Latest / 'Bone Up' / BoneUP Live @ ECTS 2025 - Part 1

Transcript

BoneUP Live @ ECTS 2025 - Part 1 · 'Bone Up'

0:00
Hello and welcome to Bone Up, the podcast all
0:04
about bones, how we make them, why we break them
0:07
and if we fully understand them. I'm David Armstrong.
0:11
Hi and I'm Richie Abel and over this series we're
0:15
going to be exploring osteoporosis, bones, what
0:18
we know and what we're yet to discover and we
0:21
hope you will join us on the journey. So for
0:24
anyone keen to learn more about our infrastructure
0:27
of calcified collagen, this is bone up hello
0:45
and welcome to another edition of bone up live
0:58
We're here in Innsbruck in the Tyrol region of
1:01
Austria to attend the European Calcified Tissue
1:05
Society Congress. Richie, excited? Oh, very excited.
1:08
This year at Bone Up Live 2025, the ECTS Congress
1:13
have been really kind to us and given us our
1:16
own room. We're in the Alborg on floor three.
1:18
We've got beautiful views of the Alps out of
1:21
the window. Listeners, I wish you could see this.
1:24
Cloud covered Alps stretching back nice and green
1:26
into the distance. It's absolutely beautiful
1:28
here. Yeah, it really is. It really is very impressive.
1:30
Lovely place to have a conference and lovely
1:33
place to talk to people about the latest news
1:36
in clinical and basic science research in the
1:40
world of bone. And we've got our famous mugs
1:42
here with us as well, Richie. We have listeners.
1:45
The guests are really keen to get their hands
1:46
on a mug. They're becoming a highly sought after
1:49
item in the bone world. We hope we bring you
1:52
a bit of a flavour of what's going on at the
1:53
conference here as we interview some high -profile
1:56
speakers and some people towards the start of
1:58
their career just presenting their work and their
2:01
papers here at the conference. David, last night
2:05
we went to the opening ceremony of the conference
2:07
and we heard a really great debate about the
2:10
treatment of patients with diabetes and whether
2:14
or not patients with diabetes should receive
2:16
a DEXA scan as part of their standard care. What
2:20
did you make of the debate last night? I thought
2:22
it was a useful debate. It was sort of couched
2:25
in terms of our representatives from the ECTS
2:27
speaking against a representative from the American
2:30
society, although in effect actually one of the
2:34
speakers was from Italy and one from Australia.
2:37
But it was very useful and regardless of the
2:39
outcome of the debate. It raised issues about
2:42
bone quality as well as just bone quantity, about
2:45
how DEXA can be useful, but how fracture risk
2:48
is a lot more than DEXA, and how we should be
2:51
looking outside just our sort of narrow vision
2:55
on bones to all the other groups of people, for
2:58
example, people with type 2 diabetes who we know
3:01
have higher fracture risk, but we may not be
3:04
certain just how... best to assess those patients
3:07
at the moment so yeah it was a very it was an
3:09
interesting start to the conference. I feel it
3:12
set a very good scene for the conference there's
3:15
a lot of research still to do to improve patient
3:17
care and we've got to get all of the basic scientists
3:19
and the clinicians together and it was nice to
3:22
see everybody coming together for the debate
3:24
last night and hopefully listeners we're going
3:26
to bring you a flavour of that debate over the
3:28
coming days. So we better get on to our first
3:33
guest. Listeners, we're very pleased to announce
3:37
our first guest of ECTS 2025, Giovanni Adami.
3:41
Giovanni is a rheumatologist and MD from the
3:43
University of Verona. Welcome. Thank you. Thanks
3:46
for having me. It's a great pleasure. We're delighted
3:50
to have you here. There's so many things we could
3:53
talk to you about. You have so much on your CV.
3:56
I tend to get into the very clinical questions,
3:59
so I'm going to ask you immediately about a paper
4:02
you published last year. which looked at the
4:08
interesting and relevant question that if you're
4:11
a patient who is receiving denosumab injections
4:14
and you have a further fracture, what do we do
4:18
at that point? And you looked at adding romosuzumab
4:22
to denosumab? Yeah, correct. So what was the
4:24
outcome from that? Well, that paper we published
4:27
in 2024 was focused on the six months results
4:31
of an observational study that is... still ongoing
4:34
and we are going to publish the 12 months results
4:38
hopefully very soon. So basically the underlining
4:41
question is what to do when you fail the nosema.
