Oral Health Promotion Interventions Amongst Stroke Survivors During Rehabilitation

The study population in this clinical trial will be patients with moderate to severe stroke undergoing further stroke rehabilitation by returning regularly to the Day Rehabilitation Center (on an out-patient basis) of the Rehabilitation Unit at Tung Wah Hospital (TWH). The TWH Rehabilitation Unit, Sheung Wan, Hong Kong has been selected as the study site because it serves a major urban area in Hong Kong. TWH is a medical teaching hospital, patient medical care is standardized and a comprehensive set of records is compiled during stroke rehabilitation. Around 500 patients with acute stroke are admitted annually to the Rehabilitation Unit. The majority of the patients are stabilized (for up to 7 days) at Queen Mary Hospital, a major teaching hospital for treating acute medical conditions, prior to admission to the Rehabilitation Unit. On admission to the Rehabilitation Unit, each stroke patient undergoes a comprehensive medical assessment that includes a functional disability assessment using the modified Barthel Index and a rehabilitation programme is formulated. Following hospital discharge, patients requiring further rehabilitation return regularly to the Day Rehabilitation Center on an out-patient basis. Stroke patients deemed fit for discharge from the Rehabilitation Unit will be invited to participate in the study in accordance with the inclusion \& exclusion criteria.

The key outcome variable of this clinical trial is the dental plaque level (the key reservoir for oral pathogens). In a previous observational study we conducted among stroke survivors, the mean PI score was 2.1 (standard deviation [SD]=0.51) at discharge and 1.9 (SD=0.56) at 6 months, yielding a change score of 0.21 (SD=0.46). On the basis of this study, and proposing a difference in plaque change scores of at least 0.3 between baseline to follow-up (and also between groups) and with the standard deviation set at 0.46, then the number of study subjects that would be required is 38 per group, based on 80% power and the statistical significance level set at 0.05. Anticipating a 20% dropout rate over the course of the clinical trial, the initial sample size for each treatment group is proposed as 47 patients per group (94 subjects in total).

Assessments of clinical oral health status, presence of oral opportunistic pathogens and subjective oral health were conducted at baseline, 3 month and 6 month. All assessments were performed in the TWH Day Rehabilitation Center. Compliance was also assessed. Data will be stored on a computer (and backed up on a compact disc) in the Faculty of Dentistry, The University of Hong Kong. Professor Colman McGrath (Principle Investigator) will be responsible for its safekeeping. All data will be kept until analyses have been performed and all research manuscripts have been accepted for publication. The data will then be destroyed.

The statistics software packages SPSS 21 for Windows (SPSS Inc., Chicago, USA) and STATA 13.0 for Windows (IBM) were used in all statistical tests. Both per-protocol (PP) and intention-to-treat (ITT) analyses was employed for both primary and secondary outcomes. In bivariate analyses, within and between groups comparisons were made over the clinical trial period. When the outcome variables were continuous and followed a normal distribution, paired t test for related samples were performed to determine significant differences over time. Student t-test for independent samples was performed to compare the mean of the outcomes between intervention groups at baseline and follow-up review. When the outcome variables were continuous but did not follow a normal distribution as detected by normality test, Wilcoxon Signed Rank test (a non-parametric equivalence of the paired t test for related samples) was performed to determine significant differences over time. Mann Whitney U test (a non-parametric equivalence of the t test for independent samples) was employed to compare difference in the rank of outcomes between intervention groups at baseline and follow-up. When the outcome variables were categorical, McNemar test were performed to identify changes in prevalence over time and Chi-square test was employed to compare variations in prevalence between intervention groups at baseline and follow up.

Regression analyses were employed to address the effect of potential explanatory factors for outcomes. When conducting the regression analysis, method of Last Observation Carried Forward (LOCF) was employed to deal with missing outcomes at follow-up reviews. Linear regression was adopted when the dependent variables were continuous and residuals of the regression followed a normal distribution. Negative binomial regression was adopted when the continuous dependent variables had over-dispersed count data. Logistic regression was adopted when the dependent variables were binary. Since the effect of intervention on the outcomes was of our core interest, the model fitting adopted a forward-wald method with the factor of intervention fixed in the model and one of the other explanatory factors added into the model each time. Each time the coefficient of the independent variables with the smallest significant p value (significance level p<0.05) was entered into the regression model. The model fitting process was repeated again until no more independent variable had significant p values.