实证医学-台湾大学课件

1、循证医学Evidence-Based Medicine 目录 循证医学介绍 (slide 1-42)及教學應用 提出临床问题 (PICO)提升文献搜索技能 (cite references) 严格文献评读 实例练习病紀引用考文獻 (PubMed or UpToDate)Medical Education in the New CenturyBioinformaticsEthicsHumanitiesPatient-centered careProblem-based learningEvidence-based medicine 定义：Use of current best evidence

2、in making decisions about the care of individual patients. EBM (EBN, EBP) is the integration of best research evidence with clinical expertise and patients unique biology, values andcircumstances (patients expectation).(Evidence-based Practice) (Sackett & Straus)Evidence-Based MedicineDrPtEvDecision

3、 Making in Health CareSearching bibliographic databasesClinical practice guideline (CPG)Evidence-based journal abstracts“Do no harm”What you learned during your professional trainingBrowse journalsTextbooksAsk colleaguesChart record cite reference 病紀引用文獻床問題型診斷 (Diagnosis)Sensitivity, specificity敏感、特

4、Predictive value (PPV, NPV) 陽性預測值、陰性預測值ROC curve, Likelihood ratio (LR+, LR-) 概似比治 (Therapy)Clinical trial (Randomized Controlled Trial, RCT, RR)預後 (Prognosis)Prediction model (Survival analysis, HR 風險比)危險因子探討 (Risk factor)Cohort study (Relative Risk, RR 相對風險比)Case-control study (Odds Ratio, OR勝算比)S

5、tep 1.Converting the need for information (about prevention, diagnosis, prognosis, therapy, causation, etc.) into an answerable question. (Ask) - PICOStep 2.Searching the best evidence with which to answer that question. (Acquire)Step 3.Critically appraising the evidence for its validity (closeness

6、to the truth), impact (size of the effect), and applicability (usefulness in our clinical practice). (AApppprraaiissaall) - VIPStep 4.Integrating the evidence with our clinical expertise and patients unique biology, values and circumstances. (Apply) - 3 EStep 5.Evaluating our effectiveness and effic

7、iency in executing steps 1-4 and seeking ways to improve them for next time. (Audit)實證醫學的五大進步驟Five Steps to Practice EBM (5 Ass)Grade of RecommendationLevel of EvidenceTherapyAFrom: Oxford Center for EBM1aSystematic review (with homogeneity) of RCTs1bSingle RCT (randomized controlled trial)1cAll-or-

8、noneB研究設計與證據的強Evidence-Based Medicine: How to Practice and Teach EBM. 2nd ed. David L. Sackett, Sharon E. Straus, W. Scott Richardson, William Rosenberg, R. Brian Haynes.Churchill Levingstone. 2000, p173-177Guideline 證據等級（Levels of Evidence）: 床指引(Guideline) 建議的形成是根據文獻的證據等級 而，建議強與證據等級有關， 但建議強並非表示建議的重

9、要2aSystematic review of cohort studies2bCohort study or poor RCT2cOutcomes research3aSystematic review of case-control studies3bCase-control study性。C4Case seriesD5Expert opinion, physiology,bench researchUS Preventive Services Task Force Rating System of Quality of Scientific EvidenceI: Evidence obt

10、ained from at least 1 properly designed,randomized controlled trialII-1: Evidence obtained from well-designedcontrolled trials without randomizationII-2: Evidence obtained from well-designed cohort or case-control analytic studies, preferentially from more than 1 center or groupII-3: Evidence obtain

11、ed from multiple time series with or without the intervention, or dramatic results in uncontrolled experiments (such as the results of the introduction of penicillin treatment in the 1940s)III: Opinions of respected authorities, based on clinical experience, descriptive studies, or reports of expert

12、 committees實證醫學沿革1972英國床病學者Archie Cochrane提出實證醫學的概。所有醫為應有嚴謹研究且證實為有效的根 據，才能將醫資源做最有效的運用，並強調 Randomizedcontrolledtrials的重要性。*無效果 之檢驗或治要做1992英國國家衛生部成實證醫學中 心，並以Archie Cochrane之名命名，進 而促使1993Cochrane Collaboration的設。(Iain Chalmers, David Sackett)Lancet 曾把Cochrane Collaboration比作床醫學的人基因組計劃Study DesignStudy

