BJJ Podcasts

The effect of antibiotic loaded bone cement on risk of revision, following hip and knee arthoplasty: a systematic review and meta-analysis

January 01, 2021 The Bone & Joint Journal Episode 34
BJJ Podcasts
The effect of antibiotic loaded bone cement on risk of revision, following hip and knee arthoplasty: a systematic review and meta-analysis
Show Notes Transcript

Listen to Andrew Duckworth &  Mike Whitehouse discuss the paper 'The effect of antibiotic loaded bone cement on risk of revision, following hip and knee arthoplasty: a systematic review and meta-analysis' published in the January 2021 issue of The Bone & Joint Journal.

Click here to read the paper

[00:00:00] Welcome everyone to our first BJJ Podcast of 2021. I am Andrew Duckworth and a happy new year and a warm welcome from your team here at The Bone & Joint Journal. We hope all of our readers and listeners have had an enjoyable festive period. And as always, we'd like to thank you for your continued comments and support as well as a big gratitude to our many authors and guest interviewers who take part.

Looking ahead to this year, which we hope will see us moving onwards and upwards from the COVID-19 pandemic and what was a very difficult 2020 for all of us, we aim to continue to deliver on a range of topics through our series, with our primary aim, to improve the accessibility and visibility of the studies we publish here at the journal for both you as our readers and listeners, as well as for our many authors. 

For this month's highlighted study, as you know, over the next 20 minutes or so we will be  covering a range of aspects of the chosen paper, emphasizing the important points of how the study has been put together, as well as some take home messages from the paper and how these potentially fit into your day-to-day clinical practices.

We also hope to give you a behind the scenes insight into how the authors have developed their study and give them an opportunity to put forward the key findings of their work. So today I [00:01:00] have the pleasure of being joined by my editorial board colleague here at the journal, Mr. Mike Whitehouse, from Bristol, to discuss their study entitled The effect of antibiotic loaded bone cement on risk of revision, following hip and knee arthoplasty: a systematic review and metaanalysis, which has been published in the January edition of the BJJ. Welcome Mike and a big thank you for taking the time to join us today. 

Thanks very much, looking forward to it. 

Lets just get straight onto the paper. So the aim of the study was to compare the effect of antibiotic loaded bone cement with plain bone cement on the overall revision rates for periprosthetic joint infection, following primary elective total hip and knee replacement.

So Mike, for our listeners, if you could just give us a brief introduction to the paper and some background to why PJI is such a devastating complication, as we all know. 

Yeah. So  we really wanted to kind of explore this from the point of view of their potential benefits and potential harms. And so obviously antibiotic loaded bone cement,  for us  in the UK, is  often the default option  and certainly within Europe, that tends to be the case. However, obviously we've got the [00:02:00] situation in the US where the FDA havent approved it as they consider it to be a kind of drug delivery system, rather than anything else. So obviously very conflicting positions around the world and strongly influenced by the fixation that people use now.

Now obviously we know from kind of work that we've done on the National Joint Registry data, that if we look at kind of time period specific effects then there's a higher risk of revision in the first three months following surgery for uncemented implants, but then they lower after that. So it kind of suggests potentially prophylactic benefit early on following surgery of  antibiotic loaded bone cement.

However, we do know that putting antibiotics into cement alters its mechanical properties. So there could be a benefit in terms of revision risk for prosthetic joint infection, but we also looked at overall revision risk to see if actually we are creating a problem in terms of aseptic loosening and later causes of failure as well.

Now, as you say, prosthetic joint infection is devastating. We've got a kind of patient group that work with us in Bristol on our infection [00:03:00] studies. And two of our patients have had cancer and infected joint replacements, and tell us the infected joint replacement was worse. It's devastating its a long-term treatment function, outcomes are very different to what we expect, and these are for patients who  went  for an operation where there's overwhelming chances of success and good outcome. So it really is something that we want to drive down on to minimize the risk of. 

Absolutely.  I think that's everybody's experience, isn't it? It's that dreaded fear of, the infection in the joint replacement and like you say, it's so rare, but so devastating  for the patient and it has such long-term effects for them moving forward. You know, they're never really right again after are they? It's just so impactful on their life. 

  Before we move on to the paper, Mike, you know, I think one of the important points you make in the introduction is about the outcome that you've used. And you chose revision, all course revision, rather than just infection as your outcome. And what was your sort of reasoning for this? Cause, you know, you say in the past, a lot of people do use just PJI as their primary outcome. 

