FDAによるPROの指針 (草稿)
臨床試験において,ある治療が有効である
と解釈できる,患者報告式アウトカム尺度
により観察された,治療群間の差や変化量
» Minimumimportant difference (MID) — The amount of difference or
change observed in a PRO <Patient-Reported Outcome> measure
between treatment groups in a clinical trial that will be interpreted as a
treatment benefit.
4
U.S. Department of Health and Human Services: Guidance for Industry: Patient-Reported Outcome
Measures: Use in Medical Product Development to Support Labeling Claims. 2006
(http://www.fda.gov/ohrms/dockets/98fr/06d-0044-gdl0001.pdf)
臨床的有意性の定義 (群間の差)
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
5.
FDAによるPROの指針 (最終版)
標的集団において,事前に定められた期間
に,重要な治療の有効性が得られたと判断
できる,ある患者個人のある尺度の得点の
変化
» Responderdefinition — A score change in a measure, experienced by
an individual patient over a predetermined time period that has been
demonstrated in the target population to have a significant treatment
benefit.
5
臨床的有意性の定義 (個人の変化)
U.S. Department of Health and Human Services: Guidance for Industry: Patient-Reported Outcome
Measures: Use in Medical Product Development to Support Labeling Claims. 2009
(http://www.ispor.org/workpaper/FDA%20PRO%20Guidance.pdf)
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
6.
混乱した用語の使い分け
群間の差 個人の変化 事例
MIDResponder [1-2]
MID MIC [3]
MCID MID [4]
MCID --- [5]
MCID [6]
注) MID = Minimal Important Difference; MIC = Minimally Important
Change; MCID = Minimal Clinically Important Difference
6
[1] U.S. Department of Health and Human Services: Guidance for Industry: Patient-Reported Outcome Measures: Use
in Medical Product Development to Support Labeling Claims. 2006 (http://www.fda.gov/ohrms/dockets/98fr/06d-
0044-gdl0001.pdf)
[2] U.S. Department of Health and Human Services: Guidance for Industry: Patient-Reported Outcome Measures: Use
in Medical Product Development to Support Labeling Claims. 2009
(http://www.ispor.org/workpaper/FDA%20PRO%20Guidance.pdf)
[3] de Vet HC et al: J Rheumatol. 2006 Feb;33(2):434-5
[4] Kirby S et al: J Biopharm Stat. 2010 Sep;20(5):1043-54.
[5] Chan LS: Am J Public Health. 2013 Nov;103(11):e24-5.
[6] Beaton DE et al: J Rheumatol. 2001 Feb;28(2):400-5.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
7.
群間の差の決定の難しさ
群間の差の重要性は,「疾患」「利用できる
治療法」「治療のリスクベネフィット比」な
ど,広いコンテクストを基にしなければ,決
定できない!
» ...(前略) thedetermination of the importance of group differences,
which can only be established in the broader context of the disease
being treated, the currently available treatments, and the overall risk-
benefit ratio of the treatment.
7Dworkin RH et al: J Pain. 2008 Feb;9(2):105-21.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
測定誤差の記載事例
方法 (二時点以上の測定/実施間隔)
» Usinga web-based system at home, the patients completed an online
questionnaire containing the four different body part-specific PROMs
[Patient-Reported Outcome Measures] at three different time-points:
T1(baseline), T2 (2 weeks after baseline), and T3 (6-month follow-up).
方法 (実施の類似性)
» The given questionnaires were identical at all three time-points, except for
two anchor questions added at T3 (see ‘ Outcome measures ’ for details).
方法 (状態の安定性)
» Data from T1 and T2 were used to determine the measurement error. We
assumed that there would be no real change in a patient's functioning
within a 2-week interval (range, 1 to 4 weeks).
20van Kampen DA et al : J Orthop Surg Res. 2013 Nov 14;8:40.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
測定誤差の記載事例
方法 (統計手法)
» Thereliability of the NDI-DV [Neck Disability Index-Dutch Version] was
assessed by rating test-retest reliability and measurement error [11]. The
test–retest interval was set at 10 days. As parameter of reliability of the
NDI-DV, the intra-class correlation coefficient (ICCagreement) was computed
using a two-way random effects model (ICCagreement = σ2
p/[σ2
p +σ2
m +σ2
r]),
where the error variance consists of a component representing the
systematic difference between the two measurements (σ2
m) and a
component for the residual (random) error (σ2
r) [12]. σ2
p represents the
systematic differences between the ‘‘true’’ scores of patients.
