1、2024/3/24,clinical study design for postgraduate 2003,臨床科研設(shè)計(jì)方案,四川大學(xué)華西臨床醫(yī)學(xué)院臨床流行病學(xué)教研室、兒科學(xué)教研室,萬(wàn)朝敏,2024/3/24,clinical study design for postgraduate 2003,,,,設(shè)計(jì)和實(shí)施,結(jié)論,研究問(wèn)題自然界的真實(shí)性,研究計(jì)劃 研究的真實(shí)性,研究實(shí)施研究中的發(fā)現(xiàn),,,design,

2、implement,,,infer,infer,科研設(shè)計(jì)、實(shí)施及結(jié)論過(guò)程*,*Diagram from Designing Clinical Research S. Hulley, S Cummings,ERRORS,,,ERRORS,,,2024/3/24,clinical study design for postgraduate 2003,前言,保證研究的真實(shí)性和可靠性，設(shè)計(jì)和實(shí)施同等重要。沒(méi)有哪種統(tǒng)計(jì)方法能挽救瑕疵設(shè)計(jì)的真實(shí)

3、性和可靠性。,2024/3/24,clinical study design for postgraduate 2003,臨床科研設(shè)計(jì)方案論證強(qiáng)度的影響因素,隨機(jī)化 randomization對(duì)照 control時(shí)向 direction盲法 blinding,,2024/3/24,clinical study design for postgraduate 2003,隨機(jī) Randomization,每一參與者有同

4、等的機(jī)會(huì)接受干預(yù)措施，使已知和未知的影響因素一致減少研究者和被研究者的主觀干擾,2024/3/24,clinical study design for postgraduate 2003,方法,簡(jiǎn)單隨機(jī)分層隨機(jī)區(qū)組隨機(jī)系統(tǒng)隨機(jī)抽樣多級(jí)隨機(jī)抽樣半隨機(jī),2024/3/24,clinical study design for postgraduate 2003,分層隨機(jī),選擇影響治療結(jié)果的因素作為分層變量分層不宜過(guò)多,一層組數(shù)

5、為4，兩層組數(shù)為16每組最好大于50例，精確性好 避免調(diào)整基線,2024/3/24,clinical study design for postgraduate 2003,分層隨機(jī)與分層分析,2024/3/24,clinical study design for postgraduate 2003,區(qū)組隨機(jī)法,對(duì)照組和治療組比例相同時(shí)間影響因素少使用靈活-區(qū)組可大可小,2024/3/24,clinical study desig

6、n for postgraduate 2003,隨機(jī)的特點(diǎn),基線一致性, 避免選擇性偏倚等量原則,2024/3/24,clinical study design for postgraduate 2003,使用隨機(jī)原則時(shí)應(yīng)避免的幾點(diǎn),半隨機(jī)無(wú)隱匿（non-concealment）的方法，減少測(cè)量偏倚自我管理分配程序非正規(guī)的隨機(jī)和無(wú)監(jiān)督的分配方法,2024/3/24,clinical study design for postgr

7、aduate 2003,有關(guān)隨機(jī)的錯(cuò)誤概念,把“隨意”當(dāng)作“隨機(jī)”隨機(jī)確保研究組的可比性基線之間的差異是隨機(jī)失敗的標(biāo)志“隨機(jī)”可以檢驗(yàn)非隨機(jī)分配的治療研究是不可靠的,2024/3/24,clinical study design for postgraduate 2003,臨床科研設(shè)計(jì)方案論證強(qiáng)度的影響因素,隨機(jī)化 randomization對(duì)照 control時(shí)向 direction盲法 blindin

8、g,,2024/3/24,clinical study design for postgraduate 2003,對(duì)照,比較組間的差異類型： - 同期對(duì)照 (concurrent control) 隨機(jī) 非隨機(jī) - 自身對(duì)照 (self control) 同期 不同期,2024/3/24,clinical study design for

9、 postgraduate 2003,- 歷史對(duì)照 (historical control) 非隨機(jī)，非同期 偏倚多- 配對(duì)對(duì)照 （matching）,2024/3/24,clinical study design for postgraduate 2003,臨床科研設(shè)計(jì)方案論證強(qiáng)度的影響因素,隨機(jī)化 randomization對(duì)照 control時(shí)向 direction盲法 blinding

