提高中草药随机对照试验的质量Ⅰ：临床试验设计和方法学

1、 提高中草药随机对照试验的质量：临床试验设计和方法学 作者：卞兆祥，李幼平，刘良，吴泰相，缪江霞，关家伦，宋丽 【关键词】 ,随机对照试验 摘要 目的：通过对中草药临床随机对照试验的设计及方法学进行质量评价，探讨如何提高中草药临床试验的质量。方法：文献检索2017年7月前发表于Cochrane图书馆的中草药系统评价共11篇，包含167个中草药临床随机对照试验。质量评价方法采用修订版CONSORT声明9项指标以及中草药剂型及质量控制标准指标。结果：所有167个临床试验都含有试验目的、方法、第1结局指标、统计学方法及中药剂型；其中163个临床试验说明了研究对象的纳入标准，只有26个临床试验说明了研

2、究对象的排除标准；只有14个临床试验详细说明了随机序列的产生方法；4个临床试验提及了随机分配隐藏；绝大部分的临床试验属于开放性的，只有%的临床试验采用了盲法设计；只有1个临床试验在试验前进行了样本含量的计算；在中草药剂型方面，%的临床试验使用的是汤剂或中药茶包，只有1个临床试验提及了制剂的质量控制。在167个临床试验中，所有质量评价指标的涉及率只有%。结论：现阶段中草药临床随机对照试验的质量还很低。建议：试验设计者及实施者必须接受正规的临床试验基础知识的培训；推荐采用临床试验设计流程图，逐一解决临床试验过程中的关键问题；在方案正式实施前进行预试验，并根据预试验的结果对临床试验设计方案进行调整；

3、对临床试验设计的最终方案进行注册登记，并预先发表临床试验设计方案；广泛开展国际合作，特别是与对中医药研究感兴趣的国际知名学术研究机构进行合作，以提高中草药临床研究的质量。 关键词 随机对照试验; 中草药; 方法学; 质量评价 Improving the quality of randomized controlled trials in Chinese herbal medicine, part : clinical trial design and methodology ABSTRACT Objective: To discuss the quality of randomized con

4、trolled trials in Chinese herbal medicine with respect to design and methodology, and provide suggestions for further improvement in future clinical trials. Methods: A search of the Cochrane Library was conducted to identify RCTs of CHM on line in July 2017. Quality of the RCTs was assessed using a

5、11item checklist modified from the revised CONSORT statement, with 2 items specific to CHM . Results: The search yielded 167 RCTs that were selected for assessment. All trials included statements about the interventions, objectives, primary outcome design, statistical methods, and herb preparation f

6、orm. Although 163 trials reported inclusion criteria, exclusion criteria were only reported in 26 trials. Fewer than 10% of trials clearly stated the random allocation sequence generation methods, and only % mentioned allocation concealment. The vast majority of trials were openlabel, while only % u

7、sed blinding. Almost half administered the CHM intervention as a tea or decoction. Only one trial reported a sample size calculation, and a single trial discussed quality control of the CHM intervention. Conclusion: The overall methodologic quality of RCTs in CHM was poor. It is essential to improve

8、 the design of future RCTs in this clinical area. Recommendations: Investigator conducting RCTs should have formal training about clinical trial design; A flow chart is recommended to ensure that all essential steps of clinical trial design are included. Conducting pilot studies prior to RCTs may he

9、lp improve their design; Registration of clinical trials and publishing their protocols prior to enrolment may reduce publication bias and solicit peer reviews of the proposed design; Collaboration between CHM investigators and traditional medicine academic research centers interested in integrative

10、 medicine may lead to quality improvement of RCTs of CHM. KEY WORDS randomized controlled trial; Chinese herbal medicine; methodology; quality assessment 1 INTRODUCTION Chinese herbal medicine is becoming increasingly popular in industrialized nations as one form of “alternative” or “complementary”

11、medicine. In order to bring it fully into the conventional medical systems of the worldand thereby utilize its considerable benefitsevidence of the safety and efficacy of herbs and herbal products are necessary. The question then is “Where can one find such evidence?” Both randomized controlled tria

12、ls and systematic reviews are commonly thought to provide the strongest level of evidence regarding treatment efficacy of competing therapeutic interventions. The credibility of the evidence to support a treatment approach such as CHM therefore depends on the quality of RCTs. Previous reports1,2 hav

13、e showed that systematic reviews in complementary medicine involving CHM lack high quality RCTs to provide clear evidence of efficacy. In fact, the quality of RCTs with intervention of CHM as a modality has been a topic of discussion for long time3,4, though no evidence was available to make a defin

14、itive judgment on the topic. If, as some may suspect, the quality of RCTs regarding CHM is not in fact satisfactory, a discussion must then follow regarding recommendations for improving this situation. We have therefore proposed a fourpart series of articles which focus on the four basic elements o

15、f RCTs in CHM: clinical trial design and methodology, control group design, quality control of CHM used in RCTs and reporting format of RCTs in CHM. Welldesigned RCTs were once widely recognized as providing the strongest evidence of the effectiveness of health care interventions. With the developme