Yeah,  so I think there are definite advantages [00:04:00] for using a relatively simple definition of the outcome for evidence synthesis, as it allows you to pull together kind of different studies that are being set up in different ways, as long as they have used a clear definition. So obviously revision is binary. It happens or it doesn't, as long as you're capturing it, then you can be confident that someone has had the revision or not. 

Definition of PJI is more controversial. Obviously there's a number of different systems out there. And part of the challenge that we face is when we're doing evidence synthesis, these studies are taken from over a prolonged period of time. And the definitions used within those classification systems have changed over that period of time. So if a study says they've used MSIS criteria you have to be careful to check which actual version of those have been used.

Talking to the patients when we kind of consider the important outcomes to them. When you kind of talk to them about, you know, what matters to them really it is their pain, their function and the number of operations they have. They don't really care whether they're still infected from [00:05:00] a kind of biochemical points of view. They care about the impact on them. And obviously, you know, a revision for infection is a major undertaking and people sometimes say, well, you know, are you sure it was infected? The fact is the surgeon has exposed their patient to the morbidity of this revision operation. And we don't think  the revisions that we do for infection are undertaken lightly. So I think, you know, I'm certainly confident in terms of the ability of surgeons to appropriately diagnose and manage these patients.

Yeah, it's interesting that isn't it?  I was speaking to Matt Costa a couple of months ago and he was saying the same thing about open fractures that we focus very much on infection as a classification system. But actually when you speak to the patients, it's actually their pain, their disability that they get, rather than if they've had an infection or not, it's actually the pain disability. And if they have to have further operations, that really is what impacts on them the most and what we really should be looking at. And that's interesting. 

So if we move on to sort of the meat of the paper, the methods, so obviously this was a systematic review with metaanalysis. It's performed, very high standard PRISMA and MOOSE guidelines [00:06:00] and the primary outcome, as we said, just to repeat, was revision rate for PJI and all cause revision between antibiotic bone cement and plain bone cement groups. And you consider them separately for total hip replacement and total knee replacement.

So, Mike, I suppose the key to any review, as you know, better than anybody, what were the eligibility criteria for one of the studies getting into the analysis and you mentioned joint registries and how did you determine whether they could be included or not? 

So I think, you know, as ever you set out to hopefully have a good number of definitive RCTs, you can bring together and get yourself a rock solid answer. Unfortunately, this is an area where there's a distinct lack of definitive RCTs in order to inform our practice. So there was one that we were able to include with just over 3000 patients in it. So, you know, we sent out fairly pragmatically to say, okay, well, in order to answer this, we're looking at a relatively rare complication and revisions with BGR is one the outcomes. Therefore we're likely to need big numbers to crack this particular nut as it were. And, you know, any difference because the effect size that [00:07:00] you get through something like antibiotic loaded bone cement, if it was that big in effect size, we'd already have the answer because it would be very apparent.

So we, you know, wanted to take a broad approach. Obviously we needed to assess the quality of the data, bringing into this to make sure that it was actually telling us what we thought it was telling us. So, observational studies obviously offer you the advantage of large sample sizes, potential for long-term follow-up. Registries are very useful source of data so research questions like this, but you have to be very careful to check actually, have they, the registry looked specifically at the exposure. Have they got the outcomes that you're interested in and how that pieces together? 

And obviously with the kind of range of follow-up available, the average follow-up available in a registry is often quite surprisingly short in comparison to the maximum follow-up just because registries increase in size as time goes on and accumulate cases.

So we included studies with other selected patients. So we didn't want these to be limited stations say just for diabetes and being selected. So this was unselected patients [00:08:00] who were undergoing primary joint replacement. So we did exclude traumatic indications, revision procedures. We're interested in primaries.

So studies had to report on revision rates of PJI all causes, and the exposure had to be antibiotic loaded bone cement versus plain bone cement with both exposure groups included rather than just studies that just had one. 


That was the reason why we ended up only being able to use Australian registry data because other registries report on the results, but not on those exposures.

So there were some studies that we found where other interventions that could influence the outcome rules included such as  allocation to administration of systemic antibiotics or not, but only one group, so we felt that might have an effect given that we are all strong believers that perioperative antibiotics are a good thing for these patients. So we excluded those studies because that could potentially influence the effects we saw.

