臨床證據的明日之星 Network Meta-analysis NMA 臨床證據的明日之星 Network Meta-analysis 陳弘洲醫師 2015-01-17

Meta-analysis研究的優缺點 優點 缺點 證據等級高 不用臨床試驗 研究時間短 研究經費低 入門稍難 選題不易 競爭者眾 只探討單一種治療

Introduction of Network Meta-analysis

What is Network Meta-analysis? 將針對某疾病所有治療方法的RCT找出 分析每種治療方法的效果並加以排名比較 例如探討某一疾病的治療: 傳統Meta-analysis Network Meta-analysis 方法 找出某種治療的所有研究 找出所有治療方法的所有研究 目的 分析某種治療的療效 分析所有治療並加以比較

一分鐘看懂 A和B誰兩位職業網球選手誰較厲害? Clinical trial= 找兩位選手出來比一場 Meta-analysis= 找出兩位選手所有對戰紀錄加以分析 Network Meta-analysis= 世界網球排名

舉例 RCT:

誰是第一種子﹖ The direct &indirect comparison

Meta-analysis的盲點 假設有ABC三種治療某疾病藥物, 想知道A與C誰較佳 找出共有10篇A vs C的RCTs Meta-analysis結果顯示療效A > C, 故A較佳 如果… 另有20篇RCTs顯示A<<B, 另30篇RCTs顯示C>>B 故C>>B>>A A與C究竟誰療效較佳???

只有direct compare夠嗎??

*Adapted by CTL from Bucher 1997, Song 2003, Glenny 2005 A versus C *Adapted by CTL from Bucher 1997, Song 2003, Glenny 2005

A versus C 單位統一 Conscious data用Standard Mean Differences (SMD, 指統一用standard deviation當單位) Dichotomous data多用Relative Risk (RR) AC差 = (權重)d x(直接 AC差) + (權重)I x(間接 AC差)  d AC = W d x d AC d+ W I x d AC I

權重怎麼算? Weight = 1/ (SE)2 SE= (upper-lower)95%信賴區間 /3.92 N 越大weight越大 SD越大weight 越小

Cipriani et al. Lancet 2009 Feb 28;373(9665):746-58 當藥物變多時… Cipriani et al. Lancet 2009 Feb 28;373(9665):746-58

Cipriani et al. Lancet 2009 Feb 28;373(9665):746-58 將所有對戰組合全部算出 Cipriani et al. Lancet 2009 Feb 28;373(9665):746-58

Cipriani et al. Lancet 2009 Feb 28;373(9665):746-58 列出所有種子的總成績 Cipriani et al. Lancet 2009 Feb 28;373(9665):746-58

如何開始 Ask your clinical question!

適合做Network Meta-analysis的題目 探討單一疾病於各種治療的效果 已有許多RCTs 有些Meta-analysis 對於何種治療較優有爭議者更好 不要太廣太大的題目

題目太大怎麼辦?

縮小題目 X 探討所有治療Schizophrenia的方法 O 探討常用第二代抗精神藥物用於治療 Schizophrenia的比較 X 探討所有治療plantar fasciitis的方法 O 探討各種injection therapy用於治療plantar fasciitis的比較

海裡撈針的技巧 Identify your trials

訂出合宜的文章篩選條件 明確闡述要找什麼樣的文章 需切合主題 不要限定搜尋SCI/SSCI journal 不要限定搜尋英文文章 病人族群為何? 介入方法文何? 實驗設計為何? 需切合主題 不要限定搜尋SCI/SSCI journal 不要限定搜尋英文文章

制訂合宜的搜尋策略 Keyword要盡可能找出所有同意字  如找plantar fasciitis: "plantar fasci*" or "plantar heel*" or "plantar foot" or "heel pain" or "painful heel*" or "hindfoot pain" or "painful hindfoot “ 如找dyslipidemia: dyslipid* or hyperlipid* or lipid or cholesterol or lipoprotein or triglyceride or TC or TG or HDL or LDL

制訂合宜的搜尋策略 盡可能多搜尋資料庫 Pubmed, Cochrane Library, Scopus… 利用資料庫功能增加找到文章的機會  Mesh, reltated article.. 利用文章reference增加找到其它文章的機會

