Joost Peerbooms

111 PRP in plantar fasciitis what we were interested in; therefore, we did not report the interactions of treatment with the quadratic and cubic time effects. A possible explanation for this discrepancy is that the assumption of linear change over time is implausible. If change in an outcome over time is not linear yet time is modelled only linearly, then spurious interaction effects between time and other variables may arise. 17 Indeed, the FFI Pain, FFI Disability, FFI Activity, AOFAS, and WHOQOL Physical Health models showed linear as well as significant quadratic and cubic time effects, suggesting that the change in these outcome measures cannot be considered linear. This finding is corroborated by visual inspection of the growth curves of both treatment groups (Figure 2). The treatment groups did not show differences with respect to their change in quality of life over time, as indicated by the nonsignificant treatment × time interaction effects for all models involving the WHOQOL-BREF outcomes. Appendix Table A2 is similar to Table 2, yet it presents the results of the 3-way interaction among treatment group, time and dosage. This test indicates whether the differences between treatment groups in the change in the outcome measures over time depend on the used injection dosage. For all outcome measures, this interaction effect failed to reach significance, both for models with linear time only and for models with linear, quadratic and cubic time effects. These results suggest that the injection dosage did not affect the differences between the treatment groups in their change on the outcomes over time. 7