10、,,2024/3/24,clinical study design for postgraduate 2003,研究時(shí)向,前瞻性 Prospective study 時(shí)點(diǎn) Point 回顧性Retrospective study,,,,,研究時(shí)點(diǎn),2024/3/24,clinical study design for postgraduate 2003,臨床科研設(shè)計(jì)方案論證強(qiáng)度的影響因素,隨機(jī)化 randomization對(duì)照

11、 control時(shí)向 direction盲法 blinding,,2024/3/24,clinical study design for postgraduate 2003,,盲法 blinding,避免測(cè)量性偏倚（主觀指標(biāo)）避免研究者和被研究者的影響隨機(jī)分配為盲法評(píng)定奠定基礎(chǔ),2024/3/24,clinical study design for postgraduate 2003,2024/3/24,

12、clinical study design for postgraduate 2003,隨機(jī)臨床試驗(yàn)研究（Randomized Clinical Trial）,開始,隨機(jī)分配,,,研究人群,,,,試驗(yàn),,對(duì)照,,,,,,,,,疾病/結(jié)果-,疾病/結(jié)果+,疾病/結(jié)果+,疾病/結(jié)果-,現(xiàn)在,,,,,,測(cè)量對(duì)比結(jié)果,,,,,將來(lái),,R,,,,,,,,,研究指向,2024/3/24,clinical study design for post

13、graduate 2003,隨機(jī)對(duì)照研究,,,結(jié)果指向,2024/3/24,clinical study design for postgraduate 2003,隨機(jī)對(duì)照研究要考慮的問(wèn)題?。。?研究人群的確定效能和有效性（efficacy and effectiveness)終點(diǎn)指標(biāo)的選擇（end-points)觀察期的確定（follow-up） 治療意向分析 （intention to treat analysis）而不是（

14、explanatory analysis) 醫(yī)德問(wèn)題,2024/3/24,clinical study design for postgraduate 2003,2024/3/24,clinical study design for postgraduate 2003,研究人群,General populationExperimental population

15、(10.7%) Unwilling Eligibility criteriaStudy population (7.6% ) Uneligible EnrollmentStudy sample

16、 (3.1%) Unwilling Randomized (2%) Death, Dropped out Completed Study,,,,,,,,,Ann

17、 Intern Med 1993;119:36-41,2024/3/24,clinical study design for postgraduate 2003,符合標(biāo)準(zhǔn)（納入、診斷標(biāo)準(zhǔn)）,,非典型病例有合并癥等,排除不愿意參加者,,潛在的志愿者偏倚？社會(huì)因素（特殊人群）,排除不合作者,重癥？或其他原因,,RCT的研究人群是高度選擇的人群,2024/3/24,clinical study design for postgraduat

18、e 2003,效能和有效性（efficacy and effectiveness),效能（efficacy） 在“理想”環(huán)境中的效果，本身的效能 有效性（effectiveness) 在“實(shí)際”環(huán)境中的效果，受到各種因素的影響，如：病人的依從性、經(jīng)濟(jì)、實(shí)用性、副作用等等。,2024/3/24,clinical study design for postgraduate 2003,終點(diǎn)指標(biāo)的選擇（end-poi

19、nts),客觀指標(biāo)、終點(diǎn)指標(biāo)盲法判斷觀察期的確定（follow-up）隨訪時(shí)間要合適。既要避免浪費(fèi)人力物力又要避免因觀察時(shí)間過(guò)短的假陰性結(jié)果。,2024/3/24,clinical study design for postgraduate 2003,治療意向分析 （intention to treat analysis）,,,Sample,Sample,,,,,,,Drop out,Cross over,Drop out,An

20、alysis according to treatment assigned,population,population,,,Drop out,,,,Cross over,,Drop out,Analysis according to treatment received,2024/3/24,clinical study design for postgraduate 2003,醫(yī)德問(wèn)題,2024/3/24,clinical study

