The COMPOSE Studies: epidemiology, characteristics, management and outcomes of femoral periprosthetic fractures

Listen to Andrew Duckworth and Paul Baker discuss two papers published in the August 2022 issue of The Bone & Joint Journal.

Papers discussed:

Epidemiology and characteristics of femoral periprosthetic fractures: data from the characteristics, outcomes and management of periprosthetic fracture service evaluation (COMPOSE) cohort study

Management and outcomes of femoral periprosthetic fractures at the hip: data from the characteristics, outcomes and management of periprosthetic fracture service evaluation (COMPOSE) cohort study

Find out as soon as the next episode is live by following us on Twitter, Instagram, LinkedIn or Facebook!

Transcript

Copy Transcript Copied

[00:00:00] Welcome everyone to one of our BJJ podcasts for the month of August. I am Andrew Duckworth, and a warm welcome from your team here at The Bone & Joint Journal. As always, we'd like to start by thanking all of you for your continued comments and support as well as a big, thanks to our many authors and colleagues who have taken part so far.

We hope that you're continuing to enjoy our podcast and all the knowledge translation work delivered from your team here at the BJJ. Our podcasts continue to focus on papers published each each month here at the journal, as well as our accompanying special edition podcast series. So today for our monthly podcast, we're actually discussing two papers from the same study.

And I have the pleasure of being joined by Professor Paul Baker from the South Tees NHS Trust to discuss two papers from the COMPOSE study that he set up that have both been published in this month's edition of the journal. The first is looking at the epidemiology and characteristics of femoral periprosthetic fractures using data from the COMPOSE study.

And the second is related to the management and outcomes of femoral periprosthetic fractures around the hip. Again, using data from the COMPOSE study. Welcome Paul. It's great to have you with, with us and really appreciate your time today. Thank you [00:01:00] for the invite, really pleased to discuss these papers and the work that we've done as part of the COMPOSE study.

That's great, Paul, thank you. So the aims of the COMPOSE study, I I'm sure many of our listeners are aware, but were to describe the demographics of patients who sustain femoral periprosthetic fractures, looking at the epidemiology and characteristics of these injuries, as well as predictors of fracture types.

And in the second paper, you, you go on to specifically look at the management and associated outcomes of patients sustaining a femoral hip periprosthetic fracture. So, Paul, can you give us a brief introduction to the COMPOSE study, maybe how it, how it came about, and also, probably related to that, some of the background of the current literature in this area, particularly looking at how these, these type of injuries are increasing and, and I likely gonna con continue to increase.

Yes. So a COMPOSE came about really through, through a number of different drivers. Firstly as part of my role on the NHR HGA commissioning panel a research brief came to me looking to try to fund some research around the management of B2, [00:02:00] B3 or B type periprosthetic fractures around the hip.

But the, the, the brief really kind of was a bit aimless and, and, and was missing some key data. So we felt that there was an opportunity to really add to the, to the, to the knowledge in that area to help future researchers and help us understand this population a little bit more thoroughly. I was also looking for a project to work alongside our training research groups.

We' ve had a successful training research group in the Northeast for a little while. And looking to really sort of amalgamate the, the experience of their group with other research groups across the, the country and, and provide a platform for them to deliver a study. So they were, they were a couple of the, the key drivers.

I mean, the other thing is that, that what came out in the research brief through the NHR was that this is an area that I think is under researched. It's an area that is clinically important. And it's an area that we're seeing an increasing burden of disease in. So the timing is really right for us to, to do more research around periprosthetic fractures.

We've seen from recent [00:03:00] publications from the group from Imperial led by Professor Bottle using HES data that the incidence is increasing. We're gonna see an increased burden of disease in this area. They were showing about a 10 to 13% annual rise in the incidence of periprosthetic fractures over the period of their study using HES data between 2015 and 18.

And some of the systematic reviews in this area have also looked at increasing rates of periprosthetic fractures particularly related to age. So, you know, in some age groups, even the over eighties, if you look at registry data 39% of the revisions done for hip revisions in those, in those age groups are done for periprosthetic fractures.

So there are actually a significant proportion of the work that's being done. And we've seen similar things when we look at our own registry around knees as well as, as hips. So this is. Something that I think we're gonna see more and more of. And the caseload is only going to go up as we do more primaries, more revisions and how service it at a more elderly population.