16、nt of systematic review and usage of metaanalytical techniques, systematic reviews of RCTs are now thought to provide the best level of evidence about the effectiveness of an intervention5. However, such claims are founded on the assumption that RCTs are of sufficient quality, especially with respec

17、t to the clinical design and methodology, and that the details are reported clearly. In fact, the methodology of RCTs determines how its results can be interpreted and the extent to which its results can be trusted. Previous studies have shown that errone123456下一页 ous conclusions can be drawn based

18、on misinterpretations of a studys design and limitations6,7. Any inadequate methodological approaches, such as patient selection, sample size calculation, randomization procedure, outcome assessment, handling of dropouts and followups, will threaten the validity of a clinical trial with potentially

19、exaggerated treatment effects8. In order to assess the quality of the clinical design and methodology in studies of CHM, the goal of this study was to review the quality of relevant RCTs in CHM, and to provide recommendations for improving them in the future. 2 MATERIALS AND METHODS Identification o

20、f randomized trials We identified RCTs related to CHM from the Cochrane Library database of systematic reviews in July 2017. The search strategy used was as follows. Step 1: “Chinese herbal medicine” was used as a search term, which yielded 25 systematic reviews. Step 2: Search records were reviewed

21、 for their relevance to CHM. There were 11 reviews relevant to CHM that were included , and 14 reviews that were not related to CHM and thus excluded . Step 3: Primary studies of RCTs of CHM from the 11 reviews were tabulated and crossreferenced to eliminate duplicates. This produced a list of 167 R

22、CTs and 2 quasiRCTs; the latter were excluded due to their design. This list of 167 RCTs was the basis for further quality assessment. Assessment of methodology quality Methodologic quality was defined as confidence that the trials design, conduct, analysis, and presentation minimized or avoided bia

23、ses in the trials findings9. In this paper, we adopted a nineitem checklist related to the methodology of RCT from the revised consolidated standards of reporting trials statement10, and added two items especially for CHM involving preparation form of herbs and quality control of CHM. Thus, the chec

24、klist was composed of eleven items . Table 1 Included Cochrane Library systematic reviews about Chinese herbal medicine Table 2 Excluded Cochrane Library systematic reviews about Chinese herbal medicine Table 3 Checklist of RCTs with Chinese herbal medicine Data extraction and analysis Two observers

25、 assessed the quality of RCT methodology according to the checklist described above. All disagreements due to inaccurate data extraction were resolved through further verification of the original articles. In all cases, consensus between the two observers was achieved prior to the analysis. The data

26、 from two reviewers were entered into an Excel file for analysis. 3 RESULTS We identified 167 RCTs related to the use of CHM for 11 different conditions. Table 4 reports the basic characteristics of these RCTs, including their indication, participants , number of CHM interventions tested, as well as

27、 the journal and year of publication. A total of 18 058 participants were studied in 167 RCTs, with inpidual trials enrolling between 20 and 374 of them. Eighteen trials not conducted in Mainland Chinawere published in English journals, while 149 RCTs conducted in Mainland Chinawere published in Chi

28、nese journals not listed in the 2017 edition of Science Citation Index & Journal Citation Reports. Table 5 reports the 11item checklist mentioned in the 167 RCTs. Overall, % of the items on the checklist for all studies were reported in the articles. Items 2, 3, 10 in the checklists were fully met,

29、while some were partially met, and others were completely omitted . All 167 trials had an explanation about interventions, general objectives, primary outcome design and preparation forms of herbs used. Although 163 trials reported inclusion criteria, exclusion criteria were only mentioned in 26 tri

30、als. Only four trials chos e the secondary outcome for evaluation of intervention. One trial reported a priori estimation of sample size. Only % of trials clearly stated sequence generation methods. Only 4 trials mentioned randomization concealment. Only 22 trials implemented blinding in the trials,

31、 while more than 86% were open trials. For those single blind and double blind trials, not all trials gave clear explanations about the blinding method. Only one trial mentioned quality control of the CHM intervention, and this topic will be discussed further in Part of this fourpart series of artic

32、les on RCTs in CHM. Table 4 Basic characteristics of included RCTs Table 5 Breakdown of all items related with methodology 4 DISCUSSION Our review discovered that many RCTs of CHM are of poor methodological quality, much like early RCTs from other disciplines in integrative medicine, such as homeopa

33、thy and acupuncture4. Despite this overall weakness, most RCTs did adequately report the specific objectives of their study, as well as the interventions, primary outcomes and preparation forms of the herbs used. Other aspects of methodology, such as selection of participants, sample size calculatio

34、n, randomization, allocation concealment and blinding, were mostly ignored. A discussion of the importance of these characteristics is as follows. Selection of participants The selection of participants serves the purpose of ensuring that all participants are equal in terms of diseases, ensuring tha

35、t the findings in the RCTs accurately represent what is for the interest of the population, and helping medical practitioners decide whether RCT results can be generalized to the patients that present to their office. Absent strict inclusion/exclusion criteria, the reader may not know whether the tr

