干预性队列研究的偏倚控制及其在中医领域的特殊性
|
摘要
作为真实世界研究方法之一,干预性队列研究具有较好的外部真实性,其证据级别仅次于随机对照试验,但最大的争议是偏倚。从研究设计、实施和统计分析的角度探讨如何控制干预性队列研究的常见偏倚。对于选择性偏倚,强调研究者的角色依然是观察者,不应干涉受试者入组意愿;清晰地界定暴露因素及强度有助于研究者判断受试者的最终暴露归属;部分特殊情况下适当采用随机化原则可减少选择性偏倚发生;可采用匹配法增加组间均衡性。对于信息性偏倚,建议采用客观、量化的指标收集资料,若难以实现,则采用"盲法"收集;在广泛收集资料的同时,需注意内容清晰易懂且不可过于繁琐,以免造成不应答偏倚。对于混杂偏倚,需要设置适当、清晰的纳入排除标准,排除已明确的混杂因素;在分析阶段,校正分析、亚组分析、多元回归分析、倾向性评分等都可用于控制偏倚。此外,中医药的干预性队列研究有其自身的特殊性,选择分中心时需注意不同地区的搭配及中医院、西医院及综合医院的搭配等。期待通过控制偏倚、优化设计以增加干预性队列研究的内部真实性及证据级别,使研究者在面对中医治疗这样个体化强、干预体系复杂、无法随机化、存在伦理学问题等情况下,转换研究思路,让更多患者因临床试验获益。
As a real-world study,interventional cohort study has good external authenticity,whose evidence level is lower only than randomized controlled trials(RCTs),but its greatest controversy is bias. This paper discussed how to control common biases from the perspective of research design,implementation and statistical analysis. In controlling selective bias,the author emphasizes that the researcher is an observer and should not interfere with the intervention assignment. Exposure factors and intensity should be clearly defined to help researchers to determine the final exposure attribution of subjects. In some special cases,the randomization principle can reduce the occurrence of selective bias. Matching method can increase the balance between groups. In controlling information bias,it is suggested to use objective and quantitative indicators to collect data. If this is difficult to achieve,then "blinded"data collection method shall be used. For wide collection of information,the content should be clear and easy to understand,in order to avoid too cumbersome to create non-response bias. In controlling compounding bias,it is necessary to properly and clearly set the inclusion and exclusion criteria. The clear confounding factors can be excluded to reduce the occurrence of confounding bias. In the analysis phase,the correction analysis,subgroup analysis,multivariate regression analysis,and propensity scores are commonly used for controlling bias. In addition,the interventional cohort study of traditional Chinese medicine has its own particularity. When choosing sub-centers,researchers should focus on the match of different regions,as well as traditional Chinese medicine hospitals,western medicine hospitals and general hospitals. The author expects to improve the internal authenticity and evidence level of interventional cohort studies by controlling bias and optimizing design. Researchers can change research ideas in case of such individualized and complicated intervention systems like traditional Chinese medicine treatment,which does not allow randomization,and often with ethical issues,and let more patients benefit from clinical trials.
As a real-world study,interventional cohort study has good external authenticity,whose evidence level is lower only than randomized controlled trials(RCTs),but its greatest controversy is bias. This paper discussed how to control common biases from the perspective of research design,implementation and statistical analysis. In controlling selective bias,the author emphasizes that the researcher is an observer and should not interfere with the intervention assignment. Exposure factors and intensity should be clearly defined to help researchers to determine the final exposure attribution of subjects. In some special cases,the randomization principle can reduce the occurrence of selective bias. Matching method can increase the balance between groups. In controlling information bias,it is suggested to use objective and quantitative indicators to collect data. If this is difficult to achieve,then "blinded"data collection method shall be used. For wide collection of information,the content should be clear and easy to understand,in order to avoid too cumbersome to create non-response bias. In controlling compounding bias,it is necessary to properly and clearly set the inclusion and exclusion criteria. The clear confounding factors can be excluded to reduce the occurrence of confounding bias. In the analysis phase,the correction analysis,subgroup analysis,multivariate regression analysis,and propensity scores are commonly used for controlling bias. In addition,the interventional cohort study of traditional Chinese medicine has its own particularity. When choosing sub-centers,researchers should focus on the match of different regions,as well as traditional Chinese medicine hospitals,western medicine hospitals and general hospitals. The author expects to improve the internal authenticity and evidence level of interventional cohort studies by controlling bias and optimizing design. Researchers can change research ideas in case of such individualized and complicated intervention systems like traditional Chinese medicine treatment,which does not allow randomization,and often with ethical issues,and let more patients benefit from clinical trials.