When there were multiple studies that we found that reported on the same cohort, we preferred the one with the longest followup or the most comprehensive data. So if a longer term [00:09:00] follow-up didn't have the exposure and outcome data we needed would revert to the shorter one. But otherwise we went for the longer term follow up. 

All right. Great.  That's really nicely put Mike , cause obviously it's such an important way where you synthesize the data down and what actual data are we including, because one of the comments about meta-analysis is it's the data you put in, it's the quality of the data you put into that will then determine  the quality of the answer that you get out from it. So that's, that's very clear. 

So before we move on to how we, how we broke that down, we're going into a bit more detail about the studies you start with and what you got to, can you just give us a brief overview of the analysis relatively simply that you performed in the study.

So, yeah, I mean, as you say, the kind of critical thing is starting with the searching and, you know, the search are all listed there in the paper for people to take away and repeat and, you know, these are the things that we tend to do again in the future. And if anyone wants to rerun the recipe, hopefully everything is there to get the same answer.

Once we identified the studies, it's important to assess the quality of the studies. So Newcastle Ottawa Scale is the score that we prefer for observational studies. In brief, [00:10:00] it looks at the quality of non-randomized study designs, focusing on selection, comparability, and outcome reporting. And we preset before looking at the studies, the level at which we felt that data was at high enough quality to be included. Again, you know, we didn't want this to be a go have a look, have a look at the level. So we set that in advance.

 For RCTs, obviously easier process to the Cochrane risk of bias tool was used. And if this study was at high risk of bias in one or more domains, then it was excluded. 

In terms of the kind of analysis itself. I mean, one of the questions that kind of came up was use of adjusted or unadjusted data. Now typically studies would prefer the use of adjusted data. But there are issues with that, so when we're considering kind of pooling in our ability to bring these studies together, a lot of the observational studies didn't report adjusted analyses. The majority didn't. So obviously we would significantly limit the pool of studies we can include if we required adjusted analyses. Unadjusted estimates give you what we call the [00:11:00] population average effect. So they assume if the allocation is unbiased and then it will tell you about the effects of the intervention and obviously selections an issue in observational studies.

However,  you know,  we kind of think really that surgeons tend to use the type of cement they use. We don't think many surgeons actually say right, this patient I'm going to do plain bone cement because of their risk profile. This one we'll use antibiotic loaded. So we were less concerned actually about going through the selection than we would be in some other areas.

If you use adjusted estimates and those give you conditional estimates, which are typically the intervention effects that will be seen in groups with particular characteristics. So that's what then is adjusted for in the analysis, but of our knowledge of the risk factors that actually influence your risk of revision for infection has changed over the years. We might actually see different adjustment strategies depending on the age of the study.

So as we've deliberately sought out studies in unselected populations, excluded studies at high risk of bias and that [00:12:00] there were inconsistent adjustments and whether adjustment had been used at all. We actually made the a priori decision to use unadjusted data for the pooling within it. We did do a sensitivity analysis when there was data there between adjusted and unadjusted to see if our assumption was correct, and actually the sensitivity analysis on the couple of studies with adjusted data gave us the same answer. 

Yeah, no I saw that. That's really good. 

So if we look on , going into the meat, the results of the paper, initially you identified 2,747 records, and you ended up including nine studies and one registry data, the Australian data, which you mentioned. So that was, you had five study data for hip replacement and six study data for a knee replacement, the two sets of data from the registry making up to 11.  And so can you just talk through for us listeners, what was the main reason for removing studies, more than anything, as you got down to that final 11?

Yeah, well, as with any search strategy you want the first number you come up with to be something that terrifies you, and then you start with your criteria for the final selection. So [00:13:00] actually just removing duplicates removed just over a thousand from that initial list. 

Obviously sequential approach to the kind of screening starting out with title and abstracts, and then more detailed for those that are more likely to be included. And we got down to 119, basically on that initial screen of the kind of title and abstract, basically by removing those that were nonclinical studies that would non-comparative studies, that were based on revision populations, were reporting outcomes for cement spaces, rather than cemented fixation or microbiology data. So that quite rapidly cut things back then until we're left with, as you say, that one Australia registry study that could be included in both and the final numbers will be hips and knees groups.

And in terms of the you've touched on it already, Mike, but what was the overall quality of data like, as a whole?

 So, I mean, there weren't many studies we had to exclude on the basis of their data alone once those criteria had been applied. And I think probably because we're [00:14:00] fairly constrained in terms of the requirements for it being a comparative study for the outcomes it was reporting but actually we probably eliminated a lot of low-quality studies already  which is kind of what you aim for really. 

