Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:10
Hello, and welcome back to
0:12
Talk Evidence. We've got Nick
0:14
writing program for you this week, we start with some
0:16
new data on colorectal cancer screening
0:19
using colonoscopy.
0:21
We then turn very excitingly to
0:23
listener request. You've been itching
0:26
for someone to send in their evidence
0:28
request. And finally, we have
0:30
one, so you can hear more about
0:32
that. And in fact, it tends to a bit of feature
0:34
then. This this episode
0:37
feels very dedicated to diabetes
0:39
and and of prediabetes, and
0:41
there's not a bit of COVID insight. And
0:43
we're going to finish off with a little bit of news
0:46
about clinical academics behaving
0:48
badly.
0:50
I'm
0:50
Helen McDonald's, research integrity editor
0:53
for the BMJ and BMJ journals, and
0:55
I'm joined today by Joan.
0:57
Hi, Joan. Do you want to introduce yourself?
0:59
Hi, everyone. This is Joe Ross. I'm
1:01
a professor of medicine in public health
1:03
at Yale and also an associate editor at the
1:05
BMJ.
1:06
I sounded very timid to that, Joe. And
1:08
I'm also joined by Juan. Juan,
1:11
do you guys introduce yourself?
1:12
Hi, everybody. I'm Juan Franco.
1:14
I'm a firm position. I'm
1:16
currently a researcher at the Heineken
1:19
University in Dusseldorf Germany, and
1:21
I'm the Attorney Chief of BMC Evidence
1:23
by medicine.
1:26
In short, a very important people.
1:34
Right. Let's get on to our first
1:36
item for today, which is colorectal cancer
1:39
screening and Juan, you
1:41
volunteered to take a closer look at this one.
1:43
This is a new study that came out in the New
1:45
England Journal of Medicine. A
1:47
few days. Okay?
1:49
Wait
1:49
a second, Helen, isn't this the B and J
1:51
Talk Evidence podcast? Why are we talking about
1:53
a paper in the England Journal of Medicine?
1:55
I'm gonna see if we can bring it back to
1:57
the MJ at the end. But we're we're open to
1:59
all evidence here. It's
2:02
evidence motivations not journals.
2:04
So Juan, tell us about this paper.
2:06
Well, I think it's it
2:08
was very popular in the
2:10
news this past
2:13
days because there is
2:15
one of the first randomized
2:18
control trials that put colonoscopy
2:21
in the under
2:23
scrutiny. And It
2:25
was a a large trial between
2:27
two thousand and nine and twenty fourteen in Poland,
2:30
Norway, Sweden, and Netherlands with
2:34
follow-up data from more than
2:36
eighty four thousand participants. And
2:41
and pay participants were randomized
2:43
to be invited to
2:45
undergo a single screening colonoscopy
2:48
or no invitation to screening.
2:51
So the
2:52
And that will be an important point later, why
2:54
don't you plan? because they were invited to
2:57
come. You didn't hear anyone's gentle
2:59
gentle voice, but I will stress that. So it's not
3:01
necessarily that they actually came. They were invited
3:04
to come. Carry on.
3:04
No one had to describe my voice as gentle
3:07
before, but thank you. I
3:11
yes. And the
3:14
the invitation part is is
3:16
crucial not only because I couldn't
3:18
imagine a trial of people being coerced to
3:20
having colonoscopy, but also
3:22
by the fact that only forty two percent
3:24
of those invited
3:27
under with screening. So very
3:30
little well, like,
3:33
I wouldn't want to put it as subjective
3:37
tone to it, but let's say it's not
3:39
an optimal uptake for
3:41
something that you're randomizing people into.
3:43
And during
3:46
the medium follow-up of ten years, Basically,
3:49
in the main analysis, the
3:54
risk of colorectal cancer
3:56
was point ninety eight
3:58
percent in the invited group and point one
4:00
two in the usual care group. So
4:02
there are say some absolute
4:05
risk reduction and the confidence interval
4:08
is between zero point seven and
4:10
zero point eight three. and
4:12
– but the risk of death
4:14
from colorectal cancer was
4:18
zero point two eight in the divided group
4:20
and zero point thirty one in the
4:22
usual care group and the confidence
4:24
interval for the risk ratio is from
4:26
zero point sixty four to one point
4:28
sixteen. And
4:31
so this is the main analysis.
4:33
And the interpretation is basically
4:35
that And
4:37
it's as consistent with these
4:39
types of screening that identify
4:43
precursors of the disease, which
4:45
is also common in cervical cancer,
4:48
so you remove the polyps, and
4:50
you therefore, you estimate that the
4:52
risk of the disease will be lower.
4:55
and this is what we see in one of the
4:57
outcomes. But unfortunately, we
4:59
can't really identify a
5:02
change in colorectal cancer
5:04
mortality. Most
5:07
of this could be explained by the
5:09
low uptake would reduce the
5:11
power of the study to
5:14
detect this
5:16
difference. And as a matter of fact,
5:18
the authors go into
5:21
an adjusted a per protocol analysis,
5:23
trying to see those people
5:26
who would have undergone screening
5:29
And in that case, they
5:31
identify a difference
5:34
in colorectal cancer mortality. with
5:38
zero point one five percent in the invited group and
5:40
zero point three percent in the usual
5:42
care group, which is an - of
5:44
an estimate of half of
5:46
did that form
5:48
colorectal cancer? So
5:50
you're trying to tease out the does
5:52
the screening program work and what the
5:54
effects are versus if you actually turn up
5:56
for this screening, does
5:58
it does it work then?
5:59
the
6:00
Joe, I can see what it to say something.