13、DesignStudy Design研究開始問題 (用途)研究種時間性過去現在未Cross-sectional(prevalence)橫斷性 觀察7收集資7Case & non-Case盛、診斷Cohort(longitudinal)縱向性(前瞻)定義世代並評 估危險因子觀察結果Y*N發生、病程預 後、病因Clinical Trial(experimental)縱向性(前瞻)作治治 組與對照組觀察結果Y*N藥物效評估Case control(retrospective)縱向性(回溯)評估危險因子Exposure: Y*N界定病組 與非病組病因尤其罕 病Repeated cross-sectio

14、nal橫斷性 觀察收集資7重複收集77隨時間改變Treatment of Class III-IV Lupus NephritisMeta-Analysis (pooling studies)Treatment of diffuse proliferative lupus nephritis: A meta-analysis ofrandomized controlled trialsAm J Kidney Dis. 2004 Feb;43(2):197-208.Cochrane Renal Group Cochrane Database of SystematicReviews.1, 200

15、4.2004 AJKDAm J Kidney Dis. 2004;43:197-208. Cochrane Renal Group Cochrane Database of Systematic Reviews 2004.0 Summary effect: Diamond-shaped symbols represent the summary estimator of overall effect pooling the weighted effect of individual RCTs.4 Null Hypothesis ( 虛無假設 RR, OR, or HR =1, mean dif

16、ference, effect size = 0): a vertical line at 1.0 representing equivalence in risk for an outcome with experimental and control treatment. Values of RR less than 1 indicate a reduction in risk for the outcome with the experimental treatment. Conversely, values of RR more than 1 indicate an increase

17、in risk.O Point estimate and 95% CI (點估計與區間估計): The RR for each outcome and its 95% CI are indicated by a solid square and a line. The 95% CIs are a measure of variability in the precision of the RR estimate and its statistical significance. The size of the solid square represents the contribution (

18、weight) of the trial to the analysis. Heterogeneity (non-combinability if p 80%) (失者少於20%) Yes治方法對病患、醫護人員、研究者是否blinded？ Yes 分析是否用Intention-to-Treat analysis (ITT)？ No vs. per protocol 除研究治項目以外，其他的治在各組間是否相同？Yes 組在治開始時的 baseline是否相似？(Table 1) Yes在閱每一篇文章時，要注意是否符合這些基本原則，如果沒有，是為麼沒 有，對於結果有沒有影響？(in Materia

19、l & Method, Result section)當有一個可信的結果，還要考慮文章的結果對病人實際上的意義為何？重 重要 ( Impact 重要性 importance:ssiize ooff tthhee eeffffeecctt, NNT, NNH )？文章常以 RRR (Relative risk reduction)表示效，但以 NNT (Number Needed to Treat), 及 NNH (Number needed to harm) 表達為直接。Yes No ? Unclear評估文章的可信 (Validity) 和實用性(*注意研究選入病人的條件)3. Critic

20、al Appraisal of the Evidence(評批判：VIP, RAMbo, CAT, or CASP checklist.)Critically appraising the evidence for its (VIP)Validity (closeness to the truth可信) 注意文章 inclusion criteriaWas the assignment of patients to treatment randomized?Was follow-up of patients sufficiently long and complete? ( 80%)Were

21、all patients analyzed in the ggrroouuppss to which they were randomized? (ITT, Intension To Treat analysis) vs. per protocolWere patients and clinicians kept blind to treatment?Were groups treated equally, apart from the experimental therapy?Were the groups similar at the start of the trial?Validity

22、 (truth): consider selection bias, information bias, confounding factorReliability of measurement (Repeatibility): intraobserver (similarity over time), interobserver, internal consistencySD, variance, point estimate - 95% CI (narrow or wide), p value ( 0.05)Impact (size of the effect, importance):

23、床的重要性NNT (number needed to treat) = 1/ARR (absolute risk reduction)NNH (number needed to harm) = 1/ARI (absolute risk increase)Practice Applicability (usefulness in our clinical practice)Integrating the evidence with our clinical expertise and patients values and preferences (expectation). (3 E)Aski

24、ng an Answerable Question Impact: NNT, NNH 計算Patient/ProblemInsulin-dependent diabeticsInterventionIntensive insulin regimenComparisonRegular insulin regimenOutcomesRetinopathySymptomatic hypoglycemiaTreatment Effects NNT 益一治Occurrence of diabetic retinopathy at 5 years among insulin- dependent diab

