跟讀練習: Sequential, Multiple Assignment, Randomized Trial Designs With Dr Kidwell - 通過影片學習英語口語

選擇練習模式

每種模式都會開啟全屏練習室。

154 句

如果句子過短或過長，請點選 Edit 進行調整。

Hello, and welcome to JAMA Evidence, our podcast series based on core issues in evidence -based medicine.

Today, we will be discussing sequential, multiple assignment, randomized trial designs.

I'm your host, Dr. Roger Lewis, statistical editor for JAMA and co -editor of the JAMA Guide to Statistics and Methods series.

I'm here with Dr. Kelly Kidwell.

Dr. Kidwell, thank you so much for joining us.

Can you go ahead and tell our listeners a little bit about yourself?

Sure.

Thank you so much for having me, Roger.

I'm Kelly Kidwell.

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,

and in particular, the sequential multiple assignment randomized trials or SMART design.

Well, again, thank you so much for joining us.

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.

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?

I completely agree with you that the vocabulary around smart design is quite daunting.

So what an excellent first question to start with.

So sequential is about a few things.

One is that we're interested in sequences of treatments.

And in this clinical trial design, we're going to have sequential randomization.

So those terms aren't next to each other, but that's really what's happening and defines this trial,

such that at least some group of participants will be randomized twice in sequence.

So usually there's some upfront randomization

and then a subsequent randomization for some group of participants based on their response to the initial treatment.

And so that multiple assignment comes into the fact that we'll have this sequential treatment assignment.

So individuals could be assigned to multiple treatments.

And this is all wrapped up in a clinical trial.

So we're going to use this SMART clinical trial tool to give us robust evidence for sequences of treatments

that adapt to individual responses or patient status.

So in contrast, in a traditional or a simpler clinical trial design.

We just randomize once, and then the patient either does well

or perhaps doesn't do so well on the treatment to which they are assigned.

And if they don't do well, we often have some sort of rescue therapy or they have to leave the trial.

So how is the experience of a patient or a participant different in a SMART design?

Right.

So in a general, standard, traditional randomized clinical trial, we're really looking at one point in time,

sort of this trial in a vacuum.

The SMART is going to say, let's look at that patient over the course of care.

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.

And in fact, we might want to randomize the treatment to

which rescue therapy they're getting so that we can collect unbiased evidence

and use causal inference to better understand what the effect of that treatment is.

And not just that treatment, but the entire treatment sequence.

And so really for the patient,

it might not be too different than a standard trial because we're going to treat you over the course of your care.

But to the investigators, we're going to make sure that

that data is collected for our trial and there might be randomization for the treatment assignment that follows for that patient.

So, let's think a little bit about how we analyze or what we can learn from those two different approaches.

So in the first, we randomize at a single point in time, and if the patient doesn't do well,

they either drop out of the trial or we assign a rescue therapy.

Whereas in the SMART design, if they require a change in therapy, we actually randomize what we do.

So what's different about what we can learn from those two different approaches?

That's right.

Often in our standard clinical trial, in that one stage where the patient drops out or receives clinician or patient choice therapy,

we no longer make inference or have any decisions or really robust results about that subsequent treatment or their outcome.

Whereas in the SMART, by randomizing that individual and continuing to collect that data,

we will be able to have really robust evidence over the course of care, over the sequence of treatments

that could have adapted based on their response to allow us

to give these sequential guidelines to clinicians to treat the patient not just up front at one point in time,

but over time, and really tailor it to how they're doing, if they responded, if they adhered, if they engaged with that treatment.

So in a sense, this more closely follows what happens in actual clinical care,

where we make a series of treatment decisions based on the patient's prior response, but in this case, you're setting it up

so we can rigorously learn about what is the best next

thing to do based on what's happened to the patient so far.

Is that a fair summary?

Absolutely.

We are looking at a type of clinical trial that more closely mimics the way in which treatment is practiced.

And so that we can provide evidence for these treatment strategies

that adapt to individual's response or patient status in the way that we would actually treat them in practice.

Let's suppose that a SMART design is run and there's two points at which patients are randomized,

maybe the initial therapy and then after some defined period of follow -up.