Krogh et al. The American Journal of Sports Medicine 2012

建構你的資料 Collecting the data

Characteristics of the selected trials. Study (Year) Patients n, (Feet n) Comparison (No. of Patients Assigned to Each Treatment Arm) Followup peroid Outcomes Measure Incl/Excl Criteria Avg Age, y Avg Disease Duration Babcock, 2005 27 patients (43 feet) Botulinum toxin A (BOTOX) 70u, 0.7mL (n = 22 feet) vs saline 0.7mL (n = 21 feet). One injection at baseline 3,8wk VAS, Maryland Foot Score, Pain relief VAS, pressure algometry Incl: adult, clinical diagnosis of PF, duration < 6mo, failure of conventional treatment. Excl: open wound of the foot or heel, history of foot surgery or fracture, or history of chronic narcotic use, other differential diagnosis 44 N/A Ball, 2013 65 Methylprednisolone acetate 20mg, 1mL, US-guided injection (n=22), palpation- guided injection (n=22) vs 0.9% saline 1mL, US-guided injection (n=21). One injection at baseline 6,12wk VAS, heel tenderness index, plantar fascia thickness Incl: clinical diagnosis of PF, failure of conventional treatment > 8wk. Excl: diagnosis of inflammatory arthritis, prior surgery or trauma to the heel, previous steroid injection. 49 6mo Crawford, 1999 106 Prednisolone acetate 25mg, 1mL + 2% Lignocaine 1mL (n=27) vs Prednisolone acetate 25mg, 1mL + 2% Lignocaine 1mL + tibial nerve block (n=26) vs 2% Lignocaine 2mL (n=27) vs 2% Lignocaine 2mL + tibial nerve block (n=26). One injection at baseline 1, 3,6mo VAS Incl: clinical diagnosis of PF. Excl: age < 18y/o, previous steroid injection within 6 mo. 57 Diaz-Llopis, 2012 56 Botulinum toxin A (BOTOX) 70u, 0.7mL (n = 28) vs Betamethasone 12mg, 2mL, + 1% Mepivacaine 0.5mL (n = 28). One injection at baseline, 2nd cross-over injection optional at 1mo with failed treatment. Final Botulinum toxin A (n=34) vs Betamethasone (n=22) 1,6mo FHSQ Excl: other differential diagnosis, previous injection within 6mo 54

Database collection Study A vs B f/u Outcome Measure Group A n Pre SD Post Post-Pre Group B Zelen 2013 0 vs 6 8 wk FACES Pain Scores Saline 13 8 1.6 4.6 1.2 mDHACM 15 8.6 0.7 Peterlein 2012 0 vs 2 18wk VAS in rest 16 -1.7 3.75 BTX 17 -1.55 3.38 Kim 2014 8 vs 9 28wk FFI-pain PRP 9 60.4 14.7 33.7 23.4 Prolotherapy 11 56.5 14 41.1 21.4 Huang 2010 3mo 25 5.4 0.6 5.2 1 5.9 0.9 2 Ball 2013 0 vs 3 12wk VAS 18 56 27.9 53.8 33.8 Corticosteroid 62 19.2 28.4 24.9 Crawford 1999 4 vs 5 6mo Corticosteroid+Local anesthesia 5.6 2.3 2.4 2.6 Local anesthesia 5.5 2.1 3.3 2.7 Diaz-Llopis 2012 2 vs 3 1mo FHSQ-pain 28 -29.1 19.5 -63.3 21.9 -34.2 21.1 -31.6 21.0 -53.7 31.2 -22.12 27.4 Elizondo-Rodriguez 2013 19 7.1 1.75 1.1 1.5 7.7 1.32 3.8 1.15 Guner 2013 3 vs 7 12mo 30 7.97 1.37 3.17 2.31 NSAID 31 8.26 1.41 2.94 2.04 Kalaci 2009 (A) 3 vs 5 7.24 2.22 0.96 1.24 6.72 1.74 3.4 2.88 Kalaci 2009 (B) 1 vs 3 ABI 6.84 2.27 3.53 3.06 6.96 2.71 1.52 2.14 Kiter 2006 7.6 1.3 1.8 7.28 2.57 2.9 Lee 2007 7.3 3.6 6.9 1.7 3 McMillan 2012 41 -35.8 20.4 -59.7 25.4 -36.8 19.9 -65.4 27.7 Omar, 2012 3 vs 8 6wk 8.8 6.5 8.2 Tiwari, 2013 6.03 0.85 2.8 0.76 0.45

結果為何? Present the results

統計軟體 STATA WINBUG

Krogh et al. The American Journal of Sports Medicine 2012 Injection Therapies in Treating Lateral Epicondylitis Krogh et al. The American Journal of Sports Medicine 2012

嚴格檢視結果 Critical Apprasial

From Oxford Center for Evidenced-Based Medicine. http://www.cebm.net Critical Appraisal of Meta-analysis Validity Were the criteria for inclusion of studies appropriate? 文章的收錄條件是否適當？ Was the search for eligible studies thorough? 搜尋的進行是否夠全面？ Was the validity of the included studies assessed? 是否有收錄的文章進行效度分析? Were the assessments of the studies reproducible? 對文章的分析是否具有再現性? From Oxford Center for Evidenced-Based Medicine. http://www.cebm.net