引文
[1]WING YK,LAM SP,LI SX,et al.Sleep-related eating disorder and zolpidem:an open interventional cohort study[J].J Clin Psychiatry,2010,71(5):653-656.
[2]PEARSON ER,FLECHTNER I,NJLSTAD PR,et al.Switching from insulin to oral sulfonylureas in patients with diabetes due to Kir6.2 mutations[J].N Engl J Med,2006,355(5):467-477.
[3]MOUNTAIN D,ERCLEVE T,ALLELY P,et al.Reatedreducing acute chest pain time in the ED:a prospective pre-/post-interventional cohort study,stratifying risk using early cardiac multi-markers,probably increases discharges safely[J].Emerg Med Australas,2016,28(4):383-390.
[4]LAU JT,LEUNG PC,WONG EL,et al.The use of an herbal formula by hospital care workers during the severe acute respiratory syndrome epidemic in Hong Kong to prevent severe acute respiratory syndrome transmission,relieve influenza-related symptoms,and improve quality of life:a prospective cohort study[J].J Altern Complement Med,2005,11(1):49-55.
[5]杨保华,李国胜,张义峰,等.经皮椎体成形和椎体后凸成形治疗骨质疏松性胸腰椎骨折疗效的回顾性队列研究[J].中国循证医学杂志,2015,15(4):377-380.
[6]BENSON K,HARTZ AJ.A comparison of observational studies and randomized,controlled trials[J].N Engl JMed,2000,342(25):1878-1886.
[7]HORWITZ RI,VISCOLI CM,CLEMENS JD,et al.Developing improved observational methods for evaluating therapeutic effectiveness[J].Am J Med,1990,89(5):630-638.
[8]BYINGTON RP.Beta-blocker heart attack trial:design,methods,and baseline results.Beta-blocker heart attack trial research group[J].Control Clin Trials,1984,5(4):382-437.
[9]韩梅,陈薇,刘建平,等.比较效果研究常用方法之二:队列研究设计[J].现代中医临床,2015,22(3):20-23.
[10]费宇彤,张颖,刘建平.再论“队列研究”在中医药临床疗效评价中的应用[J].世界中医药,2014,9(10):1261-1263.
[11]ALTMAN D,GRANATH F,CNATTINGIUS S,et al.Hysterectomy and risk of stress-urinary-incontinence surgery:nationwide cohort study[J].Lancet,2007,370(9597):1494-1499.
[12]梁万年,刘民.肿瘤病因研究中的信息偏倚及其控制[J].中国肿瘤,1995,4(7):11-12.
[13]王梅,王建华,张抗,等.中医药疗效评价队列研究的方法学质量评价[J].中医杂志,2016,57(16):1379-1383.
[14]宋永霞,洪静芳,谢伦芳,等.临床癌症患者研究中存在的偏倚及控制[J].中华护理杂志,2015,50(6):750-754.
[15]董卫.临床研究中常见的偏倚及其控制[J].中国实用内科杂志,2007,27(24):1985-1986.
[16]ROSENBAUM PR,RUBIN DB.The central role of the propensity score in observational studies for causal effects[J].Biometrika,1983,70(1):41-55.
[17]ROSENBAUM PR,RUBIN DB.Constructing a control group using multivariate matched sampling methods that incorporate the propensity score[J].Am Stat,1985,39(1):33-38.