4:45
So failing the nosema means that you have a very
4:48
high risk patients and you want to do more. for
4:51
that patient. And so you might have a very low
4:54
BND or not reaching the threshold for being happy
4:58
with their BND or having a new fracture. And
5:01
the problem with that is that the conventional
5:04
sequence treatment or whatever guidelines we
5:07
are looking at suggests to switch to an anabolic,
5:11
which makes perfectly sense for bisphosphonates,
5:15
but it's not really the... most advisable strategy
5:20
for a patient on the NOSMA because we fear the
5:24
rebound. turnover and multiple vertebral fracture
5:27
after the nosomab discontinuation. So the problem
5:31
here is that it would be advisable to discontinue
5:35
whatsoever the nosomab independently for what
5:38
you are going to give the patient after that.
5:41
So one can think, no, I can switch to xyloduronic
5:45
acid and then consolidate and then give an anabolic.
5:48
Yeah, that might be reasonable, but you're really,
5:52
really looking... for an imminent risk of fracture.
5:56
So you want to give the best treatment immediately
5:59
to that patient. To work most quickly. Yeah,
6:02
to work quickly. Yeah, that's exactly the point.
6:05
So maybe combining the treatment might be the
6:08
idea. And the working hypothesis for our study
6:12
was that maybe we can combine teriparatide or
6:15
abaloparatide in these patients, but... We do
6:19
not have the evidence for ongoing patients with
6:22
denosumab, so with a closed remodeling surface
6:26
and add on top of that teriparatide. So it might
6:30
not work properly as has been seen in the data
6:34
study. So the working hypothesis was that maybe
6:38
romazosumab, by leveraging a modeling, and so
6:42
it might be more effective. And so we tested
6:45
that. And that was an observational study. So
6:49
basically, we just enrolled patients that need
6:51
a combination treatment that was eventually found
6:55
out to be reimbursed in Italy because there's
6:59
no contraindication in the label and no contraindication
7:04
in the reimbursement criteria. So we combined
7:07
the two treatments, and we found that there was
7:09
an increase. as compared to patients that continue
7:12
denosumab. So some of the patients just continued
7:15
with the denosumab, and some you added romosuzumab
7:18
as well as the denosumab. So as well as getting
7:20
their six -monthly denosumab, they were getting
7:22
romosuzumab as well. And in those group of patients,
7:26
we always think of the spine and the hip. So
7:30
what happened at the spine? That was something
7:33
that we would not really have an explanation
7:35
for that. Maybe it's just the follow -up times
7:38
we will have the results of the 12 months, or
7:41
maybe it's just the sample size problems. But
7:44
what we found was that in the combination group,
7:47
the lumbar spine, well, very well, it increased
7:51
more than the nosomab alone, you might have said.
7:54
So if you added romastuzumab, your spine got...
7:56
this pine went up good yes by seven percent in
8:00
six months which is remarkable absolutely yeah
8:03
But the total lip and femoral neck did not perform
8:06
as expected. So they stayed quite stable over
8:11
the six months. They don't go down. They don't
8:13
go down. That's important because that was a
8:14
concern. If you added teriparatide, if you switched
8:17
from dunnusumab to teriparatide, bone density
8:20
might go down. It might go down, yeah. It didn't
8:22
go down, but it didn't go up either with the
8:24
romacizumab. Is that right? It didn't go up either.
8:26
I can tell you that we have a poster here. We
8:29
did the presentation at the WTO. We are preparing
8:33
the paper that for the 12 months, it goes a little
8:37
bit up. Breaking news, folks. So it's good results
8:43
and confirms what we would have expected. The
8:47
question now is, is the switching from the Nozumab
8:53
to Romozozumab as effective as doing the combination?
8:58
Well, actually, we did not do that. Because we
9:02
feared the rebound. Because most of the patients
9:05
that we had were on very long -term denosumab.
9:08
So basically the risk for rebound vertebral fracture
9:10
was very high. But there is a study, if you look
9:15
in the literature, by Hong, which published the
9:17
switching from denosumab to ROMO. But, again,
9:23
in that study, the mean treatment duration with
9:26
the nosema was two years. This is not that long.
9:30
And they found, actually, a rebound. They found
9:34
that CTXA is going very, very high after the
9:37
introduction of ROMO. So there's still an open
9:41
question, I think. A randomized control trial
9:43
is highly needed here. Yeah, and I think this
9:47
highlights something we may talk more about probably
9:49
at the conference, how while some people are
9:52
very focused on head -to -head studies between
9:55
drugs to show which is better now, most patients
9:58
are more interested in the long -term sequence
10:00
because they would possibly be on most of these
10:03
stronger drugs, and we just want to know how
10:06
do we combine them, when do we stop them, when
10:08
do we stop them, rather than just head -to -head
10:11
studies. So I think it's very relevant. Probably
10:15
you would say then for my patient who I've been
10:17
giving denosumab injections to for five years,
10:20
they've had a new vertebral fracture, there probably
10:22
is evidence from what you suggest that switching
10:25
to romacizumab would be preferable for a good
10:28
number of those patients from your evidence than
10:30
just continuing with the denosumab. Yeah, combining
10:34
romacizumab. Combining might be the right option.
