シャドーイング練習: Sequential, Multiple Assignment, Randomized Trial Designs With Dr Kidwell - 動画で英語スピーキングを学ぶ
練習モードを選ぶ
各モードは全画面の練習ルームで開きます。
154 文
文が短すぎたり長すぎる場合は、Editをタップして調整してください。
1
Hello, and welcome to JAMA Evidence, our podcast series based on core issues in evidence -based medicine.
2
Today, we will be discussing sequential, multiple assignment, randomized trial designs.
3
I'm your host, Dr. Roger Lewis, statistical editor for JAMA and co -editor of the JAMA Guide to Statistics and Methods series.
4
I'm here with Dr. Kelly Kidwell.
5
Dr. Kidwell, thank you so much for joining us.
6
Can you go ahead and tell our listeners a little bit about yourself?
7
Sure.
8
Thank you so much for having me, Roger.
9
I'm Kelly Kidwell.
10
I'm a professor of biostatistics at the University of Michigan School of Public Health, and my expertise is in clinical trial design and analysis,
11
and in particular, the sequential multiple assignment randomized trials or SMART design.
12
Well, again, thank you so much for joining us.
13
So the name of this approach to a trial design is a bit daunting, so let's try to break it up a little bit.
14
So for each of those terms, can you go ahead and tell us what it means and then how we put it together in designing a clinical trial?
15
I completely agree with you that the vocabulary around smart design is quite daunting.
16
So what an excellent first question to start with.
17
So sequential is about a few things.
18
One is that we're interested in sequences of treatments.
19
And in this clinical trial design, we're going to have sequential randomization.
20
So those terms aren't next to each other, but that's really what's happening and defines this trial,
21
such that at least some group of participants will be randomized twice in sequence.
22
So usually there's some upfront randomization
23
and then a subsequent randomization for some group of participants based on their response to the initial treatment.
24
And so that multiple assignment comes into the fact that we'll have this sequential treatment assignment.
25
So individuals could be assigned to multiple treatments.
26
And this is all wrapped up in a clinical trial.
27
So we're going to use this SMART clinical trial tool to give us robust evidence for sequences of treatments
28
that adapt to individual responses or patient status.
29
So in contrast, in a traditional or a simpler clinical trial design.
30
We just randomize once, and then the patient either does well
31
or perhaps doesn't do so well on the treatment to which they are assigned.
32
And if they don't do well, we often have some sort of rescue therapy or they have to leave the trial.
33
So how is the experience of a patient or a participant different in a SMART design?
34
Right.
35
So in a general, standard, traditional randomized clinical trial, we're really looking at one point in time,
36
sort of this trial in a vacuum.
37
The SMART is going to say, let's look at that patient over the course of care.
38
So in a standard trial where you might stop collecting their data and treating them off trial if they didn't respond, we actually want to use that data.
39
And in fact, we might want to randomize the treatment to
40
which rescue therapy they're getting so that we can collect unbiased evidence
41
and use causal inference to better understand what the effect of that treatment is.
42
And not just that treatment, but the entire treatment sequence.
43
And so really for the patient,
44
it might not be too different than a standard trial because we're going to treat you over the course of your care.
45
But to the investigators, we're going to make sure that
46
that data is collected for our trial and there might be randomization for the treatment assignment that follows for that patient.
47
So, let's think a little bit about how we analyze or what we can learn from those two different approaches.
48
So in the first, we randomize at a single point in time, and if the patient doesn't do well,
49
they either drop out of the trial or we assign a rescue therapy.
50
Whereas in the SMART design, if they require a change in therapy, we actually randomize what we do.
51
So what's different about what we can learn from those two different approaches?
52
That's right.
53
Often in our standard clinical trial, in that one stage where the patient drops out or receives clinician or patient choice therapy,
54
we no longer make inference or have any decisions or really robust results about that subsequent treatment or their outcome.
55
Whereas in the SMART, by randomizing that individual and continuing to collect that data,
56
we will be able to have really robust evidence over the course of care, over the sequence of treatments
57
that could have adapted based on their response to allow us
58
to give these sequential guidelines to clinicians to treat the patient not just up front at one point in time,
59
but over time, and really tailor it to how they're doing, if they responded, if they adhered, if they engaged with that treatment.
60
So in a sense, this more closely follows what happens in actual clinical care,
61
where we make a series of treatment decisions based on the patient's prior response, but in this case, you're setting it up
62
so we can rigorously learn about what is the best next
63
thing to do based on what's happened to the patient so far.
64
Is that a fair summary?
65
Absolutely.
66
We are looking at a type of clinical trial that more closely mimics the way in which treatment is practiced.