21、 design for postgraduate 2003,RCT的優(yōu)缺點(diǎn),1. 因果關(guān)系中最強(qiáng)的研究2. 盡可能的限制混雜因素（知道和不知的因素）3.內(nèi)部真實(shí)性好4.可以盲法5. 統(tǒng)計(jì)效率高6.治療性研究的標(biāo)準(zhǔn)設(shè)計(jì)方案,外部真實(shí)性差（納入標(biāo)準(zhǔn)、排除標(biāo)準(zhǔn)，嚴(yán)格的計(jì)劃） 花費(fèi)高費(fèi)時(shí) 研究問(wèn)題窄 醫(yī)德 實(shí)施及管理困難志愿者偏倚,2024/3/24,clinical study design for postgrad

22、uate 2003,交叉試驗(yàn) Crossover studies,開始,隨機(jī)分配,,,,,A,,B,,,,,,,,,結(jié)果-,結(jié)果+,結(jié)果+,結(jié)果-,洗脫期,,R,,,,,,,,研究人群,,,,,,,,,,,A,B,,,,,,,,,,,,,,,結(jié)果+,結(jié)果-,結(jié)果+,結(jié)果-,,隨機(jī)選擇方案,,,研究指向,2024/3/24,clinical study design for postgraduate 2003,配對(duì)卡方,,,結(jié)果指向,20

23、24/3/24,clinical study design for postgraduate 2003,Crossover Design,優(yōu)點(diǎn)樣本量減少自我對(duì)照減少偏倚都治療無(wú)醫(yī)德問(wèn)題可隨機(jī)，可盲法缺點(diǎn)疾病過(guò)程有變化變量受限洗脫期時(shí)間確定困難,(不能用于易治愈的疾病),2024/3/24,clinical study design for postgraduate 2003,隊(duì)列研究Cohort study,開始,測(cè)量分

24、類,,,研究人群,,,,,危險(xiǎn)因素+,,危險(xiǎn)因素-,,,,,,,,,疾病/結(jié)果-,疾病/結(jié)果+,疾病/結(jié)果+,疾病/結(jié)果-,現(xiàn)在,,,,,已有病,,無(wú)病,,測(cè)量對(duì)比結(jié)果,,,,,將來(lái),,,,,,,2024/3/24,clinical study design for postgraduate 2003,隊(duì)列研究Cohort Study,時(shí)向： - 前瞻性 Prospective - 回顧性Retrospective

25、 - 回顧-前瞻 Longitudinal隊(duì)列類型 - 固定的隊(duì)列：確定隊(duì)列后不再有新成員加入。如急性心肌梗塞的病人；爆炸后的生存者；HIV感染母親生產(chǎn)的孩子 - 開放隊(duì)列：允許隊(duì)列中的個(gè)體進(jìn)入或退出，動(dòng)態(tài)人群,2024/3/24,clinical study design for postgraduate 2003,歷史隊(duì)列研究,開始,測(cè)量分類,,,研究人群,,,,,危險(xiǎn)因素+,,危險(xiǎn)因素-,,,

26、,,,,,,疾病/結(jié)果-,疾病/結(jié)果+,疾病/結(jié)果+,疾病/結(jié)果-,現(xiàn)在,已有病,,無(wú)病,,測(cè)量對(duì)比結(jié)果,,,,,2024/3/24,clinical study design for postgraduate 2003,Cohort Study DesignDirectionality,,2024/3/24,clinical study design for postgraduate 2003,Time,,2024/3/24,cli

27、nical study design for postgraduate 2003,Relative Risk,Incidenceexp= A/A + B,Incidenceunexp = C/C + D,2024/3/24,clinical study design for postgraduate 2003,COHORT STUDY EXAMPLE,口服避孕藥與深靜脈血栓(DVT)形成的關(guān)系. OUTCOME (DVT

28、) Yes NoExposed ( on oral contraceptive ) 41a 9996b 10037a+b Not exposed (not on o.c.) 7 c 10009d 10016c+d RR=a/a+b / c /c+d =41/10037 / 7/10016=

29、5.8=6* numbers to indicate the possibility of a real association between exposure and outcome. However, the possibility of biases very often arises.,2024/3/24,clinical study design for postgraduate 2003,Cohort Study De