So that was some of the drivers for it. [00:04:00] When we, when we looked at this, the, the literature. Is is very interesting. There are some systematic reviews, but the limitations of these is that they, the nature of systematic reviews makes them ask very specific questions. So what are the outcomes of a B2 B3 fracture, for example, is, is something that the group from, from Nottingham did you know, and that gives useful information, but only on that subtype of fractures around that specific joint replacement Similarly, when you look at registry data it only reports on revisions.

When you look at HES data, it includes inpatient episodes, but often uses the the ICD code. It's it's M 96.6, which is good, but it relates to prosthetic fractures around joint prosthesis, implants, and, and plates. So we're not necessarily getting a true picture of periprosthetic fractures around joint replacements.

And similarly, the NHFD, which is now starting to report, and it doesn't mandate the reporting of all femoral fractures around periprosthetic fractures. And I think the recent estimates were maybe that they've got [00:05:00] about 20% data capture. There's lots of data out there, but none of it really covers everything that we thought needed to be covered.

So we set up COMPOSE really to try to fill in a few gaps and provide a bit more information. So yeah, so that's, that's how it all came about. That's brilliant Paul, that's a really nice overview of, of do what the various sources of data that are out there about how they, they all add a bit, but not everything and it just all needs to be linked together.

And like you say, it's, it's like a, quite a heterogeneous group, isn't it? And so actually trying to do a, a robust systematic review of something is, is actually quite difficult to do, that's a really nice overview of how it came about. So if that moves us quite nicely on to, so the study itself and the design of it.

So for those who don't know, COMPOSE is a multicenter retrospective cohort study that followed a prospective study protocol and analysis plan and data were collected from a consecutive series of patients that presented to 27 participating hospitals in the UK with a new periprosthetic fracture through the year 2018 and from the entire cohort of

788 patients. There were a subset of [00:06:00] 720 who were eligible studying femoral periprosthetic fracture that were in the initial report. So, Paul, could you maybe give us just to start with a, sort of a brief overview of the inclusion and exclusion criteria who was included in and who was not. Yeah. So when we designed this, we had, we had input from a number of the specialist societies, and, and with them, knowing that we were doing this, they were all very keen that we tried to collect as much information on every disease area as possible.

So the, the inclusion criteria were actually very broad. It was any adult patient, aged over 18 with a new periprosthetic fracture. So it had to be within four weeks of presentation, we said, and it had to relate to either a primary or a revision hip, knee, ankle, shoulder, elbow, or wrist replacement. And it had to,

what we had stipulated was that they couldn't have concurrent injuries in other body zones. So we weren't looking for patients that had done these as part of polytrauma. Cause we thought they would impact on our outcomes of mortality and length of stay so on and so forth and would [00:07:00] confound the analysis.

So that's really where we were. And, and we, we. We left it to the, the sites to, to find these patients from their own coding systems, trauma records, theater logs, admission logs so on and so forth within the, the 27 study sites. And, and how were those sites chosen Paul? So the, the, the sites were, were driven by

the research group through CORNE T the, the, the registrar collaborative that we have in the Northeast. So I say what happened was we developed the protocol and the analysis plan in combination with CORNET, myself and some clinical experts at South Tees along with the statistical input from the University of York, and then having designed a, a data capture tool through REDCap with a, with a standardized electronic data system.

We left it to CORNET to approach other trainee research groups around the, the country to put in expressions of interest to be involved in the study. What we were really [00:08:00] looking for is to any, any trauma unit mixture of MTCs people doing general trauma. That could provide a minimum of, of 10 cases per site.

So it felt that they had an annual volume you know, of at least at least 10 cases a year to contribute to the, to the overall cohort study. And, and we got a good geographical spread. I mean, as you would probably expect we got a preponderance of sites in the Northeast but we also had sites in, in the Northwest a lot around Manchester, Leeds sites,

a lot in the Southwest, Plymouth, Exeter and, and, the peninsula, south coast Bournemouth around there and also a number around London, Oxford, Cambridge also contributed. So we had, we had a, we had a good mix a reasonable geographic spread. And you know, we, we saw that in terms of the volumes put in some sites put in.

