6%
 Best overall mediator scenario2 (−2.38, −0.04)49.6%

(II) Saliva cohort

Yoga versus self-care
 Self-efficacy only (−1.85, −0.09)28.2%
 Hours of back exercise only (−1.24, 0.57)13.0%
 A + B in the model (−2.60, −0.14)45.5%
 Best overall mediator scenario3 (−3.48, −0.49)75.5%
Stretching versus Self-Care
 Self-efficacy only (−1.40, 0.01)21.9%
 Best overall mediator scenario4 (−2.08, 0.23)35.4%

RDQ: Roland-Morris Disability Questionnaire.
Model framework: each line is a separate model assessing the reduction in intervention effect after including mediator(s) in the model. Both the estimated total intervention (without any mediators) and mediator(s) models adjust for baseline variables age, sex, BMI, back pain days, work type, pain travelling down the leg, bothersomeness, and RDQ. Saliva cohort has the same baseline adjustments as the main cohort analysis with the addition adjustment of alcohol, tobacco, and medication use on the day of saliva sample collection.
Best overall mediator scenarios include the following mediators:
1Self-efficacy, conscious awareness of body, psychological distress, perceived stress, sleep disturbance, and hours of back exercise.
2Fear, self-efficacy, conscious awareness of body, psychological distress, positive states of mind, and hours of back exercise.
3Self-efficacy, Psychological Distress, perceived stress scale, sleep disturbance, hours of back exercise, and DHEA diurnal average (log scale).
4Self-efficacy, conscious awareness of body, perceived stress scale, positive states of mind, and cortisol awakening response.