Critical Appraisal of Meta-analysis Importance What are the overall results of the review? 此篇綜論的總結果為何? Were the results similar from study to study? 各文章間的結果是否相似? How precise were the results? 實驗結果是否夠精確? From Oxford Center for Evidenced-Based Medicine. http://www.cebm.net

Critical Appraisal of Indirect Evidence Transitivity: 間接證據是否真能轉換成直接證據? 病人族群、治療方法、實驗設計是否相似 Consistency: 間接證據與直接證據結果是否相似? 結果間的異質性(heterogeneity)如何? 如有Bias應加以探討並做sensitivity analysis 適時去除掉不正確之證據

Chen HC123, Lin SH4, Huang YM5, Hou WH 46 NMA研究實例 Injection Therapies in Treating Plantar Fasciitis: A Network Meta-analysis of Randomized Controlled Trials Chen HC123, Lin SH4, Huang YM5, Hou WH 46 1Department of Physical Medicine and Rehabilitation, Shuang Ho Hospital, Taipei Medical University, Taipei 2College of Medicine, Taipei Medical University, Taipei 3Center for Evidence-Based Health Care, Shuang Ho Hospital, Taipei Medical University, Taipei 4Department of Physical Medicine and Rehabilitation, Taipei Medical University Hospital, Taipei 5Department of Orthopedics, Shuang Ho Hospital, Taipei Medical University, Taipei 6College of Nursing, Taipei Medical University, Taipei

Plantar Fasciitis (足底筋膜炎) 常見造成足跟痛之原因 早晨常有“ First step pain” 與過度行走、肥胖、足內翻、阿基里斯肌腱過緊有關 自限性、但可能會持續數個月甚至數年 48

治療足底筋膜炎 第一線：非侵入式保守治療 休息、口服NSAIDs、輔具、物理治療、體外震波 第二線：侵入式保守治療 局部注射治療 (Injection therapies) 第三線：手術 只有在困難治療病患 很少施行 49

Injection Therapies 直接注射於足底筋膜上 常見注射物： 價格差異大 效果何者較佳仍未有定論 皮質類固醇、肉毒桿菌素、增生療法 (prolotherapy) 、自體血液、富含血小板血漿(Platelet-Rich Plasma) 價格差異大 效果何者較佳仍未有定論 50

Goal of our study 針對現有文獻做完整的搜尋以找出所有治療足底筋膜炎的RCT 利用network meta-analysis比較所有的注射物療效 51

Methods Databases: Pubmed, Cochrane Liberary, Scopus, CINAHL Keywords: "plantar fasci*" or "plantar heel*" or "plantar foot" or "heel pain" or "painful heel*" or "hindfoot pain" or "painful hindfoot“ Outcome: pain related score (primary), complications (secondary) Quality assessment: Cochrane risk of bias assessment tool 52

Results 53

Figure 1 54

Table 1: Characteristics of the selected trials Data items of Table 1 Number of participants Treatments compared Follow-up period Outcome measures Percentages of female participants Average ages of participants Average disease duration of participants. 55

Table 1: Characteristics of the selected trials (1) 56

Table 1: Characteristics of the selected trials (2) 57

Figure 2: Network of included comparisons 58

Table 2: Quality assessment 59

Table 3: Comparisons of the pain reduction over the 8 injectants 60

Figure 3: Consistency: compare Network (direct+ indirect) with direct Meta

Table 4: Pain score estimate in every treatment Rank Injectants SMD (95% credible interval) 1 mDHACM -6.09 (-15.62 to 3.43) 2 Platelet-Rich Plasma -3.77 (-12.87 to 5.27) 3 Prolotherapy -3.29 (-12.73 to 6.09) 4 Botulium Toxin -2.47 (-11.39 to 6.41) 5 NSAID -2.34 (-11.5 to 6.86) 6 Corticosteroid -1.80 (-10.67 to 7.03) 7 Autologous Blood -1.76 (-10.79 to 7.21) 8 Placebo -0.89 (-9.72 to 7.89) 62

Figure 4: Forest plot Injectants vs Placebo

Figure 5: Ranking

Network Meta-analysis優點與缺點 一台電腦就可以做的study 臨床極具實用價值 剛開始發展，題目俯拾皆是 投 High impact Factor雜誌的利器 缺點 架構較為龐大 統計較為複雜 許多學者仍難接受indirect evidence

謝謝聆聽！