10:39
I mean, that is quite consistent throughout the
10:44
medicine. Combining two drugs is better than
10:47
using one. And that has been shown by the data
10:50
study that was naive patients. There are many,
10:54
many convincing studies in hypertension and cardiovascular
10:58
risk, which we combine the drugs. And the rational...
11:04
behind combining an anthroposophist with a bone
11:07
anabolic, and in this case, ROMO is a very strong
11:10
bone anabolic. Don't forget about that. Well,
11:12
we have some clinical evidence, at least. I suppose
11:17
we should say this is all bone mineral density
11:19
evidence. It's not fractures. Obviously, we're
11:21
talking about it six months. Absolutely. Nevertheless,
11:23
we're reasonable to presume that if your bone
11:25
mineral density goes up, it's going to be...
11:27
That's very reasonable, yeah. No, that's very
11:30
interesting. We wait for the 12 -month data with
11:32
excitement. Giovanni, you've got another talk
11:35
at the conference. Our regular listeners will
11:38
know that romisosomab treatment lasts for 12
11:40
months because bone is homeostatic, and after
11:43
the treatment is given, bone will return back
11:45
to its normal physiology after a while. And you've
11:48
been perhaps investigating some of the reasons
11:50
why that might happen? Well, yeah, thanks for
11:52
asking. Yeah, basically... We did a prospective
11:55
study in which we observed in naive patients
11:58
with no treatment at all, and we started bromo.
12:02
We want to investigate what are the factors that
12:06
are associated with the waning of the rhombozoosoma
12:09
bufficacy over the time. And so we explore what
12:11
could be the most reasonable thing, which is
12:15
an increase in the wind inhibitors other than
12:19
sclerostin. And so we explored the... mostly
12:23
studied and probably the most important wind
12:28
inhibitor, which is TKK1. And that is, this is
12:32
again... It was surprising to look at the literature
12:35
and finding that nobody has published anything
12:38
on that. So I strongly suspect that the companies
12:42
had that type of data, but we don't know, actually.
12:45
We don't have the data. We don't have the publication
12:47
for that. So we just very simply, very easily
12:51
measured the DKK1 over the time. And not surprisingly,
12:55
we find out that was increasing. And that increase
12:59
was correlated... beyond month three of treatment
13:03
to the decrease in P1MP. So P1MP peaked, DKK1
13:08
stays stable, and then after the peak in P1MP,
13:11
we found an increase, very, very linear increase
13:18
in DKK1 over the time. So our speculation is
13:24
that, well, that's not surprising. That was somehow
13:27
expected. That's negative feedback of the bone.
13:31
that tends to go back to the normality, and was
13:35
not expected based also on studies in sclerostosis.
13:40
Sclerostosis patient has been shown to have DKK1
13:43
higher than normal, and that might be a protective
13:46
mechanism that avoids an excessive bone formation
13:51
that will kill the patient. So eventually, if
13:54
you look at the natural history of sclerostosis,
13:56
they have a plateau. in bone mineral density
13:59
that one of the possible reasons not the only
14:02
i agree but one of the possible reason is that
14:06
they have an increase in dk1 serum levels so
14:09
it might be a protective effect and What is interesting
14:13
to know from the literature, again, is that if
14:16
you combine undisclabrastin and anti -DKK1, you
14:21
have a very extremely potent drug. And there's
14:24
the phase one study from the dual inhibitor,
14:28
very recently presented at ASBMR and IOF. And
14:34
so they showed an increase in bone mineral density,
14:38
which was absolutely stunning. skyrocketing up
14:42
BND but the problem might be safety you know
14:47
there are some preclinical studies showing especially
14:51
in monkeys that the dual inhibition might result
14:54
in the thickening of the the cranial bones and
14:58
so might be harmful. We really don't know. We
15:02
have some reassuring phase one data from that
15:04
study that is safety is okay, but we really don't
15:08
know. It might be that the increase in DKK1 is
15:12
some kind of protection. Yeah, which is something
15:15
you see in sclerostosis, the thickening of the
15:17
skull bone as well. But perhaps in a few years,
15:20
we'll be talking about a DKK1 antibody here at
15:23
these meetings. Hopefully, yeah. Our regular
15:26
listeners who tuned into ECTS last year, remember
15:30
you heard it here first when we spoke to Roland
15:32
Barron. He told us that there was a study ongoing,
15:36
and it's great to hear the results are out now.
15:39
So if I follow this right then, basically what
15:42
you're saying is the osteocytes then, the bone
15:43
cells, they release sclerostin and DKK1, which
15:48
both suppress bone mass. And if you suppress
15:51
the sclerostin... with romisosomab to relieve
15:54
the inhibition on wind and increase bone mass,
15:56
the osteocytes then come back by releasing the
15:59
DKK1 to rebalance the system. Well, the main
16:02
source of DKK1 is probably not osteocytes. It's
16:07
probably released by many, many other cells.