30、signPotential Biases,檢測(cè)偏倚 兩組檢測(cè)不一致，不能采用同樣的方式，客觀地在暴露和非暴露中測(cè)量暴露和結(jié)果信息偏倚兩組獲得信息不一致無(wú)應(yīng)答或失訪，隨訪時(shí)間不足不能控制對(duì)照組的混雜因素,2024/3/24,clinical study design for postgraduate 2003,Cohort Study,Strengths暴露因素少的可得到結(jié)果,因-果時(shí)相關(guān)系清楚單個(gè)暴露多種結(jié)果可以發(fā)現(xiàn)短

31、暫時(shí)間的結(jié)果直接測(cè)量發(fā)病率因果關(guān)系強(qiáng)無(wú)醫(yī)德問(wèn)題確定自然病程可配對(duì)、標(biāo)準(zhǔn)統(tǒng)一，管理容易,Weaknesses不宜用于少見(jiàn)病研究花費(fèi)高要求隨訪率高暴露存在混雜因素對(duì)照難確定，不能隨機(jī)，盲法， 歷史對(duì)照暴露和結(jié)果不可靠,2024/3/24,clinical study design for postgraduate 2003,隨機(jī)對(duì)照與隊(duì)列研究區(qū)別,前瞻隨機(jī)分配幾乎不能用于臨床病因?qū)W研究治療性研究的標(biāo)準(zhǔn)設(shè)計(jì)方案

32、給予暴露（治療）分析結(jié)果,前瞻或回顧等不是隨機(jī)分配而是自然形成的隊(duì)列病因?qū)W研究中論證強(qiáng)度較高的設(shè)計(jì)觀察暴露后的結(jié)果,2024/3/24,clinical study design for postgraduate 2003,前-后對(duì)照研究before-after studies,開始,,,A,,,,,結(jié)果+,結(jié)果-,洗脫期,,,,研究人群,,,,,,B,,,,,,,,結(jié)果+,結(jié)果-,,2024/3/24,clinical stud

33、y design for postgraduate 2003,前-后對(duì)照研究（before-after study),兩種措施在兩個(gè)階段進(jìn)行可以是同一個(gè)體或不同個(gè)體前瞻或回顧-前瞻,2024/3/24,clinical study design for postgraduate 2003,不同個(gè)體前后研究歷史性對(duì)照,A,,,,結(jié)果+,結(jié)果-,,,,,結(jié)果+,結(jié)果-,過(guò)去,,,,,,,,,現(xiàn)在,,B,,,,,,將來(lái),2024/3/24

34、,clinical study design for postgraduate 2003,前-后對(duì)照研究?jī)?yōu)缺點(diǎn),優(yōu)點(diǎn)每一病人同時(shí)治療樣本量減少自我對(duì)照減少偏倚都治療無(wú)醫(yī)德問(wèn)題缺點(diǎn)兩階段疾病的可比性有變化洗脫期時(shí)間病情可能加重,2024/3/24,clinical study design for postgraduate 2003,交叉試驗(yàn)與前后研究區(qū)別,前瞻隨機(jī)分配為兩組同期對(duì)比,前瞻或其他不是隨機(jī)分配但可以隨機(jī)選

35、擇方案不同期前后對(duì)比（一組）,2024/3/24,clinical study design for postgraduate 2003,病例-對(duì)照研究（Case control study）,開始,測(cè)量對(duì)比,,,Case 有病/結(jié)果,危險(xiǎn)因素+,過(guò)去,,,,,Control 無(wú)病/無(wú)結(jié)果,危險(xiǎn)因素-,危險(xiǎn)因素-,危險(xiǎn)因素+,,,,,,,,,現(xiàn)在,,,,,,回顧性,2024/3/24,clinical study design f

36、or postgraduate 2003,病例對(duì)照研究的基本要求,病例與對(duì)照的確定診斷手段一致、病例的確定應(yīng)是確診病人病例與對(duì)照來(lái)源人群應(yīng)同源病例與對(duì)照的納入及排除標(biāo)準(zhǔn)一致,2024/3/24,clinical study design for postgraduate 2003,Case-Control StudyInterpretation,,OR =,acbd,,,,任何影響病例組和對(duì)照組的暴露比都將造成偏倚.,202

37、4/3/24,clinical study design for postgraduate 2003,CASE CONTROL EXAMPLE -SMOKING & LUNG CANCER,DISEASECasesControlsEXPOSURE Yes a bEXPOSURE No c dOdds Ratio = ad/bc(1 = no associat