6:02
Yeah. I mean, Juan, it was great that you
6:05
brought this to our podcast discussion
6:07
this week because it really was a
6:09
heated discussion throughout Twitter and
6:11
other social media forums. There's a bit of a,
6:13
you know, a roshock test
6:15
of of sentiment. It has it gone down,
6:17
Jay. Well, at that's what I mean. Like, it was
6:19
a bit of a Wireshark in the sense that, like,
6:21
people saw in it what they wanted to see in it. There
6:23
were the people who, you know, saw
6:25
the you know, the low participation
6:27
rate and that means that
6:29
screening programs, you know, aren't likely to
6:31
work. You know, we knew that, you know, we're probably
6:34
there's too much over diagnosis, too much
6:36
testing and treatment. This is just
6:38
one more kind of nail on the coffin for screening
6:40
programs. Others people were like, hey,
6:42
now, hey, now, like, it's more complicated than that.
6:44
And, you know, when you do do the intention
6:47
to treat or, sorry, the
6:49
per protocol analysis focused on the
6:51
people who actually underwent
6:53
screening. Right? It actually it's
6:55
it's effective at the individual level.
6:58
And to me, this raises what I think is an interesting
7:00
question. So My understanding is that in
7:02
Europe, for the most part, people don't
7:04
do colonoscopy unless they have
7:06
some symptoms of disease. But in the United
7:08
States, the sort of,
7:10
like, the ground level for colorectal
7:12
cancer screening is colonoscopy. And
7:14
I wonder if the participation rates would
7:16
have been much higher in a population of
7:19
people who are accustomed or like
7:21
prepared at age fifty or now at age
7:23
forty five, I'm supposed to undergo a
7:25
colonoscopy. And if we had seen participation
7:27
rates in the eighty percent as opposed to
7:29
the forty percent, might the results of
7:31
the trial look different?
7:33
Well, that's very interesting. And I did say that
7:36
I was going to bring it back to the BMJ didn't
7:38
I? Because a while ago, a couple of
7:40
years ago, we published
7:42
a paper in a rapid recommendation
7:45
series that looked at screening
7:49
options for colorectal cancer. And
7:51
Joe, you went back to look at that
7:53
paper to see
7:55
how to see how this
7:58
latest study might impact on those recommendations.
7:59
Yeah.
8:01
Well, so first I would just say for listeners,
8:03
you know, we'll put a a link in the show notes
8:05
to the rapid recommendations page
8:10
in the BMJ And I would encourage people to go
8:12
and look at it because the team at
8:14
the you know, at the journal do a really
8:16
terrific job of creating infographics that
8:18
people can engage with in order to look
8:20
through and sort of identify, you know, what
8:22
pathway to go down. And given the time and
8:24
effort they put into it, more people should use it.
8:27
But essentially, what you see,
8:29
right, is you know, that
8:31
when the fifteen year risk of colorectal
8:34
cancer for an individual is
8:36
below three percent there's,
8:38
you know, weak evidence
8:41
that favors no screening. But
8:43
when you when there that fifteen
8:45
year risk is above three percent there's a weak evidence
8:47
that favors screening. And of course,
8:49
that three percent risk is
8:52
estimated using the acute cancer calculator
8:54
that's also right on the site for
8:56
people to look at. Now, how
8:58
is this trial going to change it?
9:00
Well, it will probably move it
9:02
even further away in the sense
9:04
that it
9:05
one will
9:06
disfavor
9:07
colonoscopy and screening
9:10
even a little bit more.
9:12
But there are more trials
9:14
coming and I hope that
9:16
the next few years will have a much better
9:18
sense of the role of colonoscopy
9:21
in colorectal cancer screening. I don't know, Helen,
9:23
if you to talk about Well,
9:24
what I can say is that
9:26
the that sounds like a politician doesn't
9:28
know. What I can say is
9:30
that the team that produce the
9:33
rapid recommendations are going
9:35
to have a look at that
9:37
study in more detail and add
9:40
an update to
9:42
describe the the latest
9:44
evidence and to give some preliminary
9:46
indication of how things might change. But I think
9:48
they might hold on a bit longer for
9:50
those other studies that you mentioned, Joe, to
9:52
do a fuller update of
9:54
that guidance in light of that
9:56
new evidence there.
10:06
Now, let's turn diabetes or prediabetes,
10:08
I should say, and our very exciting
10:10
listener request. So
10:13
here in the KB NHS
10:15
spends about ten billion a year
10:17
on diabetes, apparently. I love this reading around,
10:19
and that's about ten percent of the entire
10:21
budget. And
10:24
our query relates to prediabetes.
10:26
And here it is. It's from
10:29
a UK GP who happens to
10:31
work in Bath, which is actually where I'm
10:33
based. So here we are.
10:37
Hi.
10:37
My name is Nicole House, and I'm a GP
10:39
in Bath. I would like more information
10:41
on the evidence for the National Diabetes
10:44
Prevention Program to discuss with my
10:46
patients living with obesity. We've
10:48
had local education Sally Day
10:50
where the team presented their
10:52
program but with no clinical outcome
10:54
data. Can the talk
10:56
evidence team tell me what is
10:58
the evidence these programs prevent,
11:01
reduce or delay the onset
11:03
of diabetes. Many
11:05
thanks.
11:09
So that was Nicole's message. For
11:12
those people who are not listening in
11:15
the UK or in England, this
11:17
is a program
11:19
that's offered through the NHS.
11:22
It's a mixture of some
11:24
coaching
11:25
information. It's delivered digitally.
11:27
It's centered around education.