16:11
One of that is the osteoblasts. Other that are
16:15
secreting are... from the macrophage lineage,
16:19
so T cells and other immune cells. So it's not
16:23
really clear where does this DKK1 come from,
16:28
but what is evident is that it plays a major
16:31
role. So if you inhibit that, and what is even
16:35
more interesting is that DKK1 is a very crucial
16:38
mediator of bone loss in multiple myeloma and
16:42
very erosive disease like... rheumatoid arthritis
16:45
so that's very fascinating thinking of you know
16:49
yeah romo might be very efficacious in multiple
16:52
myeloma but who knows what happened on the DKK1
16:55
side who knows you know the pros and cons and
16:59
need to be studied does that mean that maybe
17:02
DKK1 antibodies could potentially be a treatment
17:05
for rheumatoid arthritis as well yeah absolutely
17:08
that is very the very next future i think there
17:13
have been studies phase one studies with oral
17:17
dkk1 inhibitors in multiple myeloma That was
17:21
a very promising drug for multiple myeloma. But
17:24
unfortunately, the problem with DKK1 is, again,
17:27
it's produced by many cells all over the... So
17:30
it's ubiquitous. And you have a lot of cells
17:33
producing, so the safety concern is always one
17:36
of the big problems. Sclerostin, on the other
17:40
side, is very tissue -specific. It's just secreted
17:43
by osteocytes just in the bone, so it's more,
17:46
let's say, safe. Giovanni, you're one of the
17:51
most active people at the ECTS conference. Every
17:54
year you give lots of talks and I see you running
17:56
around everywhere. I was lucky enough to see
17:58
your talk yesterday about artificial intelligence
18:00
and you gave a really fascinating talk about
18:03
how AI might be able to improve fracture risk
18:07
prediction. Would you tell our listeners about
18:09
your research? Well, how can we improve fracture
18:13
risk prediction and why it is important to improve
18:15
fracture risk prediction? I mean, we have the
18:17
DEXA. The problem with the DEXA is that it's
18:20
not capturing fractures. It's a very good instrument
18:26
for patients that have osteoporosis, but for
18:28
patients that fracture in the osteopenic levels,
18:31
which are the vast majority of the women all
18:33
over the world, you might misdiagnose those patients
18:38
because they are at risk, actually, because they
18:42
will fracture in the future, but they do not...
18:47
osteoporosis from the classical densitometric
18:50
definition and classification. So basically we
18:55
need something else, so we included the clinical
18:59
criteria, that is the FRAX or other fracture
19:02
risk assessment tool, but that makes a big problem.
19:06
The big problem is that we need to balance the
19:09
complexity have the fracture risk algorithm.
19:13
So we really do not want a fracture risk algorithm
19:18
that makes hundreds of variables inside of that.
19:21
It's not practical. It's very unfeasible in the
19:25
clinical practice. On the other side, we need
19:28
to have precision. And the problem with that
19:31
is that the FRAX, it is improving our precision
19:35
as compared to DEXA. but it's not as precise
19:40
as we might desire. So the area under the cure,
19:44
which is a matrix for performance of the classification
19:47
accuracy of these algorithms, is around 0 .6,
19:52
0 .7. It depends on the population. So it's not
19:54
that good. We are missing a lot of patients.
19:57
And that is where artificial intelligence might
20:00
help a lot, because it can digest the complexity,
20:04
and so you can... But a lot of variables, and
20:07
in terms of seconds, have your classification
20:10
criteria and your classification of the patient,
20:13
so at risk or not at risk, and what type of risk.
20:15
And the precision improves substantially. So
20:20
basically, there are many studies showing these.
20:23
And they leverage on many features that we previously
20:28
did not think. were associated with fractures.
20:31
So that's another very, very interesting point.
20:35
Artificial intelligence is looking at patterns
20:39
inside your variables, inside your data set,
20:43
that you might never think of were associated
20:46
with fractures. So that's a very important thing.
20:49
I agree with you that sometimes we do not have
20:53
an explanation, and physicians... tend to be
20:56
very attracted to explanation, to biological
21:00
rationale. So they really do want to see some
21:05
flow chart. They do want to see what is actually
21:07
increasing the risk in my patient because I want
21:10
to modify that risk factor. But now we have some
21:14
explanatory models, so explanatory artificial
21:17
intelligence that are overcoming the black box
21:20
which is inside it. So we might have answers
21:23
even to that. It's fantastic to hear about such
21:28
cutting -edge research in the bone world. And
21:31
as well as predicting fracture, can you see AI
21:33
being able to give us an idea of how we would
21:37
treat patients in the sense of the sequence of
21:40
drugs or which patients may respond better to
21:43
different drugs that we wouldn't necessarily
21:45
see from our randomized control trials? Absolutely.