25、etic in the DCCT trialUsual insulin regimen (CER: control event rate): 38%Intensive insulin regimen (EER: experimental event rate): 13%Risk reduction (calculation): NNTAbsolute risk reduction (ARR)= CER-EER = 38% - 13% = 25%Relative risk reduction (RRR)= CER-EER/ CER = 25% / 38% = 66%Number needed t

26、o treat (NNT)= 1/ARR= 1 / 25% = 4 patientsNNT: The number of patients that need to be treated to prevent one bad outcome or get one good outcome. 增加一位病患得到某種處置好處所需的治病人 = 1/ARR,即與對照組 法相比而言，使一位病人達到實驗組治之有結果（或預防產生 結果）所需治的病人目。(NNT 越少越好)Harm: NNHThe proportion of patients with at least one episode of sympt

27、omatic hypoglycemiaUsual insulin regimen (CER: control event rate): 23%Intensive insulin regimen (EER: experimental event rate): 57%Risk increase (calculation): NNHAbsolute risk increase (ARI) =EER - CER= 57% - 23% = 34%Relative risk increase (RRI)=EER-CER/ CER= (57%-23%)/23%= 148%Number needed to h

28、arm (NNH) = 1/ARI = 1/0.34 = 3 patients (整)NNH: The number of patients that need to be treated to cause one bad outcome (being harmed). 增加一位受試者患某種醫源性傷害的治病人：即對多少病人進 實驗組治與對照組法做比較會有多一個病人產生副作用。 (NNH 目,越大越好)From RRR to ARR and NNTMeasures of the effects of treatment: RRR, ARRCEREERRRRARRNNTTreatment A50%

29、39%22%= (50-39) / 5011%TreatmentBTreatmentC5.0%0.5%3.9%0.39%22%22%1.1%0.11%9 人90 人900人*RRR (RR, OR) can not discriminate huge treatment effects from small ones!(also consider Patients baseline/control event rate, and time factor )指標 Relative Risk (risk ratio) = EER/CER vs. Absolute Risk reduction (r

30、isk difference) =EERCERvs. NNT=1/ARRHow large was the treatment effect治效果有多大？，某研究治追蹤二，對照組死亡15，治組死亡10 結果呈現的方式有下幾種：呈現的方式代表的意義Relative Risk (相對風險) (RR)治組發生風險相對於對照組的倍。RR=1 組無差別，RR 1 治會增加風險RR = 0.10 / 0.15 = 0.67RR 1 表示治可低死亡的風險Absolute Risk Reduction (ARR)治組與對照組發生風險的絕對差(絕對危險性低)ARR = 0.15 0.10 = 0.05 or 5

31、%治的處是低 5%的死亡Relative Risk Reduction (RRR)相對於對照組，治組低風險的比(相對風險性低)(最常的呈現方式)RRR = 0.05 / 0.15 = 0.33 or 33%or RRR = 1 0.67 = 0.33 or 33%相對於對照組，治可以低死亡的的機 33%Number Needed to Treat (NNT)要預防一位病人結果發生，所必需治的病人(一需治)NNT = 1 / ARR = 1 / 0.05 = 20治20位病人2，才能預防1人死亡Step 1.Converting the need for information (about p

32、revention,diagnosis, prognosis, therapy, causation, etc.) into an answerablequestion.(PICO)Step 2.Searching the best evidence with which to answer that question. (cite reference)Step 3.Critically appraising the evidence for its validity (closeness to the truth), impact (size of the effect), and appl

33、icability (usefulness in our clinical practice).(VIP)Step 4.Integrating the evidence with our clinical expertise and patients unique biology, values and circumstances (expectation). (3E)Step 5.Evaluating our effectiveness and efficiency in executing steps 1-4 and seeking ways to improve them both fo

34、r next time.重點:用簡單且病人可以聽懂的語言，對病患告知診決策的與弊4. 實證醫學的五大進步驟Five Steps to Practice EBMAppraising the evidence:Applicability 床可應用性 Are the result Valid? Is it clinical important ? (Impact: NNT.) Applicability to our patient ? (床可應用性) Is our patient so different from those in the study that its results canno

35、t apply ?Data from Taiwan, China, or Asia (種族差)? Cost-effectiveness analysis (效益分析) Do I miss any data? 搜尋考文獻 詢問專家 有沒有其他處置方式? Whatre our Pts potential benefits from CCRT ? Whatre our Pts potential harms from CCRT ? Whatre our Pts values & preferences for theoutcomes & side effect ?(3 E)Discussion, a