11:30
An educational intervention and to try
11:32
and prevent or delay people from developing
11:34
type two diabetes, and it focuses on
11:36
healthy eating, lifestyle,
11:38
help to lose weight and physical
11:41
exercise programs. And the
11:43
program both says that these have been
11:45
proven to reduce the risk of
11:47
develop up in type two diabetes. And
11:50
so what I've done when I
11:52
gave Joe and Juan the homework
11:54
on this is thought of kind
11:56
of two questions One is
11:58
what is the evidence
11:59
that programs a
12:00
bit like this work, so sort of
12:03
lifestyle and event to prevent
12:05
people from going from having
12:07
pre diabetes to diabetes. And then
12:09
we'll come back to this specific one.
12:12
that's run-in England. So
12:14
Juan, with your connections
12:16
to Cochrane and systematic reviews,
12:18
I thought you would be the perfect person
12:20
to tellers first, where the program's a
12:22
bit like this work?
12:25
Well, yes, there's
12:27
a Cocon review from
12:29
twenty seventeen
12:31
about diet, physical
12:34
activities are a combination
12:36
of both in different
12:38
programs to prevent or delayed
12:40
type two diabetes. They
12:43
included different
12:45
studies. And especially, it's it's
12:47
it's important to have, like, a different time
12:49
points with different different credentials
12:52
of diabetes. But the
12:54
main gross evidence
12:56
from eleven trials highlight
12:59
that this these
13:01
programs may reduce slightly
13:03
some of the risk of
13:05
diabetes. As particularly
13:07
in numbers, there's moderate
13:09
sense, you have evidence of every
13:13
for a thousand people, that
13:15
do not go out this program with
13:19
a two hundred and fifty seven will
13:21
get diabetes. And with
13:23
the program, one
13:24
hundred and forty six. So
13:28
a reduction that, of course, has
13:31
a pretty high confidence interval.
13:34
And what kind of time
13:34
period is that over one? Is that sort of
13:37
The
13:37
median the mean duration follow-up is
13:40
three point eight years. Okay. So
13:42
we're looking more at this sort of delaying
13:44
rather than Life
13:45
time prevention of
13:47
disease. It's difficult because most of
13:50
this study is not had
13:52
very long follow-up time.
13:54
And most of all, most
13:57
of this study is did
13:58
not
13:59
provide enough data
14:02
for other type of
14:04
important outcomes. For example, complications
14:06
related to diabetes or health quality
14:08
of life. or
14:11
possibly adverse events, which
14:13
one would expect that there would
14:15
be minor But for example, even if you take
14:17
cardiovascular mortality, that
14:20
seven studies did report on cardiovascular
14:22
mortality, the evidence
14:25
was too uncertain to
14:29
to to to figure out whether there's a
14:31
difference or not between those
14:33
receiving receiving the
14:35
program or not. So, basically,
14:39
one of the questions perhaps,
14:42
I see, I have as a
14:44
clinician thinking about this, is
14:46
how much of these programs are trying
14:48
to prevent like the chemical
14:50
diagnosis of diabetes or how much
14:52
are we preventing the actual morbidity
14:54
that is related to
14:56
diabetes and the burden of the disease?
14:58
So
15:00
yeah.
15:01
I I would I I
15:04
think that there is some is
15:06
that is positive one.
15:09
But I I mean more.
15:10
Well, it's it's really interesting. I mean, I
15:13
think much of these programs that
15:15
while, you know, that cochrane review
15:17
summarizes a wide breadth of
15:19
evidence, but the seminal
15:22
study that was actually
15:24
published a long time ago in the the Journal, I
15:26
think two thousand two, two
15:28
thousand three, something like that.
15:30
You know, that was the first three year follow-up,
15:32
you know, in the diabetes prevention program.
15:34
I identifying folks at high risk, those with
15:36
pre diabetes, you know, as
15:38
assessed by elevated
15:40
fasting and post load plasma
15:43
glucose concentrations. So they were randomized
15:45
to placebo metformin
15:47
or lifestyle interventions. And the people who were
15:49
randomized to lifestyle interventions did better
15:51
even than those who started taking metformin
15:54
in terms of the incidence
15:56
of diabetes and with weight loss.
15:59
that Lancet published their ten year outcome rates
16:01
in the lower incidence
16:04
of diabetes persisted. But
16:06
again, they don't look at cardiovascular
16:09
outcomes like WAM Notes. But it it
16:11
was those studies that actually
16:13
led in the United States
16:15
a big program evaluation led
16:18
by Medicare, the Centers for Medicare
16:20
Medicaid Services. So Medicare did this big pilot
16:22
demonstration project essentially
16:24
say, well, everyone, you know, should
16:26
have access to lifestyle
16:28
interventions as part of, you know,
16:30
diabetes prevention. Among those, quote
16:32
unquote, with prediabetes. Right?
16:34
And that's actually been a large
16:37
rollout that I think is now
16:39
mirrored in terms of what's happening in
16:41
the UK. And the
16:43
evaluation of that effort
16:45
is gonna be like what happened with the
16:47
colonoscopy trial we just discussed. So
16:49
they're gonna offer lifestyle, interventions,
16:52
and or metformin. I'm not sure if
16:54
that's even part of the program. And then it'll
16:56
depend on participation rates to know if
16:58
it actually works at the population
17:00
level as opposed to an randomized trial.
17:02
And I
17:02
did find a couple of papers, which will
17:04
put in the notes of the episode,
17:07
which were largely service evaluations
17:09
and people's experiences of having
17:11
been in the program. So I think
17:13
the answer to the precise answer in
17:16
Nicole's question is I don't
17:18
think we entirely know
17:20
the outcomes for this
17:23
program or necessarily the
17:26
answers about for how
17:28
long it might delay diabetes
17:31
or ultimately whether it fully prevents it. But I
17:33
think Joe and
17:35
Juan, I feel
17:35
I feel persuaded that
17:38
there is at least some evidence
17:40
of these programs having an
17:44
effect,
17:45
certainly feels like they're unlikely to
17:48
do harm when targeted at people
17:50
who
17:51
have pre diabetes.
