Joost Peerbooms

108 Chapter 7 individual differences in growth trajectories were taken into account by allowing both the intercept and slope to vary across all patients. Time was modelled continuously and both linear as well as quadratic and cubic time effects were investigated. Any differences between the treatment groups on the baseline scores of the outcome measures were handled with a constrained longitudinal data analysis that constrains the baseline means of the treatment groups to be equal by omitting the main effect for treatment from the statistical model. 5 Inferences regarding the difference between treatments are based on the interaction effect between treatment group and time. In the linear mixed model analysis, parameters were estimated using restricted maximum likelihood estimation. Analysis of covariance was used to test the null hypothesis of equal outcome means at the 12-month follow-up, adjusted for baseline differences. The effects of all abovementioned analyses were adjusted for the potential confounders sex, smoking, and duration of symptoms before treatment. Differences between groups in the number of patients showing at least 25% improvement in pain symptoms was assessed using a Chi-square test. P-values < .05 were considered statistically significant for the primary outcome measure (FFI Pain), while a Bonferroni correction was applied to adjust for multiple testing of the treatment effects for the secondary and remaining outcome measures. In order to retain sufficient statistical power, the Bonferroni correction was applied separately to the secondary outcomes (FFI Disability and FFI Activity, AOFAS; significance level = .05 / 3 = .0167) and the remaining outcome measures (WHOQOL-BREF; significance level = .05 / 5 = .01). Confidence intervals were calculated at the 95% level. All data were analysed by a blinded researcher (P.L.) using SPSS (v 23; IBM). RESULTS The flowchart in Figure 1 indicates that of all 115 randomized patients, 63 were allocated to the PRP group and 52 to the control group. Of the 63 patients in the PRP group, 46 completed the study and 17 patients were lost during the 12-month follow-up. For logistic reasons, 16 patients were treated with an injection made of the 30-mL PRP kit instead of the 60-mL PRP kit. The influence of dosage on the treatment effect was assessed by inspecting the 3-way interaction effect between treatment group, time, and injection dosage. No differences were seen between the 30- and 60-mL doses. In the control group, 36 patients completed the study, and 16 patients were lost to follow-up. Table 1 presents the baseline characteristics for patients allocated to the PRP and control group separately. Appendix Table A1 indicates that for all outcome measures, the Little MCAR test (missing completely at random) failed to reach significance, suggesting that the missing values on those outcome measures are likely missing completely at random. This result allowed