27Ailliet L et al: Eur Spine J. 2014 May 17
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
28.
測定誤差の記載事例
方法 (統計手法)
» Wequantified the measurement error by the Bland and Altman method
and by calculating the standard error of measurement (SEMagreement) from
the ICC formula, by taking the square root of the error variance (√[σ2
m
+σ2
r]).
結果 (級内相関係数/測定誤差)
» At the measurement time point of 6 months + 10 days, 155 patients
(60.5 %) provided the information used to assess the reliability. An intra-
class correlation coefficient (ICCagreement) of 0.88 was found.
(中略)
» The SEMagreement was 1.95.
28Ailliet L et al: Eur Spine J. 2014 May 17
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
①検出可能な最小限の変化 (MDC)
MDC =weight × SEM [1,2]
weight = 1/1.96/2.77が慣例
SEM = 測定誤差
31
[1] Turner D et al: J Clin Epidemiol. 2010 Jan;63(1):28-36.
[2] Wyrwich KW et al: J Clin Epidemiol. 1999 Sep;52(9):861-73.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
32.
方法
» The smallestdetectable change (SDC) was based on the SEMagreement. To be
95% confident that the observed change is not caused by measurement
error but can be considered real change, the SDC at individual level
(SDCind) was calculated as 1.96 × √2 × SEMagreement. The SDC expresses the
magnitude of change—with a probability of 5 %—that this change is due
to measurement error. Given this small probability, it is likely that a patient
whose score exceeds the SDC has changed [14].
結果
» The SEMagreement was 1.95. Based on this SEM, a SDC value of 5.40 was
found (calculated as 1.96 × √2 × SEMagreement).
32
①記載事例
Ailliet L et al: Eur Spine J. 2014 May 17
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
②記載事例
方法
» The distribution-basedMCID[Minimal Clinically Important Difference]
was estimated using the Cohen effect size benchmark. An effect size of
0.5 (i.e., 0.5 SD of the baseline score) indicated an important change
and was used as the minimum CID[Clinically Important Difference] in
this study.
34Lin KC et al: Res Dev Disabil. 2012 Mar-Apr;33(2):570-7.
0.5SD
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
43 5 6721
**の治療前と 比べて,
いまの自分の状態を評価して下さい
少
し
改
善
完
全
に
改
善
不
変
少
し
悪
化
か
な
り
悪
化
以
前
よ
り
悪
化
か
な
り
改
善
①変化量の平均値
アンカーで「最小限の変化群」を定義
» The MIC was defined as the mean change score in the subcategory of
patients who were “slightly improved” according to the anchor scores,
and the CI was calculated [5,25].
44
最小限の変化群
van Kampen DA et al : J Orthop Surg Res. 2013 Nov 14;8:40.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
②変化群と不変群の平均値差
群ごとに変化量の平均値を算出
» First, foreach category (much improved ― improved ―unchanged ―
worse) the mean change in VascuQol[Vascular Quality of Life] sumscore
was calculated.
47Frans FA et al: Eur J Vasc Endovasc Surg. 2014 Feb;47(2):180-6.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
48.
②変化群と不変群の平均値差
「変化群」と「不変群」の変化量の平均
値差をMICとする
» Then theMID for improvement and the MID for deterioration was
calculated as the difference between the mean change in VascuQol
sumscore of the improved or deteriorated patients and the mean
change in VascuQol sumscore of the unchanged patients.
48Frans FA et al: Eur J Vasc Endovasc Surg. 2014 Feb;47(2):180-6.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
49.
③ROC曲線
アンカーで「変化群」と「不変群」を定義
» Patients thatstated their overall satisfaction after treatment was either
“very satisfied” or “somewhat satisfied” were classified as satisfied. All other
patients were classified as being unsatisfied.
49London DA et al: Plast Reconstr Surg. 2014 Mar;133(3):616-25.
全体的な満足度
1. とても満足
2. まあまあ満足
3. 満足でも不満でもない
4. いくぶん不満
5. とても不満
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
50.