21:49
That is absolutely true, especially because,
21:53
you know, you leverage big data. And randomized
21:56
control trial is a very homogeneous population,
21:59
so you cannot really draw a conclusion on prediction
22:02
of response in different populations because
22:05
you don't have different populations at the randomized
22:07
control trial. It's very homogeneous, mostly
22:10
healthy patients. And so basically, the big data
22:16
can include a lot of variables, and you can even
22:19
predict response to treatment. So this is truly...
22:23
precision medicine you know so that type of patient
22:27
might benefit to this treatment followed by that
22:30
and that and that for the lifetime and that is
22:33
clearly the near future and will avoid a lot
22:36
of over treatment a lot of the expenditures due
22:39
to over related to over treatment and so eventually
22:43
improves the quality of care overall It's amazing
22:48
to see you pushing forward on so many fronts,
22:51
Giovanni. Thank you so much for coming to talk
22:53
to us today. Thanks. It was a real pleasure being
22:55
here with you. It's great to speak to you. Thank
22:57
you. Thank you. Listeners, we're really pleased
23:05
to introduce our next guest today. His name is
23:07
Professor Klaus -Christian Glue. He's from the
23:10
section of biomedical imaging. in the Department
23:13
of Radiology and Neuroradiology at the University
23:16
of Kiel. Klaus, welcome. Thank you, Richard and
23:20
David. It's a great pleasure to be with you here.
23:22
It's very nice to meet you and very nice to meet
23:24
your wife as well, who I understand is a keen
23:26
listener to the podcast. We both are, actually.
23:29
I think it's terrific what you've put on and
23:31
it's a great way of really broadening the awareness
23:34
of osteoporosis and musculoskeletal. Thank you.
23:37
Fantastic, thank you. There are so many things
23:40
we could ask you about. One of the things that
23:43
really impressed me with one of the talks you
23:46
give here at the conference, right from the start
23:50
of our podcast, we talked to people about estimating
23:53
fracture risk and how it's much more than just
23:55
doing a DEXA scan. We need to look at lots of
23:58
things and put them together. And we introduced
24:00
the audience to this concept of the FRAC score,
24:03
which is an algorithm to help estimate fracture
24:06
risk. But you told us about... Another a new
24:10
algorithm perhaps for fractures called hive.
24:13
Is that right? Tell us a little more about about
24:15
that and where it might fit in Yes. So, yes,
24:20
we have developed a new algorithm that you said
24:23
and the name is Hive and this comes from hip
24:26
fractures, HI, and vertebral fractures, VE. So
24:30
this is right away one of the things that we
24:33
do a little bit different from, for example,
24:35
the Frax or the Garvin or other wrist calculators.
24:39
So we focus on what we believe are the two most
24:42
important fractures and also the most typical
24:45
fractures for osteoporosis, particularly vertebral
24:47
fractures. So to get on the one hand really the
24:51
serious outcomes in the centre of attention and
24:55
secondly also improve hopefully the ability to
24:59
predict fractures and to identify patients at
25:02
risk. It was very interesting when we saw your
25:05
talk yesterday. In the HIVE algorithm, you calculate
25:09
risk at three years, whereas some of the other
25:12
risk assessments like FRAX calculated at 10 years.
25:15
Why did you go for three years? Yes, I think,
25:19
first of all, I'd like to emphasize that FRAX
25:22
has done a fabulous job over the years. I mean,
25:25
we need to appreciate that it has been around
25:28
now for more than a decade and really brought
25:32
across... the concept of fracture risk and how
25:35
important it is to make decisions about interventions.
25:39
But we thought a long time ago, what could we
25:44
do better? And we always felt when we talked
25:47
to physicians, personally I'm a physicist, so
25:50
you have to take all my words with a grain of
25:54
salt, but when we talk to clinicians, they say
25:58
this is very difficult to understand. When you
26:02
think about your health, three years is already
26:05
quite a distance, and ten years is hard. Particularly,
26:11
we quite often talk about very elderly people,
26:14
and if you're 75, this is not the horizon that
26:18
you are concerned about. There's another reason.
26:24
It's easier to predict the risk when the time
26:28
interval is shorter. So your ability... to differentiate
26:31
patients who should get treatment and not get
26:34
treatment is facilitated because of stronger
26:38
gradients of risk over the shorter interval.
26:40
That was part of the data that I showed here.
26:42
And of course now we have drugs, the osteoanabolic
26:44
drugs, which work very quickly and we're talking
26:47
about fracture risk reduction within one or two
26:49
years, whereas perhaps 10 or 15 years ago when
26:52
it was bisphosphonates, the fracture risk reduction
26:55
was over a much longer period. And where do you
26:59
get the data from for calculating the outcomes
27:03
then and how? Where did that come from? Yes.
