26 CHAPTER 1 Statistical analysis All statistical analyses were performed using IBM SPSS Statistics 25 (IBM Corp., Armonk, NY, USA). For each outcome measure (drooling quotient or VAS at the 8-week assessment), a multilevel linear model was used to account for the relation between the number of repeated injections (ordinal variable) within individuals. Sample size estimation was not suitable because of the retrospective nature of the study. This model included fixed effects for baseline (drooling quotient or VAS respectively), age, sex, cerebral palsy, degree of mobility (defined by ambulant or non-ambulant), developmental age (defined as <4y or >4y), and epilepsy and random effects for intercept and participants. Predictors were deemed as fixed as they generally did not change through time. There was no need for imputation of missing data in the outcome variables as multilevel linear models are competent in handling these missing values.19 A paired sample t-test and repeated analysis of variance (ANOVA) was used to analyse the differences in baselines of drooling quotient and VAS and the differences in the 8-week assessments between consecutive injections. To evaluate potential bias due to a difference in duration between baseline and its respective follow-up assessment (as the 8-wk assessment does not always take place at exactly 8wks after injection), a paired sample t-test was performed. A paired sample t-test was also performed to evaluate differences between injections in the duration between the onabotulinum neurotoxin A injection and the subsequent baseline measurement, because the previous injection may influence the next baseline through ongoing effect. RESULTS Of all the children treated with onabotulinum neurotoxin A injections for drooling between 2000 and 2020, 85 were treated with at least two subsequent, identical onabotulinum neurotoxin A injections. One patient turned 4 years old 3 weeks
RkJQdWJsaXNoZXIy MjY0ODMw