A Study On Repeated Measures Designs And Analysis For Clinical Trials Of Traditional Chinese Medicine

BackgroundThe preventions and treatments of diseases with Traditional Chinese Medicine (TCM) are based on the concepts of wholism and dynamics, which regards the various parts of the human body as an organic whole, and emphasizes the body being in a continuous dynamic status both in physiologically and pathologically. Therefore, the evaluations for clinical effectiveness in Clinical trials for TCM should be an embodiment of the concepts. The clinical trials of TCM were difference from the clinical research of artificial drugs, because it is on two distinct therapeutic theories. The clinical research of TCM will pay more attention to dynamical trends of clinical outcome. It should be more rational and objective that the repeated measures designs and analysis is used to assess the temporal effects on clinical outcome of treatments of diseases in the clinical trials of TCM.ObjectivesThe main interest of this study is methodological: to explore the advantages gained and problems encountered while applying repeated measures designs and analysis for the clinical trials of TCM, and in order to evaluate more roundly the dynamical trends and affecting factors of clinical effectiveness of TCM.MethodsThis study was performed its methodological practise with the example of clinical trial for chronic heart failure (CHF) to treatment with Shengmai Capsule. The clinical trial for CHF with Shengmai Capsule was designed as a double-blind randomized placebo-controlled trial with one factor of repeated measures. In the trial, we assigned 142 patients with chronic heart failure to treatment with Shengmai Capsule and 139 to placebo. Patients were required to have chronic heart failure (NYHA II-IV), previous admission for an ejection fraction of less than 45%, and to have been treated optimally with diuretics, vasodilator, digitalis, beta- blockers and angiotensin-converting enzyme inhibitors unless not tolerated. The four response variables selected for the study were measurements of left ventricular ejection fraction (EF), total scores of symptoms of TCM, quality of life (QOL) and NYHA (New York Heart Association) functional classification. Per-protocol subjects' analysis (PPS) and intention-to-treat analysis(ITT) for clinical effectiveness, safety evaluation and repeated measures analysis were performed in the trial. In data analysis for this trial, univariate analysis, multivariate analysis of variance (MANOVA), linear mixed effects model, multilevel model and generalized estimating equations (GEE) method were selected to employ. The results of PPS and ITT were compared with that of repeated measures analysis. Meanwhile, a sensitivity analysis for different repeated measures analysis and replacing missing values was dealt to, appraise applicability of these methods.ResultsThe results from PPS and ITT analysis for this trial showed that there were no statistical significant differences between Shengmai group and placebo group in all-cause mortality, incidence rate of important clinical events, measurements of EF, total scores of symptoms, QOL and NYHA functional classification. The safety evaluation in the trial also indicated that there were no significant increase at risk of adverse reactions in Shengmai group. The results of repeated measures analysis showed that time effects of EF, QOL, total scores of symptoms and changes of NYHA functional classification were high significant, but grouped effect was no significant. The resultsrevealed that there were obvious tendency in each group, which the several outcome variables were changed along with period points. There were differences among the tendencies of several variables. The temporal tendencies of EF and QOL were an approximative linearity, but the total scores of symptoms and the changes of NYHA functional classification were displayed non-linear tendencies. The results also showed that the tendencies of several variables were affected by some covariates. The result of MANOVA of repeated measures for EF, total scores of symptoms and QOL showed that age and centers were the significant covariates. The findings indicated age and centers affected the tendencies of outcome variables. Hereinto, the affecting of age factor was mostly on EF and QOL, but that of the centers factor focused on EF and total scores of symptoms. The findings from linear mixed effects modelling and multilevel modelling methods revealed that a desirable trend in temporal effect, however, there was no significant difference in the trend between two groups. The temporal tendency of EF was affected by using of digitalis and centers, that of total scores of symptoms was affected by course of CHF, and that of QOL was affected by the interaction of time point by centers. The findings from GEE method indicated that there was a clear tendency in the changes of NYHA functional classification, which was affected by baseline of functional classification. It is suitable for replacing missing values with the means in the same group through the sensitivity analysis. An affect on the results of MANOVA of repeated measures was appeared with or without replacing missing values, but the modeling methods got less impact without replacing missing values. The test power of both univariate ANOVA and MANOVA of repeated measures was higher than that of PPS and ITT analysis. Comparing between several methods of repeated measures analysis, the results from linear mixed modelling and multilevel modelling were no distinguishable change. This indicated that the results from modelling methods were reliable.ConclusionsIn summary, the power of ANOVA and MANOVA of repeated measures were higher. The repeated measures design and analysis can roundly describe and evaluate the changes over time, and analyze the tendency of changes of outcome variables and its affecting factors. Therefore, the repeated measuresdesign and analysis can entirely and dynamically evaluate the repeated measures data in clinical trials. Applying the repeated measures design and analysis to clinical effectiveness in clinical trials for TCM, it was in accord with the conception of wholism and dynamics in the therapeutical effects in TCM theories. It is important in practicality to applying the repeated measures design and analysis reasonably.