27:06
So we, of course, have German guidelines on osteoporosis,
27:13
and there's a very active group in Germany led
27:16
by Friederike Thomasius. And so a couple of years
27:22
back, we initiated a very thorough literature
27:25
review. We looked in the end at more than 2 ,000
27:30
papers in the setting that talks about risk factors,
27:34
etc. And had an external company grading the
27:39
evidence of all of these papers and then really
27:43
picked from the literature the most. relevant
27:47
in terms of strength, in terms of prevalence,
27:50
risk factors in the context of hip fracture prediction
27:54
or vertebral fracture prediction. And these were
27:56
papers from around the world? From around the
27:58
world. Using patients from around the world of
28:00
different ethnicities, backgrounds? Absolutely.
28:02
Because that is really one of the currently still
28:05
underlying assumptions that while we have different
28:08
risks in different countries, the relative risks
28:12
for a given risk factor is fairly similar against
28:16
the background in that country exactly exactly
28:19
so there may be in the future more refined models
28:22
but for the time being we consider this internationally
28:25
established risk factors so um yeah so we we
28:31
did this careful literature review and then of
28:37
course we needed to identify the base risk in
28:41
the population and we started with the German
28:44
population of course but let's say you would
28:47
like to use Hive in the future in other countries
28:50
you would only to re -establish the base risk
28:54
in that particular country and the base risk
28:57
that we established was based on more than 600
29:01
000 patients from an insurance company database
29:04
where we had lots of the risk factors and we
29:09
excluded all of the patients that had any known
29:12
risk factors to get to an estimate of the population
29:17
risk of a super healthy individual without risk
29:20
factors. So that's our base risk and then you
29:23
multiply all the other risk factors on top of
29:26
it to get to the individualized risk of that
29:29
person. And do you need... a DEXA scan? Do you
29:33
need your bone mineral density from a DEXA scan
29:35
or can you do it without that? We can do it without.
29:38
We start with the base risk. We multiply then
29:41
the risk of the two strongest risk factors in
29:45
that individual. So we really go for the concept
29:48
of personalized medicine. We ask patients What
29:53
kind of diseases, thinking of secondary osteoporosis,
29:57
et cetera, do you have? So we identify per patient
29:59
the two strongest risk factors in that patient.
30:02
That gives us then together with the base risk,
30:05
the risk without BMD. And if we have BMD available,
30:10
we currently use the femoral BMD at the total
30:15
hip. And then we have a third component that
30:18
we multiply the... risk and then we get to the
30:23
full model including base risk, clinical risk
30:26
factors and BMD. That's a really good explanation
30:30
of how the model works. I really wanted to know
30:33
how these fracture risk assessments work and
30:35
I think for patients who are listening, people
30:38
who are thinking about doing this test, it's
30:40
really good for everybody to know exactly how
30:42
the test works. What kind of risk factors came
30:45
up as being the most important risk factors for
30:48
most patients? Okay, so this literature review
30:53
that I talked about resulted in 33 risk factors
30:56
picked. Okay. So it's quite a list. Quite a lot,
30:59
yeah. Yes, and of course, you always could simplify
31:05
things by disregarding something, and each doctor
31:09
may have... patients that come from different
31:12
disease areas. So from a rheumatologist, there
31:15
may be different risk factors, most prevalent.
31:17
Lots of people with steroids or something? Exactly,
31:19
yeah. But then when you have very elderly people,
31:22
the geriatrician would have more like the false
31:24
related risk factors. So timed up and go, how
31:28
long does it take? These kind of tests. And so
31:32
the most important ones, you could say, okay,
31:36
are these the most prevalent ones in the population?
31:40
Or are these the strongest risk factors? So among
31:43
the strongest risk factors really are for example
31:47
having already vertebral fractures and maybe
31:50
having not only one but two or three. And if
31:55
we have radiographs we could incorporate also
31:57
the severity of those fractures. And Dennis Black
32:01
has written a wonderful paper years back that
32:03
the relative risk could range from anywhere from
32:06
2 to 10, depending on what the vertebral fracture
32:10
status is. And the previous hip fracture is very
32:12
important. But then also we need to look at patients
32:16
that haven't yet fractured. So there are important
32:21
aspects like Parkinson's disease, for example,
32:25
then glucocorticoid therapy. Many more. That's
32:33
really incredible. Not only that you're providing
32:36
an opportunity for people to have really individualized,
32:39
personalized care, but actually Hive fits into
32:44
the care pathways that the patients are going
32:46
through. Clinicians in lots of different fields,
32:50
you know, if bone medicine isn't your expertise,
32:52
having fracture risk assessment tools which contain
32:55
the kind of data and measurements that you're
32:57
used to using, you're used to collecting from
32:59
your patients, might make it a bit easier for
33:01
you to use as well. And can our patients go online
33:04
and measure their own HIVE score? now or we're
33:08
not at that stage yet? We're not at that stage
33:11
yet but it's exactly where I think the field
33:13
needs to head because obviously recording the
33:17
33 takes time and time is precious. Can I go
33:22
online and calculate it for my patient as well?