10 15 other sites put in 60, 70, 80 cases because they were higher volume sites. So I think we got a, a reasonable representation of, of, of, of the geography of these injuries. No, yeah, absolutely. And I think that like, like we'll come onto, [00:09:00] maybe it's obviously important that because. These are obviously, you know, injuries that sometimes require specialist intervention or, or, or often do.

And, and then they sometimes patients need to be transferred. So I think it's quite good. Like you say, to get that spread of different centers. So if, if we go to sort of the data that was collected, what was your sort of, what was the basic sort of epidemiology and classification data, and also what outcome data did you collect?

Yeah. So we, we, we sat down and, and, and we looked at, you know, as with all of these studies you know, you can collect data, ad nauseum and, and, and you, you worry that data quality and, and data completeness will drop off. So, so what we tried to do was strike a pragmatic balance of looking for baseline, patient demographics, fracture characteristics, surgical characteristics, and outcome data.

So some of the baseline patient demographics for what you would expect, age, gender, place of residence, use of walking aids, social support, pre-injury, comorbidities. And we provided a list for, for patients to for the, the teams to select from usual things that were in that [00:10:00] heart disease, COPD, diabetes, et cetera.

Smoking, smoking status, MT S, and some information about their medication use. I have to say that some of that was, was better collected than others and certainly the, the, the data where we did see some data drop off was around the collection of things like smoking, alcohol consumption, anticoagulation because of the retrospective nature.

Yeah. The fracture characteristics were largely driven by the x-rays. So that looked at the date of fracture, the mechanism of fracture from the, from the, from the clinical notes, which joint it was around, whether it was a primary or revision joint, the fracture classification, and because we were looking at all joints we used the, the AO universal classification.

Yeah. Cause that gives us a schema for, for classifying joints all of the joints of the body. And then whether it was an open or closed injury. And then the surgical characteristics looked at the date of the surgery, grade of the lead surgeon and the surgical strategy employed. What, what what if, you know, fixation replacement, what you know, [00:11:00] use of other adjuncts to, to those, the duration of surgery and then any

free text information about operative findings and other information. So that's what we collected there. And then the postoperative data, the key things we were looking at really were, were length of stay, the postoperative complications before discharge, the readmission rate, mortality rate and information about discharge destination where we could.

I mean, it's a very comprehensive amount, amount of data you collect. And I said, you don't wanna overburden people, but it's good to get all that information int before we move on, just in terms of the outcome data was the, you mentioned about completeness of data earlier. Was that reasonably well collected?

So the outcome data was very good actually. And, and data completeness for the outcome data was about 95%. So, you know where, where we did see some, some drop off was, was some of the, the demographic data, but overall, the. Yeah, the, the, the outcome data was, was, was well collected and, and well entered by the, the patient by [00:12:00] the study members.

Yeah. Good. And in terms, before we move on to the, the results, Paul, it was, is there anything, particularly with the, obviously not going into too much detail, anything, particularly with the analysis you performed, that would be important for our listeners to know where anything you'd want to highlight.

Yeah. So I think all I would say is that this was set up to, to be a descriptive analysis. So we were very careful not to start making lots of comparisons in our data. Yeah. That was not the intention. The intention was to describe the population and to look at outcomes, you know, by group, without really trying to compare A with B, because there's lots of confounding.

And as you have already mentioned, there's lots of heterogeneity in these groups. So it's really, you know, comparing, you know, a B1 with a, B2, and a C is like comparing apples with oranges. Yeah. What we did do is we did set five, a priori analyses that we wanted to look at And that was, that was things like looking at were there any factors that predicted the type of fracture?

And to do that, we had to, we created a binary outcome [00:13:00] of, of those that we felt would usually be fixed and those would be replaced, revised. Predictors of operative and non-operative treatment, predictors of length of stay, predictors of 12-month mortality and predictors of 30-day readmission. Yeah. So, so they were the, the only analyses really we applied statistical tests to and we, that was all decided ahead of time so that we would, we would limit the number of analyses that we did.

Yeah, absolutely. So that's sort of prospective plan. Yeah. So if we move on to the, the results there, like, like we've said, there were 720 patients, 27 sites. And these had all sustained in the first sort report of femoral periprosthetic fracture. And that was around a hip arthroplasty, a knee arthroplasty or one dividing between the both and of those who were 539 that were femoral periprosthetic fractures around a hip arthroplasty, 151 around a knee, and

30 dividing the two. So just briefly, what, what did you sort of find Paul in terms of in terms of occurrence in the characteristics of the, the injury in terms of things like the location and the demographics? Yeah. So you know, we, what we found is the overall, [00:14:00] essentially the, the. Summary is the overwhelming fracture type was a femoral fracture, whether it be a femoral hip, or a femoral knee periprosthetic fracture.

