Research On Methodology Of Cohort Study Of Curative Effect Evaluation Of Traditional Chinese Medicine

Purposes:Traditional Chinese medicine(TCM) is characterized by treatment based on syndrome differentiation, and in clinical practice is always applied in complex or integrated therapy. Chinese medicine practitioners make medical decisions based on the patients’ conditions and their preferences, and the treatment will always be changed with the progress of disease or syndrome changes. How to carry out researches which could reflect TCM characteristics has been a critical problem to solve. Currently, TCM clinical researches mainly focus on randomized controlled trials(RCTs), which are as “golden standards” of new drug research and development, however, there are some limitations in RCTs when evaluating complex therapies. These RCTs are far from the real world TCM clinical practice, so we need observational studies to provide more medical information. Cohort study is one type of observational methods, and the evidence level is just lower than the RCTs. The principle of therapeutic cohort study is: participants divided into exposure group and control group according to their different treatments; or according to different treatment intensity or medicine doses, and then to compare the incidence rate of different groups after a certain period. Recently, cohort study has gradually been taken seriously in TCM field, and has been widely used by significant research projects in China. The textbook of epidemiology mainly focus on introducing how to design a cohort study about exploring the risk factor of the diseases, but fewer guidance or suggestions about how do design a cohort study protocol. Many TCM clinical researchers are often confused, and some of them even use the protocol of clinical trial instead, for some cohort study details they cannot separate from clinical trial. Thus, a methodology guidance is urgent needed which integrate cohort study with TCM specific items to help the researchers to design high quality study, so as to provide more and better evidence for clinical practice.Based on what is mentioned above, the aims of this study are: firstly, to evaluate the methodology quality of current traditional medicine cohort study through bibliometrics research; Secondly, to draft a “Methodology suggestion for cohort study protocol in TCM therapeutic effectiveness evaluation”, for clinical researchers referencing when designing the cohort study; Thirdly, to introduce the statistics analysis methods of cohort study, which are different from the common clinical trial’s, always need adjustment for covariates. The final aim of this study is to provide a methodology suggestion for TCM therapeutic cohort study, enhance the clinical research quality, and get more and better evidence for clinical practice.Methods:1 Evaluate the methodology quality of cohort study in traditional medicineWe will systematically search the electrical databases for therapeutic cohort studies used in traditional medicine field, including CNKI, VIP, Sino Med, Wanfang, Pubmed, Cochrane Library, from the inception to 31 st December 2015. The study language will be limited in Chinese and English, no matter whether be public published. Two researchers independently read and selected the study, and extracted the information, including: author, published year, journal, impact factor of journal(English journal: 2014 SCI impact factor; Chinese journal: CNKI comprehensive impact factor), research country or region, funding, disease and classification according to the “International Classification of Disease(the 10 th edition)”(ICD-10). We will evaluate the cohort study methodology quality according to the Newcastle-Ottawa Scale(NOS). NOS for cohort study is composited by three parts: selection(4 scores), comparability(2 scores) and outcome(3 scores), totally 9 items. The maximum score is 9, and the higher the scores are, the better the methodology quality is. In addition, we added some other methodology items, including: sample size estimation, inclusion and exclusion criteria, treatment protocol, primary outcome, follow up, subgroup analysis, sensitive analysis, cohort migration, informed consent, ethical approval and so on. According to above key points in cohort study design, we will reclassify the study type.2 Establish “Methodology suggestions for cohort study protocol in TCM therapeutic effectiveness evaluation”We will systematically search the methodology guidance, textbooks and literatures about observational study and cohort study, and adequately understand the design principle and key points. And we will also search the methodology about the specialty of TCM clinical research, to get knowledge of the differences between TCM and western medicine. Based on the current TCM cohort study situation from the first part of our research, we would compose the draft of the suggestions for better design of cohort study in TCM therapeutic evaluation. We will invite 6 senior experts in methodology and TCM fields to review and revise the draft. We send the papery manuscript or email the electronic edition to the experts, and revise the manuscript after referring their feedback.3 Key points of TCM cohort study statistic analysesThe statistic analyses plan(SAP) is one part of research protocol and could be published. In this part, we will explore the content which should be included in the SAP. We conduct statistic analysis based on a completed TCM cohort study data, and apply some analysis of adjust the covariates. As the baseline is not comparable between groups, we will use stratification analysis and propensity score matching(PSM) to analyze the data. For the missing data, we will use repeated measure ANOVA and generalized estimating equations(GEE), generalized linear mixed model(GLMM) to solve the repeated measured continuous variables, and compare the results of different statistic models; we will use Logistic regression and GEE, GLMM to analyze the repeated measured binary variable, and compare the results.Results:1 The situation and methodology quality of cohort study in traditional medicineTotally, we included 152 cohort studies in traditional medicine therapeutic effectiveness evaluation, with 39 studies published in English covering 14 countries or regions with 16 traditional medicine therapies, such as: Chinese herbal medicine, integrated TCM and western medicine, homeopathy and so on; 113 ones published in Chinese covering 16 provinces and cities all were evaluated TCM therapeutic effectiveness, mainly in Beijing, Guangdong and Shanghai. The number of cohort study in traditional medicine was increasing year by year, and reached the peak in 2013. We totally included 190339 participants; the mean sample size was 5069 in English studies, and 190 in Chinese studies. 