33:25
No, it's not freely available yet. The point
33:28
is that we have a group of three people who have
33:35
developed this algorithm and leading that group.
33:40
And then we have done this for the DVO, the umbrella
33:43
organization of about 20 bone societies in Germany.
33:49
So the DVO has this guideline group that does
33:52
the osteoporosis guideline. And so we have provided
34:00
now this algorithm. We have tested it. in certain
34:04
populations, now here with the soft population,
34:07
but in order to apply it in the clinical setting,
34:12
you have to certify it as a medical product.
34:15
That's a very expensive, complex process. Doesn't
34:21
it seem strange? You have to certify it as a
34:23
medical product, but you do, and that's only
34:25
right, because if you're making important management
34:27
decisions based on it, then it has to be properly
34:29
certified and regulated. When clinicians make
34:34
the management decisions with the patient, does
34:37
the Hive software help them to do that? I definitely
34:43
hope so. Having just talked about that it's not
34:46
yet certified, the base version, the simplified
34:50
version of the Hive algorithm is part of the
34:54
guidelines of 2023 in tabular form. So currently
35:00
what you would have to do... is to use those
35:03
tables. They are available in the net. And it's
35:09
a fairly simple way of calculating yourself whether
35:14
you are above or below certain intervention thresholds.
35:19
So applying our HIVE algorithm in Germany was
35:22
connected to something that's not the algorithm,
35:24
but it's the guidelines. And the guidelines state
35:27
that you have now three levels of risk. that
35:32
the physician should be worried about. So if
35:34
you have a 3 % risk of hip or vertebral fractures,
35:37
you should start considering. If you're below,
35:40
then you're fine for some time. If you're above
35:43
5%, then it's modified to a very strong recommendation
35:49
to start antiresorptive therapy. And if you're
35:53
above 10, it's a strong recommendation to consider
35:58
osteoanabolic treatment. So in that sense, I
36:01
hope it will help clinicians to really not only
36:04
decide who is to treat, but also to consider
36:07
which type of therapy that we have available
36:10
should be considered. I mean, I really look forward,
36:13
hopefully, to it being more widely available.
36:15
The fact that you include falls in it, I think,
36:18
is very important. And I think the fact that
36:21
it's a three -year prediction is also... It's
36:26
useful for me as a doctor, but I think it's quite
36:28
relevant for some patients because, as you quite
36:30
correctly say, if you're seeing patients in their
36:33
70s and 80s and you give them a 10 -year prediction,
36:36
a lot of patients will say, well, I can't see
36:39
10 years ahead, but three years is very relevant.
36:43
I'm glad you like that. Your comment on falls
36:49
leads me to make one more comment about our algorithm.
36:54
Amongst these 33 risk factors, we have quite
36:57
a number that are related not to the bone aspect,
37:01
but more to the fall propensity aspect. So, of
37:05
course, falls in the last year, none, one or
37:09
more. So we have three levels there. But then
37:12
also diseases. I mentioned Parkinson's before
37:14
that enhanced the risk for falling. So we have
37:17
actually maybe seven or eight. risk factors in
37:22
that category and i mentioned earlier that for
37:25
each patient you pick only the strongest and
37:27
the second strongest yeah so if you have several
37:30
within the fall category you're only allowed
37:32
to pick one of those yes so if you take the parkinson's
37:35
box you're not allowed to take the osteoarthritis
37:38
box as well or something you have to take a box
37:40
in a different category exactly just taking steroids
37:43
or family history exactly exactly because we
37:45
really don't want to overestimate if there are
37:49
two risk factors that are not truly independent
37:51
in the assessment of risk. A big sort of topic
37:57
of conversation in the world right now is artificial
37:59
intelligence and how it's going to change the
38:01
world around us. I've noticed at the CTS some
38:04
artificial intelligence work has started to come
38:06
through and some of the presenters have been
38:09
talking about the potential role for AI in improving
38:13
fracture risk assessments and treatment assignments
38:15
for patients. Do you think that's something that
38:18
might work in the bone world? Absolutely, absolutely.
