Table 3

Intention to treat analysis—difference between group means (95% CI for difference)

Primary outcomes6 monthsP value12 monthsP value
Disability*8.65 (3.66 to 13.64)0.0017.02 (2.24 to 11.80)0.004
 Sensitivity analysis†8.80 (3.56 to 14.04)0.0016.94 (1.91 to 11.97)0.007
 Sensitivity analysis‡8.45 (3.22 to 13.69)0.0026.32 (1.56 to 11.08)0.010
 Sensitivity analysis§8.97 (3.96 to 13.98)0.0016.89 (2.07 to 11.72)0.005
 Sensitivity analysis¶7.62 (2.65 to 11.43)0.0036.54 (1.65 to 11.43)0.009
 Sensitivity analysis**8.84 (4.03 to 13.65)<0.0016.56 (1.87 to 11.24)0.006
 Per-protocol analysis††11.87 (6.59 to 17.14)<0.0019.03 (3.60 to 14.45)0.001
Effect size (95% CI)0.67 (0.27 to 1.06)0.55 (0.18 to 0.92)
Pain intensity*0.76 (–0.02 to 1.54)0.0560.65 (–0.20 to 1.50)0.134
 Sensitivity analysis†0.60 (–0.22 to 1.46)0.1480.57 (–0.31 to 1.46)0.204
 Sensitivity analysis‡0.79 (–0.07 to 1.66)0.0720.43 (–0.50 to 1.36)0.358
 Sensitivity analysis§0.77 (–0.02 to 1.57)0.0560.58 (–0.28 to 1.44)0.184
 Sensitivity analysis¶0.39 (–0.43 to 1.20)0.3470.42 (–0.46 to 1.31)0.346
 Sensitivity analysis**0.78 (0.02 to 1.55)0.0440.61 (–0.22 to 1.45)0.150
 Per-protocol analysis††1.23 (0.35 to 2.11)0.0070.87 (–0.13 to 1.86)0.086
Effect size (95% CI)‡‡0.36 (–0.03 to 0.74)0.31 (–0.05 to 0.67)

  • *Mean difference (95% CI) from linear mixed models including all three time points, controlling for within group effects (random effect) and centre (fixed effect).

  • †Based on linear mixed models including all three time points, controlling for within group effects (random effect) and centre, age, sex and time since diagnosis (fixed effects).

  • ‡Based on imputed data sets.

  • §Based on analysis of covariance (ANCOVA) of the two follow-up time points (6 months and 12 months), controlling for within group effects (random effect), and centre and baseline values (fixed effects).

  • ¶Based on inverse probability of censoring weighting.

  • **Based on general estimating equation analysis.

  • ††Per-protocol analysis based on linear mixed model analysis of data from those who completed all treatments, including all three time points and controlling for within group effects (random effect) and centre (fixed effect).

  • ‡‡Cohen’s D computed as the mean difference relative to pooled SD of baseline scores.