I mean, 70% of our entire cohort were hip periprosthetic fractures. And of those 96% of them were in the femur, 20% were knee periprosthetic fractures of which 93% were in the femur. Hence the, you know, the focus down on that because all of the other groups, really the numbers were too small to draw any inferences and make any sensible conclusions about.

So that's, that's why the, the analysis has focused down as it has done in the first paper. And, you know, actually that's why the second paper really just leads on to the hip fractures because that was the greatest proportion and well, we could have described knee outcomes as well. There was so much to say about the hip outcomes that it seemed the right thing to focus on.

Yeah, definitely. And in terms of the, the, the characteristics that you saw, you know, the sort of generally for the, the epidemiology from the first paper, it's interesting, like, like you say, these are, these are very [00:15:00] comparable to our hip fracture patients, aren't they? Yes, they are. You know, the old, mostly female quite a, a significantly higher rate of female around the knees and the hips, but mostly female.

A lot of them have a dependency for their ADLs, use walking aids. They're a frail population of patients really that, that mirror our, our hip fracture, hip fracture patients. And if anything you know, the, the, the knee ones are slightly more frail than the hip ones, but you know, the difference is, is marginal, but, you know, that's certainly what the trend would suggest in the data.

Yeah. And in terms of, before we move on to sort of the second paper, the outcomes, were there any sort of factors, cause you did look at that associated with the type of periprosthetic fracture they were, they were gonna get. Yeah. So we, we looked here at again, we, we divided this into fractures that we thought or the data suggested would be more likely to be fixed,

and those that would like more likely to be revised. So we grouped AB1 and C type together. And then the, the B2 and B3 as the ones that were most likely revision. And what [00:16:00] we found is that being females seemed to favor an, you know, a fracture where the implant would remain fixed. And that cemented hips, there was a trend towards a higher or risk of B2, B3 fractures.

Although I have to say that I didn't quite reach statistical significance, but it certainly is in keeping with other literature that would suggest that. So I think that that's probably a, a, a valid finding. Yeah, absolutely. Absolutely. So if we move on to the second paper where there's, there's some really interesting outcome data, I think in terms of looking at how the the femoral, the, there were 539 femoral hip periprosthetic fractures, as you said.

And in, in, in this second paper of those just over three quarters were managed operatively and the rest nonoperatively, so Paul. Can you give us a brief overview of the management strategies used and, and, and, and how this maybe varied, a according to site. Yeah, so I think when talk about across our 27 sites that we had in the study, there, there was really wide variation in, in how these fractures were, were treated.

Now, I think what I would caveat this [00:17:00] to say is that we didn't really go into detail looking at the, the fracture types at each site because the numbers at each site were, were in most cases were quite small at the median being about 25. But what we did see is that those treated non-operatively, the, the rates varied from zero to 64%.

Those having been offered a revision ranged from zero to a hundred percent and similarly, those having a fixation range from zero to a hundred percent. So there's of obviously you know, despite the way that these are presenting and, and, and accepting that different fractures might have presented to different sites, there's often, obviously a lot of variation geographically, about how people are managing these fractures.

Yeah, no, it's I thought that was really interesting that, and do you, I know it's, it's sort of speculation, but is that, do you think that could be to do with expertise or did you have any feel for that at all? Why that there was such a variation there in practice really potentially. Yeah, I think we, we'd just be kind of trying to, I don't think we have any solid evidence about, about why that might be, but [00:18:00] I think probably expertise is, is one of the things, probably one of the things more than anything that's driving that.

I mean, we do know that, that there there's some evidence that say that people with the trauma background will want to fix things. And people with an arthroplasty background, like myself will want to revise things. So, you know, I, I think that that is a driver, but that's information that we didn't collect.

And I think it would just be, we, we are just kind of looking for reasons in the data, rather than having a, a robust explanation as to why that might be. Definitely, definitely. And, and if we move on, just in terms of the, the other, the, the outcomes that you looked at for these patients, like you say, just to remind our listeners it's complication, you looked at length of stay, you looked at 30- and 120-day operation rates.