[18]叶晓勤.基于倾向性评分的中医复杂干预临床疗效评价[J].中国中医基础医学杂志,2012,18(2):218-220.
[19]张英,杨薇,姜俊杰,等.运用倾向性评分方法分析参附注射液对肝功能影响的队列研究[J].辽宁中医药大学学报,2016,18(9):66-70.
[20]张颖,李迅,费宇彤,等.倾向指数用于中医药非随机设计临床疗效评价的思考[J].世界中医药,2014,9(10):1272-1274.
[21]WAHLBECK K,TUUNAINEN A,AHOKAS A,et al.Dropout rates in randomised antipsychotic drug trials[J].Psychopharmacology(Berl),2001,155(3):230-233.
[22]GAMBLE C,HOLLIS S.Uncertainty method improved on best-worst case analysis in a binary meta-analysis[J].JClin Epidemiol,2005,58(6):579-588.
[23]蔡泳.病例随访资料的统计分析方法:生存分析[J].上海口腔医学,2004,13(4):315-317.
[24]ZHANG T,HE WT,ZI MJ,et al.Cohort study on prognosis of patients with metastatic colorectal cancer treated with integrated chinese and western medicine[J].Chin JIntegr Med,2018,24(8):1-6.
[25]何文婷,张彤,杨宇飞,等.中医药治疗结直肠癌临床疗效Meta分析及证型分析[J].中医杂志,2018,59(22):1929-1936.
[26]刘建平.中医药临床试验的方法学问题与挑战:循证医学的观点[J].中西医结合学报,2006,4(1):1-6.
[27]张颖,李迅,费宇彤,等.倾向指数用于中医药非随机设计临床疗效评价的思考[J].世界中医药,2014,9(10):1271-1274.
[28]田磊,管欣,马爱霞.随机对照试验研究与观察性研究的系统比较[J].中国药房,2018,29(4):493-496.
[29]LEVIN M.21st century cures act requires FDA to expand the role of real world evidence[EB/OL].(2016-12-19)[2017-08-30].http://www.jdsupra.com/legalnews/21st-century-cures-act-requires-fda-to-56141/.
[30]U.S.House Energy and Commerce Committee.Text of house amendment to the senate amendment to H.R.34,tsunami warning,education and research act of 2015[EB/OL].(2016-12-25)[2017-03-21].http://www.congress.gov/114/bills/hr34/BELLS-114hr34enr.pdf.
[2]PEARSON ER,FLECHTNER I,NJLSTAD PR,et al.Switching from insulin to oral sulfonylureas in patients with diabetes due to Kir6.2 mutations[J].N Engl J Med,2006,355(5):467-477.
[3]MOUNTAIN D,ERCLEVE T,ALLELY P,et al.Reatedreducing acute chest pain time in the ED:a prospective pre-/post-interventional cohort study,stratifying risk using early cardiac multi-markers,probably increases discharges safely[J].Emerg Med Australas,2016,28(4):383-390.
[4]LAU JT,LEUNG PC,WONG EL,et al.The use of an herbal formula by hospital care workers during the severe acute respiratory syndrome epidemic in Hong Kong to prevent severe acute respiratory syndrome transmission,relieve influenza-related symptoms,and improve quality of life:a prospective cohort study[J].J Altern Complement Med,2005,11(1):49-55.
[5]杨保华,李国胜,张义峰,等.经皮椎体成形和椎体后凸成形治疗骨质疏松性胸腰椎骨折疗效的回顾性队列研究[J].中国循证医学杂志,2015,15(4):377-380.
[6]BENSON K,HARTZ AJ.A comparison of observational studies and randomized,controlled trials[J].N Engl JMed,2000,342(25):1878-1886.
[7]HORWITZ RI,VISCOLI CM,CLEMENS JD,et al.Developing improved observational methods for evaluating therapeutic effectiveness[J].Am J Med,1990,89(5):630-638.