③ROC曲線
アンカーを確定基準,変化量をインデック
ス検査とし,感度と特異度を最大化する変
化量のカットオフ値をMICとする
» Once weidentified which subscales had discriminative ability, we then
determined the change in score (overall Michigan Hand Outcomes
Questionnaire and all subscales—the satisfaction subscale could only be
assessed by the “overall satisfaction” method) that provided the greatest
specificity and sensitivity in determining satisfaction by the minimum
distance method, and this resultant score was defined as the minimal
clinically important difference.
50London DA et al: Plast Reconstr Surg. 2014 Mar;133(3):616-25.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
51.
③ROC曲線
アンカーを確定基準,変化量をインデック
ス検査とし,感度と特異度を最大化する変
化量のカットオフ値をMICとする
» When satisfactionwas defined strictly by patients’ answers to their level of
satisfaction with their overall hand function, 135 subjects were classified as
satisfied and 51 were classified as not satisfied, with an area under the
curve of 0.85 (95 percent CI, 0.80 to 0.91) and a minimal clinically important
difference of 11.5.
51London DA et al: Plast Reconstr Surg. 2014 Mar;133(3):616-25.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
52.
④変化量の増分
1単位の意味が明瞭なアンカーを測定
» One questionassessed patient satisfaction with their pain relief (‘‘How
well did the surgery relieve pain in your affected joint?’’), and the other
assessed their satisfaction with their functional outcome (‘‘How well did
the surgery increase your ability to perform regular activities?’’). The
response to each of these questions was recorded using a 5-point Likert
scale: excellent, very well, well, fair, and poor.
52Clement ND et al: Knee Surg Sports Traumatol Arthrosc. 2014 Aug;22(8):1933-9.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
53.
④変化量の増分
アンカーを独立変数,変化量を従属変数
とした回帰式の偏回帰係数をMICとする
» Simple linearregression analysis was used to identify the MCID, using
the slope of the line for the change according to level of satisfaction, in
the OKS [Oxford knee score] and SF-12 PCS[physical component
summary] score for both pain and function.
» Simple linear regression identified the MCID for pain relief in the OKS to
be 5.0 (95 % CI 4.4–5.5) and for the SF-12 PCS to be 4.5 (95 % CI 3.9–
5.2).
53Clement ND et al: Knee Surg Sports Traumatol Arthrosc. 2014 Aug;22(8):1933-9.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
54.
Dennis Revickiの推奨
54
推奨事項
1. 複数のアンカーを使用すること
2.PROの変化量との相関が0.30-0.35以上のアンカーを選択すること
3. 臨床試験の結果を参照すること
4. 分布に基づく方法は参考程度にすること
5. 治療者の観点よりも患者の観点の臨床的有意性を重視すること
6. 複数の臨床的有意性の値を図示すること
7. 最終的な値は合意形成により決定すること
注) PRO = Patient Reported Outcome
Revicki D et al: J Clin Epidemiol. 2008 Feb;61(2):102-9.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
FDAによる累積分布関数の推奨
累積分布関数は,変化量をx軸,その変化を示
した患者の割合をy軸とする図である。この図
は,カテゴリカルな判断を与える臨床的有意
性よりも望ましい。多様な臨床的有意性の定
義は,累積分布反応曲線により同定できる。
» Such cumulativedistribution displays show a continuous plot of the
percent change from baseline on the X-axis and the percent of patients
experiencing that change on the Y-axis. This display type may be
preferable to attempting to provide categorical definitions of
responders. A variety of responder definitions can be identified along
the cumulative distribution of response curve.
56
U.S. Department of Health and Human Services: Guidance for Industry: Patient-Reported Outcome
Measures: Use in Medical Product Development to Support Labeling Claims. 2009
(http://www.ispor.org/workpaper/FDA%20PRO%20Guidance.pdf)
特徴 測定誤差 分布 アンカー 累積分布 推薦資料
臨床的有意性の定め方ではない!
累積分布関数は,アンカーに基づく方法や分
布に基づく方によって,注意深く臨床的有意
性の値を定めることの代替策とはみなせない
» However, wedo not see the cumulative distribution functions as a
replacement for the careful investigation of a PRO’s relevant responder
threshold using anchor-based methods, and supported by distribution-
based methods.
58Wyrwich KW et al: Qual Life Res. 2013 Apr;22(3):475-83.
特徴 測定誤差 分布 アンカー 累積分布 推薦資料