And along with the 30-day and one-year mortalities, what, what were sort of the, the headline take home findings from those. Yeah. So, so the, the headline take home findings for those were that overall for the hip fractures, the the median length of stay was just over two weeks. That about 8% of people were [00:19:00] readmitted within 30 days of discharge after after the treatment of their hip periprosthetic fracture, 30-day mortality was

5.2%, and 12-month mortality was 21%. And, and that's very much in keeping with what we see for hip fractures. I, I think one of the things that we did find was that there was variation dependent upon the, the fracture classification. So the highest rates of mortality and the highest the longest length of stays were in people with B2, B3 fractures, with B3 fractures, having 12-month mortality of 38.5% which was significantly higher than, than you know,

many of the other fracture groups. So yeah quite a, quite a burden and quite a significant impact from their outcomes. Absolutely. Absolutely. I think that, I think, like you say, it's the, the, the contrast and, and the comparison really to hip fracture, patients, like you say, is, is, is quite, is quite incredible actually.

So if we sort of move on to putting it all into context, [00:20:00] you know, from my point view, you know, the strengths of the study and the two papers, you know as I know our listeners will, when they read it, will find out without question, in terms of so the size of the study, it's UK wide study, it's robust a really a good setup prospective analysis plan, et cetera.

You know, it's the largest multicenter evaluation of femoral peri, femoral periprosthetic fractures about, and provides a really comprehensive description of the patient and fracture characteristics of, of this group. And, and, and really the first study that has comprehensively shown that these patients are generally elderly and frail.

They're comparable to the hip fractures as we said, and, and all that information is very useful in, in planning services in the future, and also planning research in this area in the future. And I think with regards to the femoral hip fractures, femoral hip periprosthetic fractures, sorry that, you know, with the mortality re operation readmission rates, you've just mentioned, they are very comparable to the fracture patients, you know, with that overall 12-month mortality rate over 20%.

So. What do you feel are the, the key take home messages for our listeners? And I suppose more importantly, [00:21:00] caveating that with any limitations of the study. Yeah. So I mean, I think you know, I think the key takeaway messages as you described, these are a frail elderly population of patients. I think one of the things that, you know, that we we haven't mentioned as yet is the fact that these patients typically wait quite a while for surgery.

So we have best practice tariff for hip fractures that drives, you know, people getting into theater within 36 hours. But what we found was that that even, you know, the most straightforward fracture types are typically waiting a mean of a median, sorry of four days that, you know, the B3, B2, B3 that needs revision in most situations are often waiting five or six days for theater.

So, so. We, we, we have to bear that in mind. I think that that's a, that's a key finding. I think what I would say around that is that I think we need to be careful about not just rushing to people, to theater for these complex injuries. Just try and hit a target because I think that they are slightly more nuanced and more caveated than than, than hip [00:22:00] fractures.

And I think it, it is about pairing the right hospital, the right surgical team with the right injury to get the best outcome. Cause as you've mentioned, they have very high rates of mortality, re operation rates, readmission rates, and, and that all comes at a cost both to the patient and their family and also to the NHS managing these things.

There's, there's something to unpick there. In, in terms of, in terms of what we're doing, as we mentioned earlier, there is quite a wide variation in how these are being managed. And I think more work needs to be done in that. And that's, I think it suggests that we still don't know the best way to manage a lot of these different fracture types.

Yeah. I mean, one of the things that we did see. Was that lots of different surgical strategies being employed across all the different fracture types around the hip. And you know, I think the challenge is in how we drive research in this area, when it's a hetreogenic group of fractures often presented in a variety of different ways.

Yeah. And you know, I'm someone that treats lots of periprosthetic fractures and no two are really [00:23:00] alike, either the fracture's different or the patient is different or, you know, the circumstances around their admission are different. So, so it really is looking at how we can develop research in the, in the, in the context of individualized patient care.

Yeah. In terms of the. In terms of the limitations. You know, this was retrospective by design. We did initially think about doing a prospective study, but we realized to kind of get outcome data. We were gonna have to follow these patients up for quite a while. And I think what we wanted to really understand and the aims of the study, we felt that could be delivered through a retrospective design.