And I think in terms of the kind of types of studies in there, so if you think about the kind of hip studies, the Australian, I mean, normally the registry analyses dominate in terms of population size, but in the hips analysis, The Australian registry study provided just under 13,500 cases.

And those cohort studies, just over 25,000 patients on all calls revision only from France. One cohort study of just under 83,000,  just providing data on PJI revision rate from Norway. And then the other two cohort studies, which contributed just over 250,000 from the NJR from Norway, provided data on both outcomes. 

For these slightly different, the Australian registry, provided 98,000 cases. There was one RCT from Spain with just under 3000 cases. Two cohort [00:15:00] studies providing data on PJR revision only. They gave just over 100,000 cases from Finland in New Zealand, and then two cohorts studies providing data on PJI and all cause revision outcome  just over 460,000 that were from Canada.

Perfect. Perfect. Yeah. So  big studies overall. And like to say the you've sort of whittled down the quality ones with that review process. So looking at the overall numbers, and obviously these aren't the total examined  in some ways, because it's something that we had core revision and some had PJI rates, but you had a total of  just almost 372,000 total hip replacements and just over 671,000 total knee replacements.

And what was the follow-up, Mike, for those? Was that pretty good generally? 

Yeah. Again, I think this is where the difficulty comes in with the interpretation, the information we're presented with in these studies. So, if you look at the kind of follow-up ranges, so hips, the follow-up range from zero to 20 years. Which is a nice, non informative way to express the follow-up [00:16:00] range. For PJI revision, zero  to 16, for knees zero to 13 years for both. And the average is variant, how people calculate those averages and numbers at risk. But typically you're looking at actually fairly short term followup data say the overall average was something like three to four years.

Yeah. Yeah. Okay. And so if we get to the key findings, really, what did you find? What were the revision rates in all cause PJI and the effect of the antibiotic loaded bone cement for both hip and knee replacements. So if we looked at the revision for PJI and hips, the rate in the  antibiotic loaded bone cement was 0.36% versus 0.71% for the plain bone cement. So relative risk of 0.66, and the confidence intervals were under one. So that favored the antibiotic loaded bone cement. 

The all-cause revision in hips, the rate in the antibiotic loaded cement was 1.57%. And for the plain bone cement was 5.1%, relative risk of 0.62, but this time broad confidence [00:17:00] 0.35 to 1.09. And really, you know, there's one study, the Angus arts study from 2006 with the relative risk of 0.32, fairly narrow confidence intervals that distort that a little bit. As I say, we're confident that actually with the overall pooling that you know, those confidence intervals cross one therefore difference was insignificant.

Moving on to knee. So for the revisions for PJI in knees. The rate in the antibiotic loaded cement was 0.42%. And for the plain bone, cement was 0.35%. Relative  risk is 0.92 or risk ratio, 0.92. So 0.59 to 1.45 broadly overlapping the null. And for all cause revision rates in these, the rates of the antibiotic bone cement was 1.66%.  And so the plain bone cement was 2.46% risk ratio of 0.73, and again overlaps the null. So no significant difference there. 

Great. So, you know, if we looking at that area, sort of bringing that together. So the overall pooled analyses have shown that, you know, antibiotic loaded bone [00:18:00] cement, appears to be protective against revision for PJI in primary hip replacement compared to plain bone cement. But for all the other analysis, there was no real significant seen to antibiotic loaded bone cement.  Is that reasonable? 

Yes, exactly that. 

So if we move on to that finding, I mean, I think it's really interesting in terms of, you know, if we look at this study as a whole, you know, the strengths of it are without question, it's the largest meta-analysis that's been performed, looking at this and looking at both total hip replacement and total knee replacement. But what do you feel are the key take home messages from that, how do we place that into our normal day-to-day practice? And I suppose as well, the caveat in that with any potential limitations you think there are to the data that's been included. 

Yeah. So I think, you know, this primarily is driven by observational,  not RCT data. So, you know, that really described this as a kind of a clear association and I would avoid using that kind of causation type language when we consider this even in the area where there's benefit. So I think there may be advantage in lower overall revision [00:19:00] rates of PJI for antibiotic loaded bone cement.  primary THR.  However, we have looked at overall rates and previous work we've done on NJR data suggests that the effects can be time periods specific. So given the relatively short-term nature of the follow-up period, it could be the benefit we've seen in hits will actually wane with time and come back to null. So, we don't have a kind of definitive answer on that. Is this an early protective effect we're getting, but then goes away. 