What does the result look like in the publication that results,

and how can clinicians utilize the results of the SMART in guiding their own clinical care?

One of the amazing advantages of a SMART design is

that it can answer many questions as opposed to just what is the best treatment at one time.

We can think about what is the best treatment to begin with?

What is the best subsequent treatment for those who respond?

What is the best subsequent treatment for those who don't respond?

And then we can think about that entire sequence of, okay, so what do we start with?

How do we change?

When do we change?

For whom do we change and what treatment do we give them?

And so we can really get from SMARTS this sequence of treatments adapted to patient response.

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,

how they're adhering, how they're engaging, if they're having side effects,

and use that information to better tailor the treatment assignment to that individual so that after the trial, right, we have all this information.

And we can look at what are the best sequences of treatments for which particular individuals.

The fact that the SMART design more closely matches actual clinical care,

which the sequences of therapies we offer patients is informed by their responses to prior therapies is an obvious advantage.

But what are some of the limitations or challenges in either the implementation or the interpretation of a SMART clinical trial?

I think there are a few limitations and also just lack of knowledge of this design and its opportunities that it offers.

Some of the limitations are that individuals think, oh, now that we have multiple randomizations or multiple assignments in sequence,

this is now logistically infeasible, or you're going to lose more people if you have to consent multiple times.

So one thing is that we often just consent once at the beginning of the smart design.

and we may randomize multiple times.

Alternatively, you could randomize up front to that sequence that they may follow, whether they respond or do not respond.

So logistically, this doesn't have to be very difficult.

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.

Number two is that individuals are worried about bias coming into play from a smart design

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.

下載應用

流利口語所需的一切

AI 發音評分逐句打分

音標練習掌握每個音

詞彙學習建立你的詞庫

詞彙遊戲邊玩邊學

掃碼

Download on theApp Store Get it onGoogle Play

同頻道

Balloon Angioplasty for Intracranial Atherosclerotic Stenosis

IELTS PrepTOEFL PrepUniversity Students

FOR YOU

本課的詞彙與口語要點

這節口語課以影片“Sequential, Multiple Assignment, Randomized Trial Designs With Dr Kidwell”為素材。影片中反覆出現的詞有：treatment, trial, design, smart, patient。這段影片共有 103 個句子、2153 個單詞可供跟讀。講話部分時長為 13:26。說話人語速自然，每分鐘約 160 個詞，接近日常對話。 84% 的單詞屬於英語最常用的 3,000 詞，其餘的詞建議在練習前先學一下。全部 103 個句子中有 18 個是問句，適合練習疑問語調和簡短回答。

影片中的重點詞彙

影片中 15 個值得學習的單詞，附發音和釋義：

單詞	發音	釋義
sequence 名詞	/ˈsiː.kwəns/	序列, 順序 /順序
assignment 名詞	/əˈsaɪn.mənt/	分配, 佈置
subsequent 形容詞	/ˈsʌbsɪkwənt/	隨後的 /隨後的
adapt 動詞	/əˈdæpt/	適應 /適應
collect 動詞	/ˈkɑlɪkt/	庋藏, 收藏
intervention 名詞	/ɪntɚˈvɛnʃən/	介入, 幹預
tailor 名詞	/ˈteɪlɚ/	裁縫 /裁縫, 裁縫師傅 /裁縫師傅
rescue 動詞	/ˈɹɛs.kjuː/	拯救, 救援
compare 動詞	/kəmˈpɛɚ/	比, 比較 /比較
participant 名詞	/pɑːɹˈtɪsəpənt/	參加者 /參加者, 參與者 /參與者
limitation 名詞	/lɪmɪˈteɪʃən/	限制
assign 動詞	/əˈsaɪn/	分配, 分派
whereas	/(h)wɛɹˈæz/	然而
consent 動詞	/kənˈsɛnt/	承諾 /承諾, 同意
versus	/ˈvɝsəs/	對 /對

影片中出現的短語動詞

單詞	發音	釋義
go ahead 動詞	/ˌɡoʊ əˈhɛd/	請便 /請便
figure out 動詞		弄清楚
follow up 動詞		跟進 /跟進

值得反覆跟讀的句子

影片中簡短而完整的句子，可以直接用在日常對話裡：

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?

