Interstellar pilot, colorectal cancer survival optimization wonderful reproduction: quality is the core of clinical trials, a high-quality and efficient clinical trial should have a scientific and rigorous program design. The thesis is the essence of clinical research, and a high-scoring SCI paper should start from the establishment of scientific research projects and research design.
On May 23, 2020, an "International Experience Sharing Meeting on the Design, Implementation and Publication of Clinical Studies" gathered top colon cancer experts, statistics and epidemiology experts at home and abroad was broadcast live online. The conference was chaired by Professor Liu Yunpeng, Director of the Department of Oncology, First Affiliated Hospital of China Medical University, and Professor Zhao Hong, Deputy Director of the Department of Hepatobiliary Surgery, Cancer Hospital of the Chinese Academy of Medical Sciences, as the moderator, to discuss and answer the statistical and epidemiological confusions of clinical research design, as well as the problems encountered in the writing and publication of SCI papers that clinical experts are keenly concerned about.
At the beginning of the conference, Professor Liu Yunpeng, the chairman of the conference, delivered a speech and pointed out the original intention and purpose of the conference - this is a research-oriented conference that is interested in solving related problems in clinical research in addition to solving the disease itself. Let's enjoy this academic feast of ace experts!
Professor Luo Sheng: Design and interpretation of interventional studies
Professor Sheng Luo of Duke University's Institute of Clinical Research explained the principles and principles of setting statistical indicators at the beginning of clinical research design through the interpretation of two clinical trial data on rigofenib. Both trials were phase II studies of ReDos (which assessed the safety and efficacy of low- or standard-dose regegofenib in patients with recurrent metastatic colorectal cancer) and reverse studies (which evaluated the safety and efficacy of oral reinterflation cetuximab versus oral reinterflux cetuximab versus oral reinterflux versus oral repurpose versus oral repatifunib in patients with previously received fluorouracil, oxaliplatin, and irinotecan wild-type KRAS mutations). The design elements and outcomes of the two clinical studies are as follows:
Professor Luo Sheng explained the statistical indicators that everyone is more concerned about in the clinic through the current clinical results of the two studies.
▍Type 1 error=0.2?
Type 1 error, which is the probability of falsely rejecting the 0 test (which can be understood as the probability of false positives). Typically, the Type 1 error is set at 0.05, but the Type 1 error in the ReDos study is set at 0.2, possibly for the following reasons: first, this is a Phase II clinical trial, and the US Food and Drug Administration (FDA) does not have high requirements for this indicator; second, the number of patients recruited (about 100 people) and financial considerations. In the standard dose group, the rate of patients who completed 2 courses of treatment in the initial treatment and the 3rd course was 45%, and the rate of the reduced dose group was 63%. Extrapolated from the calculation of its sample size, when Alpha =0.2, its sample size is 55 per group; when Alpha= 0.05, its sample size is 103 per group. That is, the sample size of the clinical enrollment study can be adjusted by adjusting the alpha value.
▍ Power values for different sample sizes
In the Reverse study, power values are defined as the probability of finding the effectiveness of R+C treatments. From the actual sample size and HR value (HR=0.61) of the study, the Power value of 0.68 can be calculated, that is, the Reverse study has a 68% probability of finding the effectiveness of the drug R+C. For this study, which is less than 80% of the usual probability, but the clinical result is positive, Professor Luo Sheng said that the phase II clinical trial is lucky, but it is also reasonable.
Professor Yuan Ying of Anderson Cancer Center then further explained the Type 1 error value in the ReDos study, that is, ReDos was a unilateral study, and it set the Type 1 error value to 0.2, in order to reduce the sample size required for study recruitment. At the same time, it also answered the problem of blinding mentioned by the host Professor Zhao Hong at the beginning of the meeting. At the following roundtable, Professor Wang Xicheng of the Department of Gastroenterology and Oncology of Peking University Cancer Hospital, Professor Hao Jing of Qilu Hospital of Shandong University, and Professor Li Qiu of West China Hospital of Sichuan University participated in the discussion of statistical indicators in the clinic. In a Phase II clinical study, how is the Type 1 error value considered? How are sample sizes calculated for clinical studies using multiple endpoints?
Discussion points:
At the beginning of the clinical study, the number of patients actually recruited was considered from the upper limit of the number of patients recruited; the Type 1 error value setting needed to be ≤ 0.2; the P value needed to be less than the type 1 error value originally set. Tip: Low sample size and low drug efficacy values were found.
Studies in which the clinical data looked very good but showed negative results (the reason for which the P-value did not reach the Type 1 error value originally set) was due to the fact that the study had multiple study endpoints, and for this endpoint set a low Type 1 error value, or insufficient sample size.
For the calculation of clinical trial sample sizes with multiple study endpoints, the common calculation method is to calculate the sample size by using a method that is valid for at least one study endpoint. At the same time, the adjustment, distribution and recycling of Alpha were further discussed and exchanged.
Professor Zhang Hui: A New Approach in the Age of Big Data: Real-World Research
Professor Zhang Hui of Northwestern University in the United States from evidence-based medicine to the rise, development and current development of real-world research shows us the role and significance of real-world research.
▍RCTs are the highest level of evidence-based medicine
David Sackett, the main founder of evidence-based medicine and an internationally renowned clinical epidemiologist, once defined evidence-based medicine as": "Prudent, accurate and wise evidence obtained from evidence-based medical application to determine patient treatment." "That is, the intensity of drug promotion is reproduced with evidence strength, and the higher the evidence intensity, the greater the drug promotion intensity." In evidence-based medicine, the highest-level study was RCT (randomized controlled study).