Reassuringly, and really this is the kind of thing I really wanted to get out of it. There was no deleterious effects on overall revision rates seen. So, you know, confidence actually probably the practice that is happening at the moment. And so we've got better data or data that will inform practice more is safe. So we don't need to kind of, you know, change, practice, change what we're doing on the basis as this.

So, as I say, you know, there could be selection effects going on between the antibiotic loaded bone cement and the plain bone cement. But, you know, at the individual surgeon level, when selecting for [00:20:00] interventions, I think there's very few people out there that are saying right I'll use plain bone cement for this one, or I'll get the antibiotic cement out for this patient because they're diabetic or what have you. And again, you know, we prefer the unselected population. So actually for once selection in the observational data isn't as much of the problem as it can be for quite a lot of studies during observational data. 

One of these unique situations, like you say where I dont I think surgeons are doing that selection at all. I think it just, everybody gets it. Don't they generally now, I think is one of those things. 

So just in terms of looking at the data that's already out there, then how does this fit with previous studies in the area? 

So, we've conducted the previous review that really focused on implant fixation, rather than specifically breaking down into antibiotic loaded and plain bone cement, although we did include that within the analysis, in that other review we use broader definitions of PJI. So we actually included non revision procedures, surgical site infections for that kind of more global coverage. Probably making sure you're not missing any, but also [00:21:00] certainly pulling in cases that actually aren't prosthetic joint infection. So it allowed a more comparative consideration, but reassuringly the findings were concordance with the review here. So, you know, I don't think this is a problem of the definitions of the outcomes that we are using. 

So the majority of pieces of evidence synthesis in the area have shown similar  findings to what we're saying here. There was one big US  study back in 2008  from *inaudible* group that found significant effect in favour of antibiotic loaded bone cement both for the revisions PJI and full causes. However, there were issues with the kind of the way the data was handled within that. So they didn't require a control groups. So they would take observational studies, just reporting on one exposure, not the other, which was different to how we'd handle things.

They also preferred short-term data when there were multiple reports. And I think probably because they thought that the accuracy of the outcome or the indication for the outcome might be better in those studies. They did [00:22:00] mix primaries revisions and trauma procedures, and I've got quite strong feeling, you know, those are different populations and really shouldn't be pulled together.

So I think, you know, there were reasons for why we might have come to a different conclusion, on the basis of what they have. There have been three RCTs by cheese group, two werent eligible for inclusion in this cause one was just on diabetic patients. One was on revisions only. The study of primaries they did, which we looked at to see if we could include, was small, 340, but that wasn't the kind of reason to exclude. Potentially not really reflective of our contemporary practice here in the UK, because they didn't use clean air ventilation systems in their theaters. Right. And so, you know, wouldn't be considered a standard of care but that's more controversial nowadays than it used to be. And they didn't have any blinding. So it was a high risk of bias, which was the reason why it wasn't included. And interestingly, the findings of that RCT werent concordant with the Spanish RCT, which we were able to include, which was higher quality and [00:23:00] larger numbers.

Yeah. No, I think that's right. But like you say, it does seem to very much fit with the more, the more recent studies that have been done in this area. And I think just to finish off Mike, you sort of touched on it already and I think it's a really interesting point.  I think, you know, in terms of the implications moving forward, you've already said there's no obvious negative effects or negative reasons for using antibiotic loaded bone cement.

And I suppose the question I have for you though, isn't it? If you think about, you know, going back to the beginning when we said it's  such a devastating complication,  if you ask the surgeon, how many people would you be willing to give the antibiotic loaded bone cement to, to prevent one? That number would be astronomically high, I suspect for most people. And so I think, like you said, there's no reason really to change our practice now, but I suppose  do you think we need a clinical trial in this area, is that feasible? Would it just be too large to give us the answer? And do we really need one or do we have enough data do you think out there?

 Well, certainly on the basis of this review, I haven't changed my practice. So I'm still using antibiotic loaded bone cement for all my patients. I think there [00:24:00] is justification for doing a study there. And I think really the areas most justified in would be the knees because, you know, we, haven't got a clear evidence of benefit there. And I think actually, you know, the final call, if we did have that data is very likely to be