影片中的語法

說話人最常用的結構，並附上影片中的原話：

結構	影片中的用法
被動語態 be + 過去分詞 — 強調發生了什麼，而不是誰做的	will be randomized · could be assigned · are assigned
定語從句 who / which + 從句 — 補充說明人或事物	those who respond · those who don · aim, which can

需要注意的發音

說話人用了 44 次縮略和弱讀形式，例如 we're, they're, I'm。請按聽到的簡短形式來說。

“sh” 和 “zh” 音: intervention /ɪntɚˈvɛnʃən/, limitation /lɪmɪˈteɪʃən/, interaction /ˌɪn.təˈɹæk.ʃən/, adaptation /ˌæd.æpˈteɪ.ʃən/, contrast /kənˈt(ʃ)ɹæst/
長單詞——注意重音位置: intervention /ɪntɚˈvɛnʃən/, participant /pɑːɹˈtɪsəpənt/, limitation /lɪmɪˈteɪʃən/, interaction /ˌɪn.təˈɹæk.ʃən/, subsequently /ˈsʌb.sɪ.kwənt.li/

中文母語者容易讀錯的音：

/v/ — 不要讀成 /w/，上齒輕觸下唇: adaptive /əˈdæp.tɪv/, intervention /ɪntɚˈvɛnʃən/, versus /ˈvɝsəs/, investigator /ɪnˈvɛs.tɪˌɡeɪ̯.təː/, alternatively /ɔlˈtɝ.nə.tɪv.li/
詞尾子音 — 要讀清楚，後面不要加母音: sequence /ˈsiː.kwəns/, assignment /əˈsaɪn.mənt/, subsequent /ˈsʌbsɪkwənt/, adaptive /əˈdæp.tɪv/, adapt /əˈdæpt/
/r/ 和 /l/ 的區別: practical /ˈpɹæk.tɪ.kəl/, probability /ˌpɹɑ.bəˈbɪl.ə.ti/, vocabulary /vəˈkæb.jə.lə.ɹi/

如何用這段影片練習

先完整聽一遍影片，不要開口，記下不認識的單詞。
先用 0.75 倍速逐句跟讀，熟練之後再回到正常速度。
錄下自己的聲音並與原聲對比，特別注意 sequence, assignment, subsequent 這類單詞。

什麼是跟讀法？

跟讀法 (Shadowing) 是一種有科學依據的語言學習技巧，最初開發用於專業口譯員的培訓，並由多語言者Alexander Arguelles博士普及。這個方法簡單而強大：您在聽英語母語原聲的同時立即大聲重複——就像是一個延遲1-2秒緊跟說話者的影子。與被動聽力或語法練習不同，跟讀法強迫您的大腦和口腔肌肉同時處理並模模擬實的講話模式。研究表明它能顯著提高發音準確性，語調，節奏，連讀，聽力理解和口語流利度——使其成為雅思口語備考和真實英語交流最有效的方法之一。

影子跟讀法: 閱讀完整分步指南 →

跟讀練習: Sequential, Multiple Assignment, Randomized Trial Designs With Dr Kidwell - 通過影片學習英語口語

選擇練習模式

下載應用

同頻道

Balloon Angioplasty for Intracranial Atherosclerotic Stenosis

推薦影片

Trust the Process: How to Design Teams That Actually Work

Stroke Prevention with Intracranial Stenting

Allergy Assessment for Nurses: True Allergy vs Sensitivity, and Safer Care with Dr. Sophia Thomas

Inside a Biotech Company: Roles, Teams & Where You Could Fit In | EP 14

Database Sharding! Designing Data-Intensive Applications chapter 6

The medicine myth | Denis Noble, Nessa Carey, Guy Brown

Mental Debugging: How to Build Resilience After Failure

Quick Study: How Venture Capitalists Make Decisions

Design Patterns in Java Theory

Scoliosis: How to medically classify scoliosis (the Lenke System for AIS)

Explanation of Vaccine Adjuvants

Masking in ADHD & Autism: The Hidden Cost of Looking “Fine”