36、ial results can reflect the true effect of interventions in the target population, and whether the trial results can Sample size calculation Of the 167 RCTs reviewed, only 1 conducted a priori sample size calculation. An increasing number of healthrelated journals are requiring RCTs to describe this

37、 process in accordance with the CONSORT statement. This provides readers with the details used to perform the calculation, and helps to ensure that trials will be large enough so that the investigators can observe the differences they are interested in detecting. If a sample size is too small, then

38、the research might not detect the presence of true difference between two different interventions , and the study might be a waste of resources and potentially unethical. On the other hand, a sample size must not be excessively large so as to waste resources, especially time spent in recruiting and

39、screening potential participants, nor unnecessarily expose too many participants to an experimental intervention. There are seven basic elements required to conduct a sample size calculation for a RCT in which efficacy is compared on a numeric outcome between control and treatment groups: design the

40、 study to meet the specific aims; set acceptable limits for type and type errors; determine the minimum clinically meaningful difference and variance between groups from prior literature or expert opinion ; select whether the test will be onesided or twosided ; estimate the retention rate at the fin

41、al followup point and adjust sample size accordingly ; calculate the required number of study participants in each group; revise study parameters as required12,13. Slight variations of and , and expected effect and variance in both groups will affect the final sample size calculation and therefore t

42、rial costs. For example, reducing , increasing power, and reducing the expected difference between groups will increase the sample size14,15. The appropriate effect size varies widely between studies, since it should represent the smallest effect that would be regarded as clinically meaningful and i

43、mportant. This last point must be emphasized since studies that enroll many participants may report statistically significant differences that are too small to be of interest to clinicians . Reporting these differences between the treatment and control groups may give a false impression that an inte

44、rvention is beneficial. Other factors to consider for this step include disease population, treatment cost, funding for trials, etc. Therefore, estimation of the effect size of interest should reflect both clinical acumen and the potential publichealth effect. Clinical investigators must participate

45、 in this process since statisticians are not likely to generate useful sample calculations without the required data12. The few RCTs that reported calculating the sample size did not give the method and details used to perform this calculation, severely limiting its usefulness. It is necessary that

46、sample size should be calculated before the clinical trial begins, and calculation methods should be reported clearly. Randomization Randomization is a powerful tool to assign a target sample to treatment groups balanced on all possible known and unknown confounders, thus helping to avoid systematic

47、 bias. The quality of results from RCTs is highly dependent on the quality of randomization. The important factor in the process of randomization is the allocation sequence generation; conversely, sequence generation methods are used to judge the quality of randomization. Unfortunately, our data sho

48、wed that just % trials provided clear statements about sequence generation, while % merely stated that participants were randomized without explanation as to how the allocation sequences were generated. For those 18 trials published in English journals, only 5 trials stated clearly about the sequenc

49、e generation methods. This situation is similar to other medical specialties16,17. For example, in periodontology, although 91% of trials were described as randomi Allocation concealment Undoubtedly, randomization is a powerful method to produce balanced treatment groups, while allocation concealmen

50、t is the critical factor to ensure the success of randomization. Random sequence generation merely means that the group assignment sequence of a participant was generated randomly, but does not ensure that the allocation sequence was consistently followed during implementation. In order to allocate

51、the participants to balanced treatment groups, the sequence should be concealed to investigators, participants, and all other study personnel. If not, selection bias may be introduced, whereby the treatment assignment is no longer truly random and an imbalance in prognostic factors may occur between

52、 treatment groups. Previous studies have shown that inadequate or unclear allocation concealment can exaggerate clinical effects up to 40%, especially in poorly conducted trials6,7,20, and it can also cause greater heterogeneity in results21. Thus proper randomization should involve both random sequ

53、ence generation and complete implementation of that sequence to minimize bias. Our study reported that although 167 trials claimed to be randomized, only % reported the methods for allocation concealment, compared with 17% trials reporting allocation concealment in periodontology16. The reason why t

54、he rate of randomization concealment is so low is unknown, though some researchers believe that with blindingand especially double blindingallocation concealment is not necessary. But blinding and allocation concealment are different22. Blinding concentrates on preventing study participants and pers

55、onnel from determining the group to which participants have been assigned ; it safeguards the sequence after allocation. In contrast, allocation concealment concentrates on preventing selection and confounding biases; it safeguards the assignment sequence before and until allocation. Some researcher

56、s may not understand the two concepts. Whatever the reason, RCT investigators need to understand these two principles and the rationales for using them. Commonly used methods to conceal allocation include calling a central, coordinating office for each patient assignment at the time that the patient

57、 presents for study inclusion; using sequentially numbered, opaque sealed envelopes; and using numbered bottles or containers23. Blinding Why is blinding necessary for RCTs? Briefly, the reason is related with the aim of clinical trial: to find out the objective efficacy of the target intervention. In order to maintain this objectivity, we should make sure that the results were not diluted or misled by the subjective preferences from the participants, investigators, or ass