COCHRANE ROB 2 · 偏誤風險評讀
4 週總睡眠時間變化
← 回到「失眠」的圖表
總體判斷 OVERALL-High(高風險)
Domain 與 overall 判斷由程式依 Cochrane RoB 2 官方演算法計算;AI 只負責回答下方各訊號問題。符號:+ 低風險、? 有些疑慮、- 高風險(不只靠顏色)。
風險偏誤紅綠燈
- +Domain 1 隨機化過程:Low(低風險)
- +Domain 2 偏離預定介入:Low(低風險)
- -Domain 3 缺失結局資料:High(高風險)
- +Domain 4 結局測量:Low(低風險)
- +Domain 5 選擇性報告:Low(低風險)
- -總體(Overall):High(高風險)
中文摘要(給一般讀者)
這篇論文的偏誤風險與 limitation
這是雙中心、中央隨機、三盲、安慰劑對照的平行組試驗,比較辨證論治個別化中藥顆粒與安慰劑對原發性失眠 4 週總睡眠時間的影響。隨機與分派隱匿做得好,基線相近。病人、醫師、評估者皆設盲,安慰劑做成 4 種口味輪流發放以模擬處方調整,偏離介入的風險低,且主要分析採 ITT。主要疑慮在缺失資料:約 12% 受試者退出,中藥組退出者全因「自覺無效」,卻以隨機缺失假設做多重插補,可能高估中藥效果。另外,安慰劑組在第一次期中分析就提前終止,但報告的 P 值(0.001)並未跨過事前訂的第一次期中門檻(0.0002),提前因療效終止的試驗也常高估效果;期中分析後的 40 位直接分到中藥組、未經隨機,本評讀只評前段真正隨機的 76 位。整體而言,偏誤主要來自缺失資料的處理。
詳細評讀(給專業讀者)
Domain 1:隨機化過程 — 判斷:Low
偏誤來源:隨機分組的過程——分配序列是否隨機產生、是否隱蔽到分組為止、基線(第一期開始時)是否平衡。
1.1 Was the allocation sequence random?
分配序列是否隨機產生?
- 回答:是(Yes — 有明確證據支持)
- 原文引述:Treatment randomization was conducted with a central randomization system (Institute of Basic Research in Clinical Medicine, CACMS) by a specified investigator who was not involved in any other study procedures. ... Block randomization and complete randomization were used for the treatment assignment and the visit order for 3 CM clinicians for each participant, respectively.治療分派由中央隨機系統(中國中醫科學院中醫臨床基礎醫學研究所)執行,負責的研究人員不參與其他任何研究程序。……治療分派採區塊隨機,每位受試者看 3 位中醫師的順序採完全隨機。
- 理由:使用中央隨機系統與區塊隨機法產生分派序列,屬於真正隨機的方法。
1.2 Was the allocation sequence concealed until participants were enrolled and assigned to interventions?
分配是否隱蔽到分組為止?
- 回答:是(Yes — 有明確證據支持)
- 原文引述:Treatment randomization was conducted with a central randomization system (Institute of Basic Research in Clinical Medicine, CACMS) by a specified investigator who was not involved in any other study procedures.治療分派由中央隨機系統執行,負責的研究人員不參與其他任何研究程序。
- 理由:中央隨機系統由不參與收案與評估的專人操作,收案者在納入前無法得知下一位的分組,分派隱匿適當。
1.3 Did baseline differences between intervention groups suggest a problem with the randomization process?
(第一期開始時)基線差異是否暗示隨機化有問題?