I think one of the keys with doing it retrospectively is being able to identify all the patients. And I guess one of the limitations is that we have no control over quality control over, over which sites identified what patients and whether they missed datas and data and how they coded to pick up these cases.

So yeah, that, that's a challenge. What I, what I will say around that is. [00:24:00] I was very pleased to see that we, we saw so many people that were treated non-operatively, which does make me think that people were, were picking up cases relatively robustly. Because you wouldn't necessarily expect those to be picked up if you were just looking at theater records, so on and so forth.

So I, I suspect that people did get the majority of them, although we have no way of verifying that. And then I guess the other things were around how these fractures were classified because you know, there is some difference of opinion in, in how you know, fractures, particularly the B types might be classified; is the implant loose, is it not?

 You know, how how'd you go about doing that? There's been some recent papers even in, in the BJJ around that from, from the Leeds group. Yeah. Around some of that. So So, so again, we, we didn't, we didn't, weren't able to validate any of the, the fractures and see how they've been classified. We, we, we took them as, as data entered, really.

Yeah. To try and do that, we did have a, a trainee and a consultant at every site. So there was senior input into, into that. Should there have been any, any doubt [00:25:00] about the fracture type. So I guess they're, they're, they're the key sort of limitations of, of the study. Yeah, no, no, I think that's very, very nicely sort of summarized, but I, I agree in, in many ways, I don't think in, in any way of sort of limited the findings, which I think are really important.

And I think we've already sort of mentioned how this potentially fits into the current literature and how it's already advanced it. So maybe just to finish up Paul, if. You know, you've already mentioned it, but you sort of, and, but you mentioned it also in the paper, you discussed this idea of creating a national or the need, really to create a national framework for data collection for this hetero heterogeneous group of patients.

And I just, maybe just your thoughts or musings on how that might look and how it, how could it potentially maybe fit into, into our existing infrastructure that's out there. So I think this is challenging. I think the one advantage that we have in this area is that we do have data collect already being collected through things like the NJR and the NHFD. I know, you know, through other work that I'm doing [00:26:00] with the major revision center and revision knees, that we've got a real challenge in how we collect data around those.

Cause it's a similar heterogeneous group of patients. So I think probably what we have to do is not try to reinvent the wheel and look to either look to adapt something that's probably already out there. I, I think probably the best way to do it would be to to try to increase data collection through the NHFD for all femoral fractures.

Yeah. Which I know is the, is the aim and then look at building in some appropriate outcome metrics and, and data around that. That's to me seems the, the most sensible way to do it. I guess the other way of, of doing it would be to create some sort of bespoke longitudinal cohort study, you know, akin to what has been done with hip fractures in the WHiTE studies to be able to collect a cohort data and then nest in trials within that.

I think one of the challenges is. As we've already discussed, this is such a heterogenic group of fractures. Yeah. To design decent comparative randomized studies, comparing different treatments [00:27:00] in this group is gonna be really challenging. And, you know, and some of the studies that you we've done in WHiTE aren't gonna be, you couldn't replicate those in this patient group.

I agree. And probably what we need to do is, is use something like NHFD as a, as a driver as they have done in hip fractures to drive up standards of care, look at pathways, reducing time to theater as a, as a, as a real sort of carrot and stick for, for service improvement for the benefit of patients in this area.

Yeah. Yeah. I, I totally agree Paul, I think your, your comments about, you know, how, how being naturally being part of NHFD or similar makes a lot of sense. And, and like you say, trying to design robust trials, certainly at the moment, I think would be very, very, very, very tricky. And, and it'd be difficult to ask the, the right questions or, and deliver those.

But, well, Paul, I, I, I think that's all we have time for today, but, but thank you so much for taking time to join us and, and congratulations really on an outstanding study that I like other recent studies we've published here at the journal you know, I think are real benchmarks for, for the UK nationwide collaborative research of this nature and, and data collection throughout the UK and have, [00:28:00] and, and without doubt of that is so much literature in this area.

So thanks so much, Paul it's, it was been great having you with us. Absolute pleasure. Thank you very much for your time. And to our listeners, we do hope you've enjoyed joining us, and we encourage you to share your thoughts and comments through social media in a light, feel free to tweet or post about anything we've chatted about here today.

And thanks again for joining us. Take care everyone.