67
And so that we can provide evidence for these treatment strategies
68
that adapt to individual's response or patient status in the way that we would actually treat them in practice.
69
Let's suppose that a SMART design is run and there's two points at which patients are randomized,
70
maybe the initial therapy and then after some defined period of follow -up.
71
What does the result look like in the publication that results,
72
and how can clinicians utilize the results of the SMART in guiding their own clinical care?
73
One of the amazing advantages of a SMART design is
74
that it can answer many questions as opposed to just what is the best treatment at one time.
75
We can think about what is the best treatment to begin with?
76
What is the best subsequent treatment for those who respond?
77
What is the best subsequent treatment for those who don't respond?
78
And then we can think about that entire sequence of, okay, so what do we start with?
79
How do we change?
80
When do we change?
81
For whom do we change and what treatment do we give them?
82
And so we can really get from SMARTS this sequence of treatments adapted to patient response.
83
Not only that, we can collect baseline information, so information up front at the start of the trial, and information about the patient and how they're doing with the treatment,
84
how they're adhering, how they're engaging, if they're having side effects,
85
and use that information to better tailor the treatment assignment to that individual so that after the trial, right, we have all this information.
86
And we can look at what are the best sequences of treatments for which particular individuals.
87
The fact that the SMART design more closely matches actual clinical care,
88
which the sequences of therapies we offer patients is informed by their responses to prior therapies is an obvious advantage.
89
But what are some of the limitations or challenges in either the implementation or the interpretation of a SMART clinical trial?
90
I think there are a few limitations and also just lack of knowledge of this design and its opportunities that it offers.
91
Some of the limitations are that individuals think, oh, now that we have multiple randomizations or multiple assignments in sequence,
92
this is now logistically infeasible, or you're going to lose more people if you have to consent multiple times.
93
So one thing is that we often just consent once at the beginning of the smart design.
94
and we may randomize multiple times.
95
Alternatively, you could randomize up front to that sequence that they may follow, whether they respond or do not respond.
96
So logistically, this doesn't have to be very difficult.
97
We also have lots of technology that helps us with complex trials, if you want to say that it's complex, although it's not.
98
Number two is that individuals are worried about bias coming into play from a smart design
99
because of these multiple randomizations and the subsequent randomization depending upon patient response or status.
100
This is not a concern because we do not compare these treatment pathways or you first get A,
101
you respond and you continue versus you first get A, you do not respond and you switch to B.
102
We're not going to compare those people.
103
We know that they're different.
104
One responded and one didn't, and we can have some issues if we do those comparisons.
105
We're looking at what we call adaptive interventions or dynamic treatment regimens that include that entire treatment sequence that's adapted.
106
So start with A, and if you respond, continue.
107
And if you don't, receive B.
108
We're going to think about what are the effects of that entire sequence that's tailored to the response.
109
Another misconception is that a SMART is an adaptive design.
110
Usually a SMART is a fixed design, not an adaptive design, meaning
111
that we're not going to use information from patients who have
112
been through the trial to change the trial in some way for future patients.
113
So we're not going to change randomization probabilities or eligibility criteria based on the information we've gained from the trial thus far.
114
What does happen in a SMART is that patients' treatment might change based on their own response.
115
So there might be a within -person adaptation, but there's no between -person adaptation.
116
And so generally, these aren't adaptive designs.
117
Since we're trying to learn more from a clinical trial that's a smart design versus a traditional design
118
that just has a single point of randomization, it seems like we might need larger sample sizes to answer the additional questions.
119
What are the challenges, if any, with respect to sample size requirements for a smart design?
120
If you see pictures of smart designs, they look a little bit intimidating because there's a fair amount of branching that occurs due to these sequential randomizations.
121
And often investigators are quite worried that we're going to require thousands of patients.
122
The thing is that to size a SMART is just like sizing any other trial.
123
And so it's really dependent on the primary aim, which can be similar to a standard one -stage trial and compare,
124
for example, the initial treatments.
125
Alternatively, the aim can be based on those embedded tailored sequences of treatments or the adaptive interventions.
126
And for those perhaps more patients may be required than a standard trial.
127
However, this really isn't an apples -to -apples comparison of a SMART design to a standard or traditional one -stage trial.
128
As I mentioned before, the questions
129
that we can answer in a SMART are much different and many more than we could with a standard design.
130
And so perhaps we may need more depending on the primary aim, but we're going to get a lot more information,
131
data, robust results, and more practical information that we can use after a smart design.
132
In your article, you discuss a study that was performed by Fortney et al that used a smart approach.