17:53
And and, also, in general, right, they're
17:55
focused on lifestyle modifications
17:58
-- Yes. -- which are general -- I'm saying
17:59
wow. -- not just decided,
18:02
but for lots of lots of
18:04
reasons. So, yes, I think, you know,
18:06
feel pretty good about that. Well, we can
18:08
look forward to Nicole's feedback maybe.
18:11
on whether we have
18:13
helped her remotely have a conversation
18:15
with her patients. You
18:16
know, perhaps I would add
18:19
is how this affect the whole system
18:21
in the sense.
18:22
When you shift the
18:25
resources and time and
18:27
efforts into prevention, I would have
18:29
to be absolutely sure that the
18:31
people with the condition in this case with
18:33
diabetes are receiving the
18:35
most option care because if and
18:37
so thinking about the whole system,
18:39
if you if you're a policy communicator
18:41
have to allocate money, where
18:44
would you put it in this case?
18:46
So there's some
18:48
let's look a bit of
18:50
an equity situation. Right? So
18:52
it depends on the system. Since I'm not in the NHS,
18:54
I cannot say that.
18:56
So but I I would argue that
18:58
in other systems, which people with
19:00
with diabetes are struggling to
19:03
get their medication or
19:05
or have a strong pathway
19:08
of care, allocating
19:10
resources to the prevention of the
19:12
condition may seem slightly
19:15
You are
19:15
really confusing everyone's CPT
19:18
now, hon.
19:27
So
19:27
coming up in December, we've got BMW
19:29
Research Forum at BMA House. And
19:31
joining me is dot to Helen Serrano from the events team.
19:33
Hi, Helen. Hi, Helen. Helen's
19:38
meet. So tell us a bit about this
19:40
event. You're speaking. So
19:42
the event is to bring together
19:44
the research community. We
19:47
recognize that there's loads of people
19:49
involved in research, not just
19:51
the people who submit to us and the people
19:53
who edit the journals here at B and J.
19:56
But there's funders, there's policymakers,
19:59
patient experts,
19:59
But of course, the researchers are the main part
20:02
of that. And we're just having this
20:04
event alongside the B and J
20:06
editors meeting to make
20:08
sure all those people are talking with each other
20:10
about the same things rather than
20:13
just to their own groups. Some of the
20:15
speakers we've got are
20:17
from some of the big funding organizations.
20:19
We've got Ken Gabriel
20:21
from Welcome Leap, who
20:24
is talking about how they fund for
20:26
impact by using big amounts
20:29
of venture capital money
20:31
alongside investment from organizations like
20:33
welcome. We've got
20:35
Chris Witty, the Chief Medical Officer, talking
20:37
about how research happens in
20:39
the UK and how governments involved
20:42
in it. And we've got a
20:44
really interesting panel
20:46
from
20:46
from
20:47
researchers and funders and
20:50
patients talking about Parkinson's disease
20:52
research and how that is becoming
20:54
a real spectrum and
20:56
a virtuous circle of patients
20:59
getting involved with research
21:01
questions and funders looking
21:03
at impact as well as
21:05
basic science research and further
21:09
clinical
21:09
research as well. So there's
21:11
loads, there's something for everyone
21:13
involved in research, and
21:15
there's some detailed workshops
21:16
and sessions along
21:19
different lines for different groups
21:21
as well.
21:23
Well, I will certainly be there and you've got
21:25
a special offer I understand for
21:28
our loyal listeners of
21:31
Talk Evidence we have. Well, we know that
21:33
the talk evidence community
21:35
are very interested
21:37
in how research gets done. So
21:39
we really hope that some of you will be able to
21:41
join us. And we've got a twenty
21:43
percent discount code, which
21:45
you can use when you
21:48
register at the website. The website
21:50
is BMJ research forum dot BMJ
21:52
dot com. That's all one word. And the discount
21:55
code is BMJRF
21:58
twenty two. So
21:59
that's BMJ research forum is the thing
22:02
you need to remember. And from that, you'll be able to
22:04
remember the website and the
22:06
discount code. so we really hope to see loads of your listeners
22:08
there. Well, thanks for
22:09
dropping and telling us all about it, Helen. Brilliant.
22:11
Thanks for having
22:12
me.
22:18
We are going to stay with
22:20
diabetes for
22:21
our next item. and
22:24
Joe, we discovered that you are a
22:26
co author of a paper that
22:28
has recently been published
22:30
and with your big
22:33
data obsession, you
22:35
have tried to emulate
22:37
a trial looking at the effectiveness
22:39
of adding different
22:42
second line drugs to
22:45
metformin, which is often use first line
22:47
for type two diabetes
22:49
because that's not enough for
22:52
many people. Tell us about your paper and
22:54
what you did.
22:56
Yeah. So
22:56
and thanks for giving the chance to talk a
22:58
little bit about it. So among
23:01
the many hats I wear,
23:03
one is I lead a research
23:06
group at Yale that
23:08
jointly works with folks at the Mayo
23:10
Clinic to collaborate with the FDA
23:12
on regulatory science research projects
23:14
that are a priority for the agency
23:16
And as you both are
23:19
aware, there's a
23:20
the
23:22
deep interest in
23:25
leveraging real world data sources
23:27
to try to better understand product,
23:29
medical product, safety and
23:32
effectiveness, particularly once products are approved
23:34
for use. And so how can
23:36
we, you know, with
23:38
the better data that are available to
23:41
us today, you know, advance our thinking and
23:43
and generate better evidence. So this
23:45
was actually part of a project
23:47
that we collaborated with with
23:49
folks in office of
23:51
medical policy to try to
23:53
understand, you know, essentially, lots
23:55
of people have looked back and said,
23:57
alright, these clinical trials have been done. Could
23:59
we have gotten the same results if we
24:01
had used observational data.