[8]BYINGTON RP.Beta-blocker heart attack trial:design,methods,and baseline results.Beta-blocker heart attack trial research group[J].Control Clin Trials,1984,5(4):382-437.
[9]韩梅,陈薇,刘建平,等.比较效果研究常用方法之二:队列研究设计[J].现代中医临床,2015,22(3):20-23.
[10]费宇彤,张颖,刘建平.再论“队列研究”在中医药临床疗效评价中的应用[J].世界中医药,2014,9(10):1261-1263.
[11]ALTMAN D,GRANATH F,CNATTINGIUS S,et al.Hysterectomy and risk of stress-urinary-incontinence surgery:nationwide cohort study[J].Lancet,2007,370(9597):1494-1499.
[12]梁万年,刘民.肿瘤病因研究中的信息偏倚及其控制[J].中国肿瘤,1995,4(7):11-12.
[13]王梅,王建华,张抗,等.中医药疗效评价队列研究的方法学质量评价[J].中医杂志,2016,57(16):1379-1383.
[14]宋永霞,洪静芳,谢伦芳,等.临床癌症患者研究中存在的偏倚及控制[J].中华护理杂志,2015,50(6):750-754.
[15]董卫.临床研究中常见的偏倚及其控制[J].中国实用内科杂志,2007,27(24):1985-1986.
[16]ROSENBAUM PR,RUBIN DB.The central role of the propensity score in observational studies for causal effects[J].Biometrika,1983,70(1):41-55.
[17]ROSENBAUM PR,RUBIN DB.Constructing a control group using multivariate matched sampling methods that incorporate the propensity score[J].Am Stat,1985,39(1):33-38.
[18]叶晓勤.基于倾向性评分的中医复杂干预临床疗效评价[J].中国中医基础医学杂志,2012,18(2):218-220.
[19]张英,杨薇,姜俊杰,等.运用倾向性评分方法分析参附注射液对肝功能影响的队列研究[J].辽宁中医药大学学报,2016,18(9):66-70.
[20]张颖,李迅,费宇彤,等.倾向指数用于中医药非随机设计临床疗效评价的思考[J].世界中医药,2014,9(10):1272-1274.
[21]WAHLBECK K,TUUNAINEN A,AHOKAS A,et al.Dropout rates in randomised antipsychotic drug trials[J].Psychopharmacology(Berl),2001,155(3):230-233.
[22]GAMBLE C,HOLLIS S.Uncertainty method improved on best-worst case analysis in a binary meta-analysis[J].JClin Epidemiol,2005,58(6):579-588.
[23]蔡泳.病例随访资料的统计分析方法:生存分析[J].上海口腔医学,2004,13(4):315-317.
[24]ZHANG T,HE WT,ZI MJ,et al.Cohort study on prognosis of patients with metastatic colorectal cancer treated with integrated chinese and western medicine[J].Chin JIntegr Med,2018,24(8):1-6.
[25]何文婷,张彤,杨宇飞,等.中医药治疗结直肠癌临床疗效Meta分析及证型分析[J].中医杂志,2018,59(22):1929-1936.
[26]刘建平.中医药临床试验的方法学问题与挑战:循证医学的观点[J].中西医结合学报,2006,4(1):1-6.
[27]张颖,李迅,费宇彤,等.倾向指数用于中医药非随机设计临床疗效评价的思考[J].世界中医药,2014,9(10):1271-1274.
[28]田磊,管欣,马爱霞.随机对照试验研究与观察性研究的系统比较[J].中国药房,2018,29(4):493-496.
[29]LEVIN M.21st century cures act requires FDA to expand the role of real world evidence[EB/OL].(2016-12-19)[2017-08-30].http://www.jdsupra.com/legalnews/21st-century-cures-act-requires-fda-to-56141/.
[30]U.S.House Energy and Commerce Committee.Text of house amendment to the senate amendment to H.R.34,tsunami warning,education and research act of 2015[EB/OL].(2016-12-25)[2017-03-21].http://www.congress.gov/114/bills/hr34/BELLS-114hr34enr.pdf.