Are longitudinal randomised controlled oral health trials properly analysed? A meta-epidemiological study

Longitudinal designs with multiple outcome measurements are commonly encountered in oral health randomised controlled trials (RCTs). The aim of this meta epidemiological study was to assess whether optimal statistical analysis approaches have been used in longitudinal oral health RCTs.

PubMed search was undertaken in September 2021 for longitudinal oral health RCTs with at least 3 repeated outcome measurements which have been published between 2016 and 2020 in the highest impact general and specialty dental journals.

Study selection and data extraction were accomplished independently and in duplicate. The statistical methods undertaken in the selected articles were tabulated, and the association between study characteristics and use of optimal analyses were assessed using X² or Fisher's exact test and logistic regression.

Five hundred and five oral health RCTs were deemed eligible for inclusion. Of these, only 28.3% RCTs used optimal statistical analyses for a longitudinal trial design. For the trials with an optimal statistical approach, the most frequent test used was repeated measures analysis of variance (RM-ANOVA) followed by mixed effect models (MEM). The use of optimal statistical tests was predicated by the involvement of a statistician (OR: 2, 95% CI:1.27-3.18, p < 0.01), the journal impact factor (OR:1.19, 95% CI;1.1-1.29), continent of first author (likelihood ratio test p = 0.01), number of the authors (OR:1.22, 95% CI;1.12-1.3, p < 0.001), protocol registration (OR: 1.48, 95%CI; 1 to 2.2, p = 0.05), funding(OR:2.4, 95%CI; 1.6-3.7, p < 0.001), and dental specialty (likelihood ratio test p < 0.001).

Most longitudinal oral health RCTs did not use optimal statistical analyses. Greater awareness of optimal analyses used to assess longitudinal data reported in oral health trials is required to circumvent the reporting of suboptimal inferences, selective reporting and research waste.

Further progress is required to avoid suboptimal statistical analyses and fully utilise the benefits of the repeated measurements over time in oral health RCTs.