38、ion, > 1 = possible association, < 1 = protective effect)DISEASECasesControls (lung cancer)EXPOSUREYes 56 230(smoking)No 7 246The odds ratio would theref

39、ore be 56 x 246 = 13776 = 8.6. 7 x 230 1610,2024/3/24,clinical study design for postgraduate 2003,優(yōu)點(diǎn),研究稀少病種和潛伏期長(zhǎng)的疾病的暴露與結(jié)果間關(guān)系最合適的研究方案，確定潛在因果關(guān)系 不需要隨訪，

40、結(jié)果已很好的定義適用于潛伏期長(zhǎng)的疾病病例數(shù)需要的少，（比橫斷面研究）省錢，省時(shí)提出假說(shuō)無(wú)醫(yī)德問(wèn)題同時(shí)研究多個(gè)危險(xiǎn)因素，有幾個(gè)假說(shuō),2024/3/24,clinical study design for postgraduate 2003,缺點(diǎn),病例的選擇容易出現(xiàn)偏倚對(duì)照與病例的人群來(lái)源不一致。暴露可能與生存率有關(guān)而不是疾病的發(fā)生不能計(jì)算患病率、發(fā)病率、相對(duì)危險(xiǎn)度，因果關(guān)系不清短暫的關(guān)系很難確定存在生存率偏倚、樣本偏

41、倚，以及各種測(cè)量偏倚，共75種,2024/3/24,clinical study design for postgraduate 2003,病例對(duì)照研究的偏倚舉例,不易明確的確定對(duì)照組有時(shí)不能采用同樣的方式，客觀地在病例組和對(duì)照組測(cè)量暴露和結(jié)果回憶偏倚等,2024/3/24,clinical study design for postgraduate 2003,減少樣本偏倚,配對(duì) Matching人群樣本 – 人群登記選擇對(duì)照

42、 – 隨機(jī)撥號(hào) – 鄰居對(duì)照 多個(gè)對(duì)照病例和對(duì)照的發(fā)病率最好與一般人群中相似,2024/3/24,clinical study design for postgraduate 2003,減少測(cè)量偏倚,盲詢問(wèn)者 – Case/Control Status – Risk Factors盲病人和對(duì)照 – Case/Control Status:,2024/3/24,clinical study design for postg

43、raduate 2003,病例對(duì)照研究與回顧性隊(duì)列研究區(qū)別,不能確定因果關(guān)系首先確定結(jié)果（病例）,因果關(guān)系時(shí)向清楚首先確定暴露（病因）,,,,,？,？,,,,,暴露,疾病,,,,,？,？,,,開始,開始,疾病,暴露,,病例對(duì)照研究,回顧性隊(duì)列研究,2024/3/24,clinical study design for postgraduate 2003,橫斷面研究cross-sectional survey,開始,測(cè)量,,,研究人群

44、,,,,,無(wú)病,,有病,,,,,,,,,危險(xiǎn)因素-,危險(xiǎn)因素+,危險(xiǎn)因素+,危險(xiǎn)因素-,時(shí)點(diǎn),,,,,2024/3/24,clinical study design for postgraduate 2003,橫斷面,隨機(jī)選擇、適宜應(yīng)答率真實(shí)性和重復(fù)性,2024/3/24,clinical study design for postgraduate 2003,橫斷面研究Cross-Sectional study,測(cè)量：在同一時(shí)間內(nèi)

45、，檢查結(jié)果與危險(xiǎn)因素等暴露因素的關(guān)系得到患病率（Prevalence）容易、快捷、便宜、無(wú)醫(yī)德問(wèn)題不能說(shuō)明因果關(guān)系，只能表明聯(lián)系回憶偏倚，混雜因素分布不均勻、樣本不等形成假說(shuō),2024/3/24,clinical study design for postgraduate 2003,優(yōu)點(diǎn),無(wú)隨訪時(shí)間（時(shí)點(diǎn)）是唯一可得到患病率的研究類型假說(shuō)的形成前瞻性隊(duì)列研究的開始 prospective cohort study,202

46、4/3/24,clinical study design for postgraduate 2003,缺點(diǎn),只能建立因果聯(lián)系時(shí)相關(guān)系不清楚不使用于發(fā)病率低的疾病,2024/3/24,clinical study design for postgraduate 2003,敘述性研究（descriptive study）,結(jié)果,,,,病因+,病因-,,,研究方向,2024/3/24,clinical study design for p