24:03
Right? And and confirmed essentially,
24:05
you know, that when you apply the same inclusion
24:07
and exclusion criteria, you you get
24:10
basically the same findings for the most
24:12
part. Right? This and that's been shown over
24:14
and over John Ionidas did a great
24:16
systematic review of all of the studies that have done
24:18
that in they they they pretty much match up, but
24:20
there's always this interesting risk of
24:22
bias when you're looking backwards as opposed to
24:24
forwards. And so Working
24:26
with the FDA, we actually identified two
24:29
ongoing clinical trials
24:31
that were kind of relevant to
24:33
the agency where we asked,
24:35
well, could we emulate
24:37
that trial with observational data
24:39
sources? And we wondered whether our
24:41
results would look similar.
24:43
And so we did one trial looking at cardiovascular
24:45
safety outcomes for a
24:47
prostate cancer treatment. And then we did
24:49
this trial, which emulated
24:52
what was considered to be kind
24:55
of the landmark grade
24:57
study. This was a study that
24:59
was launched way back in
25:01
twenty thirteen that asked the question,
25:03
okay, you have a patient with
25:05
diabetes. They're already prescribed
25:07
metformin. What do you give
25:09
them as a second line
25:11
agent? and they tested
25:13
four different medications
25:17
representing four different class
25:19
classes. I think tell
25:20
us the classes, Joe, because I feel like
25:22
the drugs vary in the different places, but
25:24
tell us the classes.
25:25
Yeah. So the
25:27
a short acting insulin or sorry.
25:29
A a
25:32
m they
25:35
tested insulin, insulin glargine,
25:37
the synthetic insulin, they tested
25:39
a sulfonylurea, in this case, glumab
25:41
glumab peride. They tested
25:45
GLP-one receptor, liraglutide,
25:47
and they tested
25:49
a DPP-four cidaglyptan. Now
25:53
you guys
25:54
see patients all the time.
25:56
What's missing? I tell you from that.
25:59
No.
25:59
Well, Juan, I'm
26:02
sure you quickly realize they're
26:04
missing the -- The t's. -- the new this
26:07
sGLT2 inhibitors, which are like the
26:09
actual diabetes medication
26:11
that now is often --
26:13
most often used for second line treatment
26:15
because it's -- there's been good evidence that
26:18
demonstrates not only does it work for glucose
26:20
lowering for patients with diabetes, but it prevents
26:22
cardiovascular risk, heart failure risk, and other renal
26:24
outcomes. And so you know,
26:26
this is the madness of clinical
26:28
trial design. Right? Way back, you know,
26:30
over ten years ago, they thought they
26:32
identified the four best sort of second
26:34
line agents and it
26:36
turns out they were wrong. Right? And
26:38
they missed one. So
26:40
what we did is we we those
26:43
results were actually published
26:45
very recently. But before those results were
26:47
published, we – our study
26:49
was published, which looked
26:51
at testing these these
26:53
four these four different medication classes using,
26:56
you know, essentially health insurance
26:59
claims data that were linked
27:01
to electronic health
27:03
work data. So we could use the lab
27:05
values to identify the
27:08
sort of main outcome, which
27:10
is of the time to a hemoglobin A1C
27:12
greater than seven seven
27:14
percent. Oh, no. You want to jame down?
27:18
Right.
27:18
right So you want
27:21
the
27:21
- essentially, you start with
27:23
a patient. It's very complicated the
27:25
way they designed it, but they essentially said
27:27
they looked for failures of treatment. Okay.
27:30
Yeah. It it so we've tried to
27:32
emulate exactly We did wanna make this easy
27:34
for anyone to understand. You wanted
27:37
to tell bus drug
27:39
that was most least
27:41
likely to cause you to fail. Is that
27:43
is that right? Right. because because you
27:45
want it because essentially what they tested
27:47
is which drug or drug class was
27:49
most effective at maintaining a
27:52
controlled hemoglobin -- Mhmm. -- agency level. Okay.
27:54
It's making a good service now. Yeah.
27:56
Yeah. But it's all everything is complicated
27:58
and and all you know, everything that we did
28:00
was also complicated. I'll try to just keep
28:02
it simple by focusing it on
28:04
two points. the first of which
28:06
was we could not
28:08
actually test the insulin glargine
28:10
arm because when you're doing this
28:12
using observational data sources, your
28:14
data reflect real world practice. And
28:16
as it turns out, no one
28:18
was prescribing insulin engardene as a second
28:20
line agent. Everyone basically got
28:23
some other, you
28:25
know, oral therapy, you know,
28:27
one of these other therapies So we just
28:29
didn't even have enough patients even with
28:31
a massive data set
28:33
to emulate that arm of the
28:36
trial. The
28:36
second thing is, in
28:38
the real world, people don't have their
28:40
A1C measured every three months like
28:42
they do in a trial. So that was
28:45
super complicated to operationalize in
28:47
terms of like, how many A1C
28:49
measures did you have to have in order for
28:51
us to look at you? And is that biasing
28:53
the sample? and all sorts of stuff like that.
28:55
But essentially, we what
28:57
we found was that liraglutide was more
29:00
effective at maintaining glycemic
29:02
control than both glimepyride or sotaglyptin.