133
Can you tell us a little bit about what they did, what they found, and how they analyzed their data?
134
Sure.
135
So in this study, they had a design where individuals were sequentially randomized to better understand some effects of direct
136
and indirect telecare.
137
And what they did is they answered questions that are sort of more similar to one stage design.
138
So they said, okay, so what's the best first thing to start with?
139
And then subsequently, what do we do if individuals aren't responding well to that initial treatment or intervention.
140
So they haven't yet pieced together that full sequence.
141
It seems like that's subsequent analyses for this group in terms of figuring out,
142
so what is the best overall sequences of treatment that can adapt to the individuals?
143
While they answered those individually, what they could find when they move forward with the analysis of the adaptive intervention is
144
that there might be interactions between the initial treatment
145
and the subsequent treatment for
146
which we can only really quantify in these types of SMART designs as opposed to individual randomized clinical trials.
147
And so what they likely will subsequently do is look at
148
that sequence of treatment over the course of the trial
149
or those adaptive interventions to see if there are any interactions between the first step and the second step.
150
I'm Roger Lewis, and I've been speaking with Professor Kelly Kidwell about sequential multiple assignment randomized trials.
151
You can find a link to her article in this episode's description.
152
This episode was produced by Daniel Musisi at the JAMA Network.
153
To follow this and other JAMA Network podcasts, please visit us online at jamanetworkaudio .com.
154
Thanks for listening.
アプリをダウンロード
流暢に話すために必要なすべて
AI発音採点文ごとに採点
IPA練習すべての音をマスター
単語学習自分の単語帳を作る
単語ゲーム遊びながら覚える
📺 同じチャンネル
✨ おすすめ動画
C11
346 文
Trust the Process: How to Design Teams That Actually Work
Daily ConversationUniversity StudentsWorking Adults
C11
12 文
Stroke Prevention with Intracranial Stenting
Daily ConversationUniversity StudentsWorking Adults
C12
442 文
Allergy Assessment for Nurses: True Allergy vs Sensitivity, and Safer Care with Dr. Sophia Thomas
Daily ConversationUniversity StudentsWorking Adults
C11
53 文
The Science of Early Childhood Development
Daily ConversationUniversity StudentsWorking Adults
C11
710 文
The truth about peptides, NAD, & other longevity treatments | Rachele Pojednic, Ph.D.
Daily ConversationUniversity StudentsWorking Adults
C13
31 文
Usability Heuristic 2: Match Between the System and the Real World
Daily ConversationUniversity StudentsWorking Adults
C11
689 文
Database Sharding! Designing Data-Intensive Applications chapter 6
Daily ConversationUniversity StudentsWorking Adults
C11
113 文
Architectural Design Process Explained
Daily ConversationUniversity StudentsWorking Adults
C11
60 文
Quick Study: How Venture Capitalists Make Decisions
Daily ConversationUniversity StudentsWorking Adults
C11
406 文
Scoliosis: How to medically classify scoliosis (the Lenke System for AIS)
Daily ConversationUniversity Students
C12
221 文
Masking in ADHD & Autism: The Hidden Cost of Looking “Fine”
Daily ConversationUniversity StudentsWorking Adults
B21
623 文
How evolution works, in 54 minutes | Sean B. Carroll: Full Interview
Daily ConversationUniversity StudentsWorking Adults
このレッスンの語彙とスピーキングのポイント
このスピーキングレッスンは、動画「Sequential, Multiple Assignment, Randomized Trial Designs With Dr Kidwell」を教材にしています。 繰り返し出てくる語は次のとおりです：treatment, trial, design, smart, patient。 この動画には、シャドーイング用の文が103文、単語が2153語あります。 音声の長さは13:26です。 話す速さは1分あたり約160語で、日常会話に近い自然なペースです。 単語の84%は英語の頻出3,000語に含まれます。残りは練習の前に確認しておきましょう。 全103文のうち18文が疑問文なので、疑問文のイントネーションや短い受け答えの練習に向いています。
この動画の重要語彙
動画に出てくる覚えておきたい単語15語を、発音と意味つきで紹介します。
| 単語 | 発音 | 意味 |
|---|---|---|
| sequence 名詞 | /ˈsiː.kwəns/ | 順序 |
| adapt 動詞 | /əˈdæpt/ | 適応する |
| collect 動詞 | /ˈkɑlɪkt/ | 集める, 収集する |
| intervention 名詞 | /ɪntɚˈvɛnʃən/ | 介入 |
| tailor 名詞 | /ˈteɪlɚ/ | 仕立て屋, テーラー |
| rescue 動詞 | /ˈɹɛs.kjuː/ | 救う, 助ける |
| compare 動詞 | /kəmˈpɛɚ/ | 比べる, 比較する |
| participant 名詞 | /pɑːɹˈtɪsəpənt/ | 参加者, 関与者 |
| assign 動詞 | /əˈsaɪn/ | 割り当てる |
| consent 動詞 | /kənˈsɛnt/ | 承諾する, 同意する |
| versus | /ˈvɝsəs/ | 対, バーサス |
| interaction 名詞 | /ˌɪn.təˈɹæk.ʃən/ | 相互作用 |
| depend 動詞 | /dɪˈpɛnd/ | 依存する |
| podcast 名詞 | /ˈpɑdˌkæst/ | ポッドキャスト |
| oppose 動詞 | /əˈpoʊz/ | 阻む, 阻止する |
動画に出てくる句動詞
| 単語 | 発音 | 意味 |
|---|---|---|
| go ahead 動詞 | /ˌɡoʊ əˈhɛd/ | どうぞお先に |
繰り返し練習したい文
動画に出てくる短く完結した文です。日常会話でそのまま使えます。
- Let's try to break it up a little bit.