47、ostgraduate 2003,敘述性研究（descriptive study）,措施/病因,,,,結(jié)果+,結(jié)果-,,研究方向,,2024/3/24,clinical study design for postgraduate 2003,敘述性研究（descriptive study）,個(gè)案報(bào)道或病歷報(bào)道（case report or case series)詳細(xì)描述單個(gè)病例或多個(gè)病例的特點(diǎn)目的：用于少見(jiàn)病例、新發(fā)現(xiàn)、教學(xué)等新治

48、療方案的嘗試、預(yù)試驗(yàn)等不是研究無(wú)對(duì)照一般不能說(shuō)明因果關(guān)系,2024/3/24,clinical study design for postgraduate 2003,病歷報(bào)道有時(shí)可以提示病因聯(lián)系或提供強(qiáng)的證據(jù),結(jié)果很少并且具有特點(diǎn)暴露因素去除，情況改善，再暴露，再出現(xiàn)脫發(fā)與乙肝疫苗基因治療持久的媒介媒介的結(jié)果和靶器官的毒性作用.基因治療明顯改善預(yù)后如AIDS病治療,2024/3/24,clinical study

49、design for postgraduate 2003,2024/3/24,clinical study design for postgraduate 2003,,2024/3/24,clinical study design for postgraduate 2003,2024/3/24,clinical study design for postgraduate 2003,,強(qiáng),弱,2024/3/24,clinical stud

50、y design for postgraduate 2003,研究目的與研究設(shè)計(jì)的選擇,2024/3/24,clinical study design for postgraduate 2003,研究課題的設(shè)計(jì),臨床觀察視力差和戴眼鏡似乎與缺乏胡蘿卜素有關(guān)科學(xué)理論視紫質(zhì)的形成需要胡蘿卜素，視黃醛與夜間視力有關(guān)假說(shuō)視力的好壞與進(jìn)食胡蘿卜的量有關(guān),2024/3/24,clinical study design for postgr

51、aduate 2003,病例-對(duì)照研究,開始,測(cè)量對(duì)比,,,Case 視力差組,食胡蘿卜多,過(guò)去,,,,,Control 視力好組,食胡蘿卜少,食胡蘿卜少,食胡蘿卜多,,,,,,,,,現(xiàn)在,,,,,2024/3/24,clinical study design for postgraduate 2003,橫斷面研究,開始,測(cè)量,,,選定人群,,,,,視力好,,視力差,,,,,,,,,胡蘿卜多膳食,時(shí)點(diǎn),,,,,胡蘿卜少膳食,胡蘿卜多

52、膳食,胡蘿卜少膳食,2024/3/24,clinical study design for postgraduate 2003,隊(duì)列研究,開始,測(cè)量分類,,,研究人群,,,,,食多胡蘿卜,食少胡蘿卜,,,,,,,,,視力差,視力好,視力好,視力差,現(xiàn)在,,,,,視力差,,,測(cè)量對(duì)比結(jié)果,,,,,將來(lái),2024/3/24,clinical study design for postgraduate 2003,隨機(jī)對(duì)照,開始,隨機(jī)分配,,,

53、研究人群,,,,,食多胡蘿卜,食少胡蘿卜,,,,,,,,,視力差,視力好,視力好,視力差,現(xiàn)在,,,,,視力差,,,測(cè)量對(duì)比結(jié)果,,,,,將來(lái),2024/3/24,clinical study design for postgraduate 2003,幾點(diǎn)說(shuō)明,所有的設(shè)計(jì)都有優(yōu)缺點(diǎn)選擇最適合研究問(wèn)題的設(shè)計(jì)盡量避免劣勢(shì)最大限度的保證真實(shí)性和可靠性嚴(yán)格地計(jì)劃每一細(xì)節(jié)過(guò)程設(shè)計(jì)Overlap,2024/3/24,clinical st

54、udy design for postgraduate 2003,References,Robert H.Fletcher ect Clinical Epidemiology. The essentialsRCT: 1. Pocock, S.J. Statistical Aspects of Clinical Trial Design. The Statistican, 31:1-18,1982. 2. Yus