29:04
And now we're actually in
29:06
the process of comparing our results to
29:08
the actual trial results, which just came
29:11
out. And we're actually
29:13
gonna extend our emulation to include
29:15
the SGLT2 inhibitors because if we
29:17
feel like, you know, what we did match
29:20
the actual grade study, then we can feel pretty confident in what we
29:22
find with the SGLT2 inhibitors. And
29:24
we hope that this evidence is going to be
29:26
useful. But at the end of the day,
29:28
This is these types of exercises I think
29:31
are kind of what's going to help
29:33
FDA and other regulators in
29:35
terms of like how it use
29:37
these types of data, these types
29:39
of evidence to complement, you know, what we
29:41
get from the clinical trials.
29:43
Do you
29:43
think you can ever use them if you don't
29:46
have clinical trials? because it feels a
29:48
bit like almost you
29:50
need quite
29:52
a good idea of what
29:54
you're gonna do in the trial to
29:57
design this
29:59
the It
30:00
is very complicated because
30:03
how, for instance, do you
30:05
emulate a placebo arm
30:07
right, people who have the disease but don't
30:09
get treated. Like, there's there is
30:12
bias happening there among the, you know, in
30:14
terms of using that as a
30:16
population, as control arm. That's
30:18
very hard to do. If you're comparing
30:20
two active treatments, it's a little
30:22
bit easier because you can do kind of a new incident
30:25
user study with big data
30:27
sets. Right? But it's
30:28
very difficult to just say,
30:31
these these types of data analysis
30:33
are gonna replace a clinical trial. It's
30:35
very hard to emulate
30:38
that opportunity to randomize a
30:40
person and truly get
30:42
a more unbiased the the most
30:44
unbiased estimate of the treatment
30:46
effect. So I am very reluctant
30:48
to think that we'll get there. But
30:50
if you have, you know, datasets
30:52
from nine different countries representing
30:54
nine different populations and
30:57
you examine it and
30:59
pretty consistently find that drug
31:01
a were better than drug b, I think pretty close to the
31:03
truth. We may not have the exact estimate of the
31:05
fact that we get from a clinical trial. So you would
31:07
see kind of the dream that you have the
31:08
trials and then
31:10
these big data things. They can confirm
31:13
things. They can maybe look
31:15
at
31:15
some subgroups.
31:18
Some populations who are representative?
31:20
Yeah. Divested population. So all the
31:22
people all the people. Exactly.
31:25
Uh-huh. I see how you're working here. And
31:27
of course, safety. Right? because clinical
31:29
trials are generally too
31:31
small to get real signals of safety if
31:33
they exist. So -- So how do you
31:35
think? -- the FDA decided what
31:37
they they favor? Oh, as if
31:39
they would say it bold and outright and be
31:42
very clear in their intentions, that's
31:44
not how federal agencies is worth And
31:47
what does what professor
31:49
Ross, what are you pro prescribing in
31:51
your colleague, Alan? Oh, I
31:53
generally You know, I
31:55
I think that the SGLT2 inhibitors,
31:58
not because of the work that we're done, but
31:59
because of the basis of the clinical trial evidence
32:02
tends to have the most rigorous
32:04
evidence behind it in terms of the
32:06
benefit to patients. So I think
32:09
Those are the ones I would prescribe as second
32:11
line. I don't know one, if you agree.
32:13
But to me, there's there's more
32:15
trial evidence that supports their
32:17
Juan, he's bringing you in hand. Do you agree? Yeah.
32:19
I mean, the the
32:22
SGL to
32:24
inhibitors were they
32:28
they came later in my
32:30
practice in that's
32:32
a few years. And for for the country,
32:34
it's working. It was very
32:37
difficult to to have them as second
32:39
line. and when the national guidance in Argentina
32:41
were being updated. When
32:45
I have years go. Basically, there was
32:47
a big push by the societies to put
32:50
in the second line, but it
32:52
was it was not feasible,
32:54
basically, because of the cost. So
32:57
But, yes, they they do come
32:59
as a revolution like one
33:01
of the people compared to
33:03
to statin. rights and further
33:05
effects that go beyond
33:07
the the change in glucose and
33:10
and
33:10
the possible effects of cardiovascular
33:13
fidelity and and they're very
33:15
promising. Perhaps what one question that I
33:17
had was so you said
33:19
that you back now to
33:21
extrapolate, to follow the same
33:23
analysis and include
33:27
the SGLT two
33:31
inhibitors, would that make so
33:35
but that would that would
33:37
be the extrapolation, right, not without the
33:39
comparing it to trial data?
33:41
Right. We would essentially repeat
33:44
the emulation -- Mhmm. -- including an arm
33:46
for people who took an SGLT2 inhibitor.
33:48
And I just I don't want to be
33:50
remiss the, you know, this work, which
33:52
I'm up part of is being
33:54
led by Rosalina McCoy
33:56
at the Mayo Clinic, an incredible
33:58
endocrinologist and clinical researcher
33:59
along with Yi Hung Dan. So there and
34:02
there's a big team of other people. So I'm just wondering Sorry. We
34:04
won't give you too much credit. Joe, you
34:06
can really borrow from me now anyway. But
34:08
while done
34:11
all those people.
34:18
So to our
34:20
final item, chemical
34:23
academics behaving badly, I'm gonna be I'm
34:25
gonna do something which
34:27
I really do, which is I'm gonna be
34:30
silent and see see
34:32
what Joe and Juan wants to
34:34
say because that this is think
34:36
you guys are gonna ask me some questions.