36、pply and AuditPlain language summary (absolute % difference, NNT, NNH)PtEv DrEvidence-Based Practice Clinical ScenarioPatients Clinical Problemraise clinical question學習目標Perform Five Steps in EBM (5 As)Ask (PICO): ask a clinical question (P-I-C-O)Acquire: search database (cite reference)Appraisal (V

37、IP): Valid? Important? Practical?Apply: to patients problem (3E evidence, experience, expectation)Plain language summaryAudit: effectiveness (Explain treatment options to you patient)Apply (Integrate with ourclinical expertise and patient values)1.目前醫學上的證據並支持孕婦服用生菌 預防氣喘的發生（文獻等級1a）。而您本 身第一胎有氣喘，屬於高危險族

38、群，雖然 生菌對於過敏疾病高危險嬰兒可能減少濕疹 的機，但對氣喘預防並沒有效果（文獻等 級1a）。New Meta-analyss RR 1.47 (1.04-2.09)2.2008有一篇研究，甚至發現孕婦生產前4-6周及嬰兒服用生菌(LGG ) 6個月反而 會增加二歲時氣喘的危險百分之16.9%，服 用生菌者中每位 (NNH = 6)會多增加一 位得到氣喘的危險（文獻等級1b）。Plain language summaryTwo Fundamental Principles ofEBMEBM posits a hierarchy of evidence to guide clinical d

39、ecision making. (Level of evidence)Evidence alone is never sufficient to make a clinical decision. 3E:Consider the patients value (Expectation of the Patient) 告知 同意：用病人可以聽得懂的語言 (explain)Integrate clinical expertise (Expert opinion)Trade the benefits and risks (NNT, NNH)Costs ($)Inconvenience 研究效果需要因

40、應個別病人做調整如治 Patient NNT = 1/ (RRR PEER)PEER = patient expected event rate (your case)Critical appraisal ChecklistCATOxford CEBM Critical Appraisal Sheets (include RAMBo sheet)/index.aspx?o=1913CCAATT maker: /index.aspx?o=1216Report: CAT (Critical Appraisal Topic) formCASPCritical Appraisal Skills Pro

41、gramme (CASP) appraisal toolhttp:/www.phru.nhs.uk/pages/PHD/resources.htmGATE frame (critical appraisal with picture)Q & A 報告實習診斷檢驗Diagnostic testSensitivity. Specificity, PPV, NPV, LR, ROC curve疾病(Disease)有(Present)無(Absent)檢查(Test)陽性(Positive)真陽性TPa偽陽性bFP陰性(Negative)偽陰性cFNd真陰性TNDiagnosisDiagnostic

42、 test(ferritin)Disease (IDA)PresentAbsentPositive(陽性)731TP abFP270Negative (陰性)78FN cdTN 1500SnSpSensitivity (Sn) = a/a+c = 731/809 = 90%Specificity (Sp) = d/b+d = 1500/1770 = 85%(敏感會受疾病嚴重程影響)PPV a/a+bNPV d/c+dPositive predictive value (PPV) = a/a+b = 731/1001 = 73% (=post-test probability)Negative

43、predictive value (NPV) = d/c+d = 1500/1578 = 95%但 診斷試驗的預測值 (predictive value)受疾病的盛 (prevalence) 影響。 Positive predictive value (PPV) = Sen . P / Sen . P + (1-Sp) . (1-P)(貝貝氏氏定) P= 0.5,PPV= 0.80.5 / 0.80.5+0.20.5= 0.8 =80.0%P= 0.05,PPV= 0.80.05 / 0.80.05+0.20.05=17.4%P= 0.005, PPV= 0.80.005 / 0.80.005

44、+0.20.005 =0.2%同一診斷工具, 在同盛情況下, 其 Predictive value 結果同。 LR 概似比Specificity 高，但運用在盛低的族群時，大部分陽性結果是假陽性。 Sensitivity 高，但運用在盛高的族群時，大部分陰性結果是假陰性。 ： Ovarian cancer CA-125: PPV 2% in screen vs. 97% in pelvic mass casesLRPrevalence (different clinical situations) affect predictive valueDiagnostic testDisease (H

45、IV)PresentAbsentPositive (陽性)9TP100FPNegative (陰性)1FN9890TN： Ovarian cancer CA-125: PPV 2% in screen vs. 97% in pelvic mass casesIncreasing the Prevalence of Disease Before Testing : When the prevalence of disease in the population tested is relatively high more than several percent the tteesstt ppe