- 回答:否(No — 有明確證據支持)
- 原文引述:A total of 116 participants were randomized to the CM group (n =96) or placebo group (n =20, Figure 1); baseline characteristics were similar (Table 1).共 116 位受試者分入中藥組(96 位)或安慰劑組(20 位);基線特徵相近(表 1)。
- 理由:表 1 期中分析的中藥組 56 位與安慰劑組 20 位,年齡、性別、Spiegel 分數、TST(5.2 vs 5.4 小時)、PSQI 等都相近,看不出隨機出問題。56:20 接近「3 位醫師+安慰劑」四組各一份的區塊隨機預期比例。要注意的是論文把期中分析後直接分到中藥組的 40 位也算進「隨機 116 位」,這 40 位其實沒有和安慰劑同步隨機;本評讀只評真正隨機的 76 位,網站圖表數值也來自這一段(表 2)。
Domain 2:偏離預定介入 — 判斷:Low
偏誤來源:實際執行時偏離預定的介入——盲性如何、是否依「隨機分到的組別」分析(ITT)。
2.1 Were participants aware of their assigned intervention during the trial?
參與者是否(在每一期)知道自己的分組?
- 回答:可能否(Probably no — 綜合判斷後傾向「否」)
- 原文引述:Therefore, a randomized, placebo-controlled, triple-blinded (patients, clinicians, and outcome assessors) parallel-group trial was designed ... 4 placebos with different tastes, colors, and odors packaged identically to those in the CM group were produced.因此設計成隨機、安慰劑對照、三盲(病人、醫師、結局評估者)的平行組試驗……製作 4 種味道、顏色、氣味不同的安慰劑,包裝與中藥組相同。
- 理由:病人設盲。安慰劑含 10% 原方並加色素、苦味劑、焦糖,做成 4 種口味輪流給,模擬辨證處方隨診次變動,事前也測試過安慰劑品質。但論文沒有報告試驗後的盲性檢驗,個別化中藥與固定安慰劑仍可能被察覺,故答「可能否」。
2.2 Were carers and people delivering the interventions aware of participants' assigned intervention during the trial?
照護/執行者是否(在每一期)知道分組?
- 回答:可能否(Probably no — 綜合判斷後傾向「否」)
- 原文引述:Participants visited all 3 clinicians at the site during each visit in a random order. According to their own CM theories and experiences, each clinician provided individual treatment for every participant受試者每次回診時依隨機順序看該中心全部 3 位醫師,每位醫師依自己的理論與經驗為每位受試者開立個別處方。
- 理由:三位醫師對每位受試者(包括安慰劑組)都照常辨證開方,不知道誰的處方會真正發給病人、也不知道病人是否拿到安慰劑,論文亦稱醫師設盲。施治者知情的可能性低。
2.3 Were there deviations from the intended intervention that arose because of the trial context?
是否因試驗情境造成偏離?
- 回答:不需評估(NA)
- 原因:條件題:依 2.1/2.2 的答案,本題依 RoB 2 官方路由不需評估(NA)。
2.4 Were these deviations likely to have affected the outcome?
這些偏離是否可能影響結局?
- 回答:不需評估(NA)
- 原因:條件題:依 2.3 的答案,本題依 RoB 2 官方路由不需評估(NA)。
2.5 Were these deviations from intended intervention balanced between groups?
這些偏離在兩組間是否平衡?
- 回答:不需評估(NA)
- 原因:條件題:依 2.4 的答案,本題依 RoB 2 官方路由不需評估(NA)。
2.6 Was an appropriate analysis used to estimate the effect of assignment to intervention?
是否用了適當分析(ITT)?
- 回答:是(Yes — 有明確證據支持)
- 原文引述:All analyses were conducted in an intention- ... to-treat population, which included all randomized participants.所有分析都在意向治療族群進行,涵蓋所有隨機分派的受試者。
- 理由:主要分析為 ITT,納入所有隨機受試者並以多重插補補缺值,屬於估計指派效果的適當分析。
2.7 Was there potential for a substantial impact (on the result) of the failure to analyse participants in the group to which they were randomized?
未依隨機分組分析,是否可能對結果造成重大影響?
- 回答:不需評估(NA)
- 原因:條件題:依 2.6 的答案,本題依 RoB 2 官方路由不需評估(NA)。
Domain 3:缺失結局資料 — 判斷:High
偏誤來源:結局資料缺失——是否幾乎所有人都有資料、缺失是否可能與病情(結局真實值)相關。
3.1 Were data for this outcome available for all, or nearly all, participants randomized?
是否幾乎所有受試者都有此結局的資料?