▍ The rise and development of real-world research
A 2013 paper published in the NEJM (New England Journal of Medicine) by American statisticians pointed directly to the shortcomings of clinical trials, and the study showed that based on observational data or stronger clinical trial evidence can be obtained, which also stimulates the rise of real-world research. In order to further unify everyone's understanding of clinical trial evidence, in 2016, the US FDA set off a discussion on clinical trial evidence. The boom in real-world research in recent years is largely due to changes in human lifestyles.
In 2016, the US FDA promulgated the "21st Century Treatment Act", affirming the significance and utility of real-world research and approving that drugs studied by the real world can be applied for marketing.
▍ Real world role
1. Data are widely sourced
2. Fill the current clinical trial gap
Maximize the merit of data/evidence from settings that are more reflective of clinical practice;
Fill the gaps in current standard clinical trials.
At the following roundtable, Professor Yuan Ying, Professor Song Yan of the Department of Internal Medicine of the Cancer Hospital of the Chinese Academy of Medical Sciences, Professor Liu Zimin of the Department of Oncology of the Affiliated Hospital of Qingdao University, and Professor Zheng Liqiang of Epidemiology and Health Statistics of Shengjing Hospital affiliated to China Medical University participated in the discussion on real-world research in the clinic. Which data is more credible for the large sample size of a real-world study compared to the relatively small sample size of an RCT study? Is it necessary to set up controls for retrospective studies, how are the corresponding sample sizes calculated, and how are the resulting partialities resolved? Does data selectivity bias have a large impact on sample size selection? Discussion points:
Real-world studies are complementary to RCT studies. For the study of the effectiveness of new drugs, the RCT study is the gold standard, and its findings are more authoritative; for drugs with known efficacy, real-world studies are more representative if you want to further understand the actual efficacy of the drug in a larger disease population.
Retrospective studies still require control, and statistical matching can be used to calculate the required sample size. The resulting bias should be avoided as much as possible, and cannot be optimized by statistical methods such as multivariate regression.
The effect of data selectivity bias on determining the sample size can be partially mitigated or eliminated by statistical methods. (1) For the missing data, in the case of small impact deviation, the impact of the missing data will not affect the test results through comparative analysis; (2) in the case of large influence deviation, the missing data can be filled, and after the statistical model is established, the data description analysis can be carried out; (3) the existing sample size and the filled sample size are typed separately, and the consistency before and after the data selection is provided through comparison.
Professor Wei Qingyi: Timing and techniques for writing SCI articles
Professor Wei Qingyi
Professor Wei Qingyi of Duke University School of Medicine statistically analyzed the proportion of these top 100 papers in each country through a review of the epidemiology of colorectal cancer (CRC), the number of CRC annual publications and the top 100 papers cited, and the top 100 journals that published the top 100 CRC papers.
Professor Wei Qingyi also pointed out at this conference that the publication of SCI papers is not achieved overnight, and a high-scoring SCI paper should start from the establishment of scientific research projects and research design. At the same time, we must always pay attention to international hot spots in the field of research and pay attention to case collection.
Finally, Professor Wei Qingyi called for how to start from the actual situation of colorectal cancer research in China, collect data, solve clinical problems, solve clinical problems, solve problems for colorectal cancer patients, and return to the essence of clinical research.
At the final roundtable, Professor Cao Baoshan of the Department of Cancer Chemotherapy and Radiology of peking university third hospital, Professor Chen Xiaobing of the second department of gastroenterology of Henan Cancer Hospital, and Professor Yan Dong of the Cancer Center of Beijing Luhe Hospital affiliated to Capital Medical University participated in the discussion on the relevant issues related to the writing and publication of SCI papers. How to deal with a paper that has been returned that requires substantial revision, and what are the techniques for revision? What are the tips for publishing SCI articles for rare cases? What is the experience of writing SCI article content? Discussion points:
SCI Minor Experience: For papers that are returned and require substantial revision, the probability of being accepted is generally higher. For the part that is requested to be revised, we need to honestly answer the questions raised in the reverted process, and give a clear explanation, do not avoid talking about it; we must also argue for the questions that we think are correct but are questioned.
For rare cases, given the scarcity of cases, in order to obtain high-quality SCI articles, we should expand our own data samples in the content of the papers: integrating relevant data from other hospitals, using public databases for analysis, etc.
SCI articles require that the experiment itself be innovative, or analyze using innovative statistical methods, which is a necessary condition for the article to be received. In addition, attention should also be paid to the expression of text content, the professionalism and accuracy of language description, etc.
In the final free discussion session, Professors Luo Sheng and Professor Wei Qingyi explained in detail and in a simple way how to solve the data drift in the retrospective study, and discussed the elements of how to obtain high-scoring SCI articles by using public databases to study how to obtain high-scoring ARTICLEs in SCI by using public databases to discuss with Professor Zhao Hong of cancer hospital of Chinese Academy of Medical Sciences, Professor Qu Xiujuan of the First Affiliated Hospital of China Medical University, and Professor Lu Ming of Qilu Hospital of Shandong University. Discussion points:
In fact, a considerable number of clinicians need to build their own databases, from the design of trials, to the collection of data, to the analysis of results, quality control in clinical research is very important for high-scoring articles.
Another important problem in using public databases to write articles is "verification", comparing their existing data or similar databases with data in public databases, and presenting conclusions will increase the credibility of the results.
If you can't find an external database to verify, you can try to "halve" the same database, one for modeling and one for verification. Validation in the same database has a lower confidence level than validation from an external database.
The meeting was enthusiastic, and the experts attending the meeting said that they would benefit from it a lot and still not finished. Professor Liu Yunpeng, chairman of the conference, also said that he hoped to subdivide the problems again and conduct more in-depth discussion and learning.