Joost Peerbooms

110 Chapter 7 and cubic time effects. The linear time models assume that patients change linearly over time on the outcome measures. For these models, the FFI Pain, FFI Disability, FFI Activity, and AOFAS outcomes showed significant interaction effects between treatment and time, suggesting that the treatment groups differ in their change in these outcomes over time. However, when inspecting the results of the models that also include quadratic and cubic time effects, it turns out that the treatment groups now only show significant differences in their change on the FFI pain scores over time and no longer on the FFI Disability, FFI activity, and AOFAS outcomes. Table 2 F Test on the Interaction Between Treatment and Time α Outcome measure Time modeled linearly Time modeled linearly, quadratically and cubically F -test P value F -test P value FFI Pain b.c F (1,96.634) = 8.140 0.005 d F (1,178.118) = 6.884 0.009 d FFI Disability b F (1,117.727) = 11.227 0.001 d F (1,210.506) = 2.856 0.092 FFI Activity b F (1,122.090) = 7.176 0.008 d F (1,202.982) = 1.393 0.239 AOFAS b F (1,126.818) = 12.955 <0.001 d F (1,211.989) = 2.922 0.089 WHOQOL-BREF Overall QOL and General Health F (1,116.178) = 0.579 0.448 F (1,213.492) = 1.967 0.162 Physical Health b F (1,122.849) = 6.201 0.014 F (1,175.289) = 0.360 0.549 Psychological Health F (1,113.228) = 0.168 0.683 F (1,167.993) = 0.376 0.541 Social Relationships F (1,104.092) = 0.252 0.617 F (1,187.356) = 2.393 0.124 Environment F (1,126.934) = 0.940 0.334 F (1,192.926) = 0.037 0.847 α For all outcome measures, the results of the F test on the interaction between treatment and time for models with and without adjustment for quadratic and cubic time effects. AOFAS, American Orthopaedic Foot & Ankle Society; FFI, Foot Function Index; WHOQOL-BREF, World Health Organization Quality of Life. b These models showed significant quadratic and cubic time effects. c Primary outcome measure. d Statistically significant after the Bonferroni correction (see the Statistical Analysis section). Note that in the models including linear, quadratic and cubic time effects, we report in Table 2 only the interaction effects between treatment and the linear time. The interactions between treatment and quadratic and cubic time were included in the model, although their significance was similar to those of the linear time × treatment interaction. We decided to report the results of the linear time × treatment interaction to test the null hypothesis that the 2 groups were equal in their change on the outcome over time, after accounting for nonlinear change in the outcome measure. This is exactly