93 studies followed up a period after treatment, 29 in English and 64 in Chinese. The methodology quality of the 152 included studies were general poor. The NOS scores were manly 4-7(138 studies, 90.97%), and the mean score was 5.32±1.27. The quality of English published studies was higher than the Chinese ones. The English studies were 5.92±1.31, and the Chinese ones 5.12±1.19, which showed significant difference between them(t=3.558, P=0.001). We reclassified the type of the included studies: 78 were really cohort studies(37 in English; 41 in Chinese), and 19 ones in them were databases studies(7 in English, 12 in Chinese); five Chinese studies mentioned both “cohort study” and “random allocation” in the text, which acutely should be RCTs; 69 studies didn’t have the characteristics of observational cohort study, for example, they excluded the participants of poor compliance or changed the treatment protocol, or selected the participants according to strict inclusion and exclusion criteria to evaluate the specific effectiveness of simple treatment method, which should be clinical controlled trials(2 in English and 67 in Chinese).2 Methodology suggestions for cohort study protocol in TCM therapeutic effectiveness evaluationIn TCM clinical practice, cohort study could help to find the dominant disease in TCM, or evaluate the effectiveness and safety of treatment, or explore the factors that may affect the treatment effectiveness. The researchers could choose the study type according to their research purpose and the data situation, such as: prospective, or retrospective, or ambispective cohort study, or registry study or database study. The sample size estimation should base on the formula of cohort study, and the researchers should enlarge the sample size sometimes when considering factors like loss to follow up or cohort migration. In cohort study, the groups being compared should be as similar as possible with respect to factors may be related to the disease except the exposure factor. Therefore, the subjects both in the exposure and the control groups should be selected from the same source population. Cohort study should simplify the inclusion and exclusion criteria, and should not reject the subjects of bad treatment compliance. Subjects can be effectively continuous sample from the hospital, be members of a community, or from a doctor’s individual practice. The participants in exposure group made their own choice to get into this group, or their clinicians chose based on the participants disease situation, but not the researchers assigned the group. For the definition and measure of exposure, the cumulative dose, or total amount of exposure over a specified time period, is often optimal for adequately defining exposure. There is no placebo group in cohort study, and also no blinding of clinicians and participants, however, blinding of the outcome evaluators is needed. Usually, the primary outcome may focus on terminal outcomes, such as survival rate, mortality. Intermediate or surrogate outcomes as primary outcomes in cohort studies, may lead to potential bias that the researchers additionally added the unnecessary medical examination project in order to prove the therapeutic effectiveness sometimes. One thing needs to be mentioned is that the researchers should pay equivalent attention and use the same measurement across different exposure groups. The quality of follow up will influence the results in TCM cohort study, so the duration of follow up should be adequate, and the follow up information should include exposure factors, outcomes, and confounding factors. Multiple pathways could be used for contacting with the participants, e.g. Telephone, the patients revisit the hospital, internet questionnaire or Wechat or QQ software. The protocol of the TCM cohort study should be submitted to the local Ethics Committee for ethics approval, and the participants should sign the informed consent. The protocol of the cohort study could be registered in the international website which could control the selective reported outcome bias and publication bias, and improve the research quality.3 Adjusting covariates in TCM cohort study statistic analysesThis study introduced the content and statistical methods which should be included in the TCM cohort study SAP. The SAP of cohort studies were different from that of clinical trials, as the one way statistics analysis methods were not suitable for large sample, multivariable cohort study, but the adjusting covariate methods were needed, e.g. Logistic regression and Cox proportional hazard model for analyzing confounding factors. When the baseline of different groups were not comparable, propensity score could be used, and for the missing data, generalized estimating equations(GEE) and generalized linear mixed models(GLMMs) could be used. Take data of a TCM prospective cohort study as an example, through our statistical analysis we found that statistical result was different with or without adopting stratification analysis, so did the propensity score matching analysis. We also compared the statistic results of repeated measures ANOVA with GEE and GLMM for continuous variable with lots of missing data, the statistic results were also opposite to each other.Conclusions:1 The methodology quality in traditional medicine needs to be improvedThis study showed that the cohort study design has been increasingly adopted in TCM, however, there were some shortcomings in traditional medicine cohort studies. Currently, NOS methodological evaluation showed the methodology quality in traditional medicine cohort studies were generally poor, and the Chinese published studies were worse then those published in English. As there is no guidance for clinical researchers to design therapeutic cohort study, some researchers were easily confused cohort study with clinical trials, even misused. Thus, it is urgently needed to form a methodology suggestion and to strengthen cohort study training, in order to enhance innovation capability, and ensure clinical research quality.2 Formulate “Methodology suggestions for cohort study protocol in TCM therapeutic effectiveness evaluation”We combined TCM with modern epidemiology, and formulated “Methodology suggestion for cohort study protocol in TCM therapeutic effectiveness evaluation”, which covered 10 items including research design, participants selection, diagnosis and inclusion/ exclusion criteria, exposure definition and measurement, sample size evaluation, outcome, follow up, ethical approval and consent inform, statistic analyses, and protocol registration. This suggestion will guide the researchers to design high quality cohort study, and will contribute to the evidence used in healthy care and clinical decision.3 TCM cohort study statistics analyses needs adjusting covariatesOur study introduce the contents that should be included in the TCM cohort study SAP: statistic set ascertainment, data verification, missing data handling, participants description, baseline comparability, statistic analysis methods, safety and drug combination. Our research showed inappropriate statistics analyses may induce incorrect conclusions, so the analyzers should consider the outcomes, exposure, covariates and statistic models to choice proper statistical methods.