46、errffoorrmm wweellll.Community-wide HIV screeningTest: 90% sensitivity, 99% specificPopulation 10,000Prevalence 0.1% = 10/10,000PPV = 9/(9+100) = 0.08 = 8%NPV= 9890/(1+9890) = 0.9999 = 99.99%PPV同一診斷工具, 在同盛情況下, 其 Predictive value 結果同Note: spectrum of patients, age, gender, risk factors, clinical find

47、ings (prevalence)Increasing the Prevalence of Disease Before Testing Prevalence affect the PPVTest ResultDisease (DM)PresentAbsentPositive (陽性)900TP90FPNegative (陰性)100FN8910TNCommunity-wide HIV screeningTest: 90% sensitivity, 99% specificPopulation: 10,000Prevalence: 0.1% = 10/10,000PPV = 9/(9+100)

48、 = 0.08 = 8%NPV = 9890/(1+9890) = 0.9999 = 100%Population = 10,000Prevalence = 10% = 1000/10,000 Sensitivity = 900/1000 = 90% Specificity = 8910/9000 = 99%PPV = 900/990 = 0.91 = 91% NPV = 8910/9010 = 0.99 = 99%Diagnostic testDisease (HIV)PresentAbsentPositive (陽性)9TP100FPNegative (陰性)1FN9890TNDiagno

49、sis LR: likelihood ratio (multi-levels odds)Diagnostic test(ferritin)Disease (IDA)PresentAbsentPositive(陽性)731TPabFP(I) 270Negative (陰性)78FN(II) cdTN1500Sensitivity = TP/(TP+FN) = a/a+c = 731/809 = 90%SnSpSNout, SpPinSpecificity = TN/(FP+TN) = d/b+d = 1500/1770 = 85%Positive predictive value (PPV) =

50、 a/a+b = 731/1001 = 73% (=post-test probability)Negative predictive value (NPV) = TN/(FN+TN) = d/c+d = 1500/1578 = 95%LR+ for a positive result = Sens/(1- Spec) = a/(a+c) / b/(b+d) = 90%/15% = 6陽性概似比 LR+ = 敏感/(1-特) = TP/FP = 有病者與無病健康者,檢驗呈陽性的機勝算比LR- for a negative result = (1-sens)/spec = c/(a+c) / d

51、 /(b+d) = 10%/85% = 0.12Pre-test probability (prevalence)= a+c/a+b+c+d= 31%Pre-test odds = prevalence / (1-prevalence) = 31%/69% = 0.45* Post-test odds = Pre-test odds LR = 0.45 6 =2.7PPosttest probability = Posttest odds / (odds + 1) = 2.7/(2.7+1) = 73% (= PPV 73%)+ PV- PVDiagnosisUse of Sensitivit

52、y Test(高敏感檢查的運用)Treatable diseaseScreening: maximize sensitivity while optimizing specificity未被檢查出會有嚴重後果者 treatable or transmissiblee.g. screen donated blood for HIV, Pap smear, mammogramsRule out disease (SNout) e.g. ANAUse of Specificity Test(高特定檢查的運用) 當假陽性結果會傷害患者身體、情緒、財物時e.g. cancer chemotherapyR

53、ule in disease (SPin)Diagnosis: maximize specificity while optimizing sensitivity. e.g. anti-ds DNAROC (receiver operating characteristic) curve 以真陽性(TP, 敏感)為縱軸，假陽性(FP, 1-特)為橫軸，即將有病及無病者檢 驗結果呈陽性的機做一比較。ROC curve下方的面積越大，診斷工具的準確越好Optimum cutoff point correlated with the best Youden index = (sensitivity

54、+ specificity - 1)c.f. LR+ for a positive result = sensitivity / (1- specificity): The probability of that test result in people with the disease divided by the probability of the result in people without diseaseAccuracy: proportion of correct results = (a+d) / (a+b+c+d) = (prevalence x sensitivity)

55、 + (1-prevalence) x (specificity) = TP+TN / (TP+TN+FP+FN)Diagnosis StrategiesSerial vs. parallel testSerial Test(高特定)The result of test 1 are considered before test 2, and so onInorder to be considered positive, all test in the series must be positiveHighly specific but insensitiveUseful when false positive are undesirableSuch as treatment is highly invasive or toxic e.g. cancer chemotherapyParallel Test(高敏感)Any positive is considered a positiveSensitive but not specificUseful when rapid di