- One is that we're interested in sequences of treatments.
- We can think about what is the best treatment to begin with?
- How do we change?
- When do we change?
この動画の文法
話し手がよく使っている文型を、動画の実際の表現とともに紹介します。
| 文型 | 動画での表現 |
|---|---|
| 受動態 be + 過去分詞 — 誰がするかより、何が起きるかに焦点を当てる | will be randomized · could be assigned · are assigned |
| 関係詞節 who / which + 節 — 人や物について情報を加える | those who respond · those who don · aim, which can |
注意したい発音
話し手はwe're, they're, I'mなど、短縮形や弱形を44回使っています。聞こえたとおりの短い形で発音しましょう。
- 「sh」と「zh」の音: intervention /ɪntɚˈvɛnʃən/, limitation /lɪmɪˈteɪʃən/, interaction /ˌɪn.təˈɹæk.ʃən/, adaptation /ˌæd.æpˈteɪ.ʃən/, contrast /kənˈt(ʃ)ɹæst/
- 長い単語（アクセントの位置に注意）: intervention /ɪntɚˈvɛnʃən/, participant /pɑːɹˈtɪsəpənt/, limitation /lɪmɪˈteɪʃən/, interaction /ˌɪn.təˈɹæk.ʃən/, subsequently /ˈsʌb.sɪ.kwənt.li/
日本語話者が苦手な音：
- /r/ と /l/ の区別 — /l/ は舌先を歯茎につけ、/r/ はどこにもつけない: practical /ˈpɹæk.tɪ.kəl/, probability /ˌpɹɑ.bəˈbɪl.ə.ti/, vocabulary /vəˈkæb.jə.lə.ɹi/
- /v/ — /b/ にならないように、上の歯を下唇に当てる: adaptive /əˈdæp.tɪv/, intervention /ɪntɚˈvɛnʃən/, versus /ˈvɝsəs/, investigator /ɪnˈvɛs.tɪˌɡeɪ̯.təː/, alternatively /ɔlˈtɝ.nə.tɪv.li/
- /f/ — 「フ」ではなく、上の歯と下唇で出す: define /dɪˈfaɪn/, inform /ɪnˈfɔɹm/, inference /ˈɪn.fə.ɹəns/, upfront /ʌpˈfɹʌnt/, quantify /ˈkwɑn.tɪˌfaɪ/
この動画での練習方法
- まず声を出さずに動画を最後まで聞き、知らない単語をメモします。
- まず0.75倍速で一文ずつシャドーイングし、慣れてきたら通常の速度に戻します。
- 自分の声を録音して元の音声と比べます。sequence, adapt, collectなどの単語に特に注意しましょう。
シャドーイングとは？英語上達に効果的な理由
シャドーイング（Shadowing）は、もともとプロの通訳者養成プログラムで開発された言語学習法で、多言語習得者として知られるDr. Alexander Arguelles によって広く普及されました。方法はシンプルですが非常に効果的：ネイティブスピーカーの英語を聞きながら、1〜2秒の遅延で声に出してすぐに繰り返す——まるで「影（shadow）」のように話者を追いかけます。文法ドリルや受動的なリスニングと異なり、シャドーイングは脳と口の筋肉が同時にリアルタイムで英語を処理・再現することを強制します。研究により、発音精度、抑揚、リズム、連音、リスニング力、そして会話の流暢さが大幅に向上することが確認されています。IELTSスピーキング対策や自然な英語コミュニケーションを目指す方に特におすすめです。