38:21
So a big part of my research in the past eight
38:25
years has been on developing AI methods, particularly
38:29
for imaging. And to just give you an example,
38:33
in the FRAX algorithm and our algorithm, we're
38:36
currently basing the image information on the
38:40
results from the DEXA device. But there are other
38:43
methods out there. and for example we did a study
38:47
reported at asbmr that just taking a dexa image
38:52
and doing ai on the very same dexa images already
38:56
improves things so but then you have x -rays
38:59
you have other imaging modalities like ct etc
39:04
and i think one great option is to widen the
39:09
availability of diagnostic tests by incorporating
39:14
equivalent ways across all of these different
39:17
imaging modalities. And in that setting, AI is
39:20
a terrific tool to establish, let's say, what
39:23
is a risk threshold equivalent to minus 2 .5
39:27
on a CT of whatever you're measuring. So many
39:33
options in there. And then the second way would
39:37
be to incorporate this. do the modeling of the
39:42
entire algorithm using AI, which also will lead
39:47
to new ways. And in a third way, the way of interacting
39:50
with the patient, think of chat GPT and other
39:53
tools, could be totally different. The ECTS,
40:00
and especially the ECTS Academy, are really good
40:03
at trying to bring through young scientists,
40:05
people building their careers. I guess there
40:08
might be a lot here who might be looking at AI
40:10
and thinking, well, this is clearly an important
40:12
future for our field. How would they get into
40:15
it? What kind of training and expertise should
40:17
young people pick up if they want to get into
40:20
AI? That's a very, very good question. I'm extremely
40:24
glad that you mentioned the Academy. I was fortunate
40:28
to initiate 10 years ago the founding of the
40:32
ACTS Academy and just yesterday we celebrated
40:36
10 years and it was so great that this really
40:40
caught on. And one of the great things of the
40:43
academy is the interdisciplinary nature. There
40:46
are people that have a surgical background and
40:48
internal medicine background, et cetera, and
40:50
everything. And to come back to your question,
40:53
that is the challenge because you need to figure
40:56
out ways how computer people and medical people
41:01
develop some common language to understand what
41:05
AI can do and what it not can do. My whole scientific
41:09
life, I've been sort of, in between the technology
41:13
side and the application side in medicine and
41:17
yes it is very difficult for each side the the
41:21
technology side has great ideas but doesn't know
41:24
what are the main problems that need to be resolved
41:27
and on the medical side you can say this much
41:30
better but but you know what you need yeah but
41:34
you may not be aware of the latest greatest technologies
41:37
so we are always in the middle as medical physicists
41:41
to sort of bring this about. And then training
41:44
needs to go along those lines, bring those people
41:47
together so they understand what the tools can
41:49
do. It sounds like you had a really interesting
41:52
career. How did you go from sort of physics into
41:56
osteoporosis? Very good question. So I did a
42:04
plain physics study. I did my diploma in Hamburg,
42:07
Germany, where I still live. And then my supervisor
42:12
asked me, okay, are you interested in doing your
42:14
doctorate? I said yes. And so he said, okay,
42:18
here are a couple of options. One was a medical
42:20
application of synchrotron radiation, first time
42:22
ever developed just in Stanford. And I said,
42:25
that sounds exciting. And he was excited. So
42:27
I got into a medical application that was coronary
42:30
angiography using synchrotron radiation. And
42:35
so that got me into medicine. And then I wanted
42:38
to do a postdoc in America, in the U .S. So I
42:41
checked out sites, 50 sites. Long story. In the
42:45
end, I worked with Harry Genant in San Francisco.
42:51
I worked there for seven years total. And he
42:54
was a great mentor. So I'm really... Very thankful
42:58
that he got me on the route to assess bone, to
43:01
assess osteoporosis, fracture risk. He unfortunately
43:04
passed away a few years ago, but he was a terrific
43:08
mentor, and we're trying to live up to the things
43:11
that he sort of trained us in. That's really
43:17
wonderful. Thank you for sharing your time with
43:21
us today, and thank you for starting up the ECTS
43:24
Academy. As the regular listeners will know,
43:27
It was through the ECTS Academy that we arranged
43:30
recording the podcast from ETS Live. And they've
43:32
been great supporters. And we've interviewed
43:35
some members over the last couple of years. So
43:37
thank you very much. Well, it was a wonderful
43:39
experience. Thank you for putting together this
43:42
podcast. I mean, not this very one, but a series.
43:46
And I'm really excited about this. And hopefully
43:50
we can work together also at the International
43:53
Federation of Musculoskeletal Research Society.
43:57
I think we would like to enhance the exposure
44:01
worldwide. And yeah, that's hopefully a bright
44:06
future. We would very much appreciate that. That's
44:08
great. And you've definitely done well enough
44:10
to get one of the bone -up mugs. Okay, this made
44:12
my day. Thank you, Klaus. Thank you very much.
44:17
Thank you very much. listeners thank you for
44:24
tuning in to episode one of bone up live ects
44:27
2025 we hope you enjoyed it watch out for episode
44:31
two which will be along soon bye now