34:40
So
34:41
though I
34:44
started following this this
34:46
debate on online because I
34:48
saw this editorial of the
34:50
British Journal sports medicine about the
34:53
retraction of many of the
34:55
early trials and was a little
34:56
bit confused at first because I
34:59
started seeing all pop up of subtraction,
35:02
subtraction, mass massively the
35:04
journal. And In ABMJ journal,
35:06
no less. In ABMJ
35:08
journal. And of
35:11
course, I I was
35:13
immediately intrigued. Calvin, could you can you
35:15
tell us a little bit Well,
35:18
I'm not sure how much is the backstory, Matt,
35:20
that is all out of it. keep it
35:24
brief. Yeah. Helen, tell us what you've been doing for the past two
35:26
years. He's not as long as she
35:28
is. So a few
35:30
months back, we did a
35:31
track piece.
35:33
It
35:33
was a comedy piece written by a
35:36
clinical academic called Dr.
35:37
Paul Macquarie. He is
35:39
a sports
35:39
medicine doctor House
35:42
in Australia. And he was also the in chief the Journal
35:44
of Sports Medicine,
35:44
BTSM, which is published by
35:47
The Journal. And
35:48
the
35:49
we retracted that 30s piece
35:51
because he plagiarized some content from the early
35:53
thousands. And soon after we
35:55
retracted that piece, we
35:57
were alerted to
35:59
some
35:59
other potentially
36:01
problematic pieces, which might
36:03
contain plagiarism.
36:04
There was
36:06
an instance of misrepresentation
36:09
of someone else's
36:11
quotation in
36:11
Macquarie's work
36:13
and some
36:15
examples of potentially duplicate publications, self
36:20
plagiarism. And so for
36:21
the last few months, once strand
36:23
of my job has been working with the
36:25
current editor in chief of BGSM,
36:27
and
36:27
he and my colleague
36:29
in
36:29
research integrity PMJ have been looking into the allegation.
36:31
We've now provided an update on our
36:34
investigation and our
36:36
next steps.
36:37
So the news
36:39
is is so the
36:41
news
36:41
is we've retracted nine more
36:43
papers on which Macquarie was
36:45
a soul or single office, and
36:48
he was just writing these pieces alone. They were all commentary style pieces,
36:50
so not academic research
36:52
pieces.
36:55
five
36:56
contained plagiarism,
36:58
so similar to
36:59
the first example. Three were
37:01
duplicates or redundant publications where
37:03
he'd been inappropriately reusing
37:06
chunks of his own work
37:08
without reasonable attribution.
37:11
And in one
37:12
case, there was this instance where he
37:15
misquoted the view of another clinician
37:17
from nineteen fifty. And this
37:18
was a US doctor called
37:21
Thondike. He was seeing patients, he had
37:23
sports injuries at a university
37:25
and and the full driver's recording
37:27
and
37:28
making recommendations based
37:30
on his
37:31
experience observation. This
37:32
is
37:33
quite an old paper but it
37:35
included some cases of concussion. And Thond dupe
37:38
made a recommendation that after
37:39
a certain number, the
37:41
barriers to concussions,
37:42
there's a concussion somebody,
37:44
basically, should have played contact sports again.
37:46
That was the insinuation
37:48
of the of the piece.
37:50
But the way that Macquarie had quoted it was
37:54
inaccurate but it also altered the
37:55
mirror because
37:58
Macquarie quoted Thondike is saying
37:59
that people in these
38:02
circumstances shouldn't play sports until
38:04
the next season.
38:06
And that and
38:07
that wasn't quite
38:09
right. So that
38:10
and That dealt
38:11
with allegations, but we also needed
38:13
to take stock of
38:15
where we are to
38:17
communicate our findings,
38:20
to others,
38:20
and
38:22
and our concerns
38:22
because our trust is broken
38:25
in in this esoteric chief.
38:27
And we
38:27
needed to think about how we handle the
38:30
rest
38:30
of Macquarie's work published in our
38:32
journals
38:32
and by others to to
38:34
play their part in that.
38:36
And so on that basis, we decided to place expression
38:38
of concern upon any content,
38:40
which is written
38:41
by Macquarie
38:44
Ware he is a single or sole author. So his commentary
38:46
style pieces, but
38:47
he's also published a lot of other content,
38:50
particularly in
38:52
BGSM.
38:53
where he's published
38:56
research papers and also this
38:58
consensus statement for
39:00
concussion sport. which is very key to that
39:02
community
39:02
of doctors and and management
39:04
of that particular condition. and
39:08
we're going to ask his institution to look into the standing of
39:11
his research. And for
39:12
the makers of their consensus statement to
39:15
look into whether they consider
39:18
that there's anything
39:18
to be concerned about
39:19
in that regard. We we scream through
39:21
just for plagiarism in the
39:23
in the latest
39:26
twenty sixteen. version and
39:28
didn't find anything there. But
39:30
I still think there's a considerable way
39:32
to go
39:33
on this journey.
39:36
So assuming the best, for
39:39
example, and it's just editorial my
39:41
my Because you are kind,
39:43
haunted present. Wow. Yeah.
39:46
and and that no no harm to patients or will
39:48
come out of it. And to
39:51
what degree do you think
39:54
it is something that we're just seeing one end
39:56
of the spectrum and of
39:57
something that might be happening more
39:59
more globally. Because
39:59
I imagine
40:02
there might be a lot
40:04
of academics listening to this podcast who are quote quoting
40:08
and paraphrasing all the
40:10
time. No. No. No. No.
40:12
No. No. No. It won't
40:14
be very very paranoid because we
40:16
have this. III thought
40:18
in authentic CADE system
40:20
and and Scholar one and the art system
40:22
that sort of catches like
40:24
this gross
40:26
plagiarism. But
40:28
if so, I I once had a
40:30
paper that came and said thirty percent
40:32
overlap and he started seeing all this
40:34
color coded system of the plagiarism system. And
40:37
it was very confusing, like,
40:39
plain jenga. And and but
40:42
beyond that, I
40:44
I couldn't quite III understood that
40:46
the authors were trying to do the
40:49
best of paraphrasing, but they were
40:52
doing very badly. But it's And also, Helen,
40:54
I'm just curious before you answer that is,
40:56
like like how much of this was
41:00
self plagiarism where the,
41:01
you know, essentially, the author's writing the same thing kind of over
41:03
and over as opposed to plagiarizing somebody
41:05
else's writing, which I think is Yeah.