- 回答:否(No — 有明確證據支持)
- 原文引述:Discontinued (n =6) Reported no efficacy by patient (n =6) ... Discontinued (n =3) Had no time to continue (n =2) Reported no efficacy by patient (n =1)(中藥組)退出 6 位,皆為病人自述無效;(安慰劑組)退出 3 位,2 位沒時間繼續、1 位自述無效。
- 理由:期中分析段中藥組 56 位退出 6 位(10.7%),安慰劑組 20 位退出 3 位(15%),合計 9/76(11.8%)缺主要結局。TST 缺值用多重插補補上,但依指引插補值不算「有結局資料」,不符「幾乎所有受試者」。
3.2 Is there evidence that the result was not biased by missing outcome data?
是否有證據顯示結果未受缺失資料偏誤?
- 回答:可能否(Probably no — 綜合判斷後傾向「否」)
- 原文引述:Missing data were assumed to be missing at random and imputed using multiple imputations (MI). Except for SOL, WASO and sleep efficiency, all variables were imputed using MI.缺失資料假設為隨機缺失,以多重插補處理;除 SOL、WASO、睡眠效率外,所有變項都以多重插補補值。
- 理由:多重插補建立在「隨機缺失」假設上,本身無法排除缺失偏誤;論文的重複測量模型只是換分析方法,沒有針對缺失機制做敏感度分析,無證據顯示結果未受缺失資料影響。
3.3 Could missingness in the outcome depend on its true value?
缺失是否可能取決於結局真實值?
- 回答:是(Yes — 有明確證據支持)
- 原文引述:Discontinued (n =6) Reported no efficacy by patient (n =6)(中藥組)退出 6 位,皆為病人自述無效。
- 理由:中藥組退出者全數因「自覺無效」,缺失與睡眠改善程度(結局真值)直接相關。
3.4 Is it likely that missingness in the outcome depended on its true value?
缺失是否很可能取決於結局真實值?
- 回答:可能是(Probably yes — 綜合判斷後傾向「是」)
- 原文引述:Discontinued (n =6) Reported no efficacy by patient (n =6) ... Discontinued (n =3) Had no time to continue (n =2) Reported no efficacy by patient (n =1)中藥組 6 位退出皆因自述無效;安慰劑組 3 位退出中 2 位因沒時間、1 位因無效。
- 理由:兩組退出原因分布不同:中藥組 6/6 因無效,安慰劑組只有 1/3。以隨機缺失假設插補中藥組這 6 位,等於假設他們的改善跟留下的人差不多,可能高估中藥組效果。人數不多,但依指引退出原因與結局相關且兩組不同,答「可能是」。
Domain 4:結局測量 — 判斷:Low
偏誤來源:結局的測量——測量工具是否恰當、評估者是否盲化、評估是否可能受分組知情影響。
4.1 Was the method of measuring the outcome inappropriate?
結局測量方法是否不適當?
- 回答:否(No — 有明確證據支持)
- 原文引述:The primary outcome was TST from baseline based on the result of the pilot study(16) and previous research of insomnia, (3-5,8) assessed using a Sleep Disorders Institute sleep diary.主要結局為總睡眠時間相對治療前的變化,依前導研究與既往失眠研究選定,以睡眠障礙研究所睡眠日誌評估。
- 理由:睡眠日誌記錄 TST 是失眠試驗常用且適當的測量方法。
4.2 Could measurement or ascertainment of the outcome have differed between intervention groups?
(每一序列內)結局測量是否可能因介入而異?
- 回答:否(No — 有明確證據支持)
- 原文引述:The mean daily TST over the past 4 weeks was evaluated. TST was calculated as the awakening time subtracted from the sleep onset time.評估過去 4 週每日 TST 的平均值,以醒來時間減去入睡時間計算。
- 理由:兩組用同一份睡眠日誌、同一種計算方式與同樣的時間點,測量方式沒有組間差異。
4.3 Were outcome assessors aware of the intervention received by study participants?
結局評估者是否知道受試者接受的介入?
- 回答:可能否(Probably no — 綜合判斷後傾向「否」)
- 原文引述:Therefore, a randomized, placebo-controlled, triple-blinded (patients, clinicians, and outcome assessors) parallel-group trial was designed因此設計成隨機、安慰劑對照、三盲(病人、醫師、結局評估者)的平行組試驗。
- 理由:TST 由病人自填睡眠日誌,評估者就是病人本人;病人使用外觀一致、輪換口味的安慰劑而設盲,負責收集結局的醫師也設盲。未做盲性檢驗,故答「可能否」。
4.4 Could assessment of the outcome have been influenced by knowledge of intervention received?
結局評估是否可能受介入知情影響?
- 回答:不需評估(NA)
- 原因:條件題:依 4.3 的答案,本題依 RoB 2 官方路由不需評估(NA)。
4.5 Is it likely that assessment of the outcome was influenced by knowledge of intervention received?
結局評估是否很可能受介入知情影響?
- 回答:不需評估(NA)
- 原因:條件題:依 4.4 的答案,本題依 RoB 2 官方路由不需評估(NA)。
Domain 5:選擇性報告 — 判斷:Low
偏誤來源:選擇性報告結果——是否依事前訂定的分析計畫分析、有無從多個結果中挑選有利者報告(交叉另含:是否因偵測到殘留而只報第一期)。
5.1 Were the data that produced this result analysed in accordance with a pre-specified analysis plan that was finalized before unblinded outcome data were available for analysis?
是否依事前訂定(取得未盲化資料前定版)的分析計畫分析?
- 回答:可能是(Probably yes — 綜合判斷後傾向「是」)
- 原文引述:Prior to enrollment, the trial was registered at ClinicalTrials.gov (NCT01613183). The protocol was published by Yan, et al. ... For the primary outcome (TST changes from baseline at 4 weeks), the t -test was used according to the protocol.試驗在收案前已於 ClinicalTrials.gov 登錄(NCT01613183),計畫書由 Yan 等人發表。……主要結局(4 週 TST 相對治療前變化)依計畫書使用 t 檢定。
- 理由:收案前登錄並發表計畫書,主要結局的分析方法明言依計畫書。但要注意:論文事前訂的第一次期中分析 O'Brien-Fleming 門檻是 P=0.0002,而期中分析 TST 的 P=0.001 並未跨過此門檻,安慰劑組卻「依事前標準」提前終止;此提前終止的依據與計畫書是否一致無法從本文確認,提前因療效終止也可能高估效果。主要結局的分析本身仍符合計畫書,故答「可能是」。
5.2 Is the numerical result being assessed likely to have been selected, on the basis of the results, from multiple eligible outcome measurements (e.g. scales, definitions, time points) within the outcome domain?
此數值結果是否可能依結果,從多個量表/定義/時間點中選出?
- 回答:否(No — 有明確證據支持)
- 原文引述:The primary outcome was change in total sleep time (TST) from baseline.主要結局為總睡眠時間相對治療前的變化。
- 理由:主要結局與測量時間點(4 週 TST 變化)在計畫書與樣本數估算中已預先訂定,論文也同時報告 4 週 TST 絕對值與追蹤期結果,無挑選量測的跡象。
5.3 Is the numerical result being assessed likely to have been selected, on the basis of the results, from multiple eligible analyses of the data?
此數值結果是否可能依結果,從多種分析中選出?
- 回答:可能否(Probably no — 綜合判斷後傾向「否」)
- 原文引述:For the primary outcome (TST changes from baseline at 4 weeks), the t -test was used according to the protocol. To assess robustness, a repeated-measures model was adopted主要結局依計畫書用 t 檢定;為評估穩健性另採重複測量模型。
- 理由:主要分析依計畫書,並同時報告敏感度分析與二分法(≥0.5 小時)結果,數值一致,看不出從多種分析中挑選。