41:07
Mostly Even plagiarism of
41:10
other people's
41:10
content, which I think we can we can all accept
41:14
is is worse
41:15
than reusing the
41:18
AI content.
41:20
And I think, you know, there there are efforts going
41:22
on to help researchers, you
41:24
know, well intentioned people with that paranoia
41:27
of am I
41:29
and my par you know, do I
41:31
need to paraphrase this a bit more so I'm
41:33
not plagiarizing myself? And I
41:36
think this is ah publishers
41:38
are have to
41:38
be pragmatic about these things. If you're so
41:40
so
41:41
the things that we might start
41:43
to look at is
41:45
you know, if you're
41:46
reporting work from
41:48
a research study and the methods
41:50
of that research study
41:53
are the same as methods that are used in a different paper,
41:55
there isn't a need to rephrase
41:58
that for
42:00
the
42:00
sake. trying to
42:01
evade the plagiarism detection
42:04
software. You just need to reference the
42:06
methods and say they're the same, and it's
42:08
absolutely okay to reproduce those.
42:10
That's that's fine. That's
42:12
transparent and honest and useful, I
42:14
think. And That's
42:15
what sorry. That's what called
42:17
text recycling. Text recycling.
42:18
Yeah. So that so it I
42:20
suppose it's sort of like a euphemism, but it is
42:22
but it is I think I think it's
42:24
important to conceptualize it as
42:27
a totally different thing. And
42:29
it's also okay
42:30
to write about the
42:32
same research in different
42:33
ways. You know, you could publish your research
42:36
paper. Joe was
42:36
talking about his research today. He
42:38
might
42:39
then want to write an
42:41
opinion piece or he might want to write an
42:43
editorial for a diabetes journal and
42:45
to talk about that study
42:47
and he might repeat some of the results. And again, there's
42:49
no need to be constantly paraphrasing
42:52
all of that information. You just need to
42:55
make it clear and be clear that
42:57
you're not publishing redundant materials.
42:59
So if Joe were to take his research paper
43:01
and so then try and publish it somewhere else,
43:03
that would clearly be inappropriate. That would
43:05
be And just for
43:06
the record, I would never do that. I
43:09
would never do that. I
43:12
didn't want I'm coming off the major
43:14
declaration. No. But there is this practice and
43:16
we see it with trials. I was
43:18
just looking at this when
43:22
when when Ross was summarizing the diabetes
43:24
prevention trial,
43:26
there were, like,
43:27
over forty publications in,
43:30
like, a very short period of the And this is, of
43:32
course, many of them might
43:34
be super relevant, but there's
43:38
a very gray
43:38
concept that is this salami slicing where you have a study
43:41
and you make a lot of papers out
43:43
of it and which
43:45
is set up to the
43:47
point of redundancy. Right? And some
43:49
of these decisions aren't
43:51
easy. And I think this relates to,
43:53
I mean, the art sort of the
43:55
answer to your question, although I've kind of forgotten where your first question started by,
43:57
which is that in general, this
43:59
work is
44:02
is
44:02
quite complex. You have
44:04
to to weigh up a
44:06
a
44:06
lot of information. There's a lot of
44:08
judgment involved. Sometimes it's very
44:12
obvious. things have been cut and pasted. But
44:14
as you say, looking at different publications and thinking, okay, what are the
44:16
new outcomes here or what new audience
44:20
you're trying to reach, you have to you have to weigh those things
44:22
up and come to a decision just like you would
44:24
with any with any sort of
44:28
primary content. So would you
44:28
say just to wrap up, would
44:31
you say that copy and
44:33
pastes someone else's comment
44:36
or publishing your article several times is a
44:38
no no, but perhaps text
44:40
recycling for the methods that
44:44
far. They're always the same crowd. I think we should do a future
44:46
episode on plagiarism and
44:49
fraud. No. Fraud
44:51
is like that another category. I
44:52
feel like this episode has not a
44:55
very practical information for sort
44:57
of continuing professional development. CPD,
44:59
as we call it, over
45:01
here in the UK. And I think while though it's something that you'd
45:03
have to go and reflect upon
45:06
now. It's hard to have a
45:07
good all think about. How many c
45:09
me c m credits you. The listeners
45:11
get it. Well, it depends on how that we
45:13
talk to. Well, not
45:17
sure it should depend on that. It shouldn't depend on the
45:19
quality of the conversation. But Joe tells
45:21
us there's somewhere else he has to be enough. So
45:23
we we we've gotta wrap it up
45:25
for this week. bring the show to
45:27
a friend. So I would like to thank you both, Juan and Joe for
45:29
joining
45:29
us. I'd like to thank
45:31
you for listening. you
45:34
can subscribe wherever you
45:35
listen to a podcast, you can rate us
45:38
especially if you like to rate us
45:40
very highly.
45:42
we really enjoyed looking into Nicole's query
45:44
this month. And if
45:45
you have a query that you'd like us
45:47
to look
45:48
into, or you'd like
45:50
to give further query to me so I can
45:52
give mine and do some homework to do,
45:54
then that would be wonderful.
45:56
So until next month, it's goodbye for me. Goodbye
45:59
for me.
45:59
Goodbye for me.
46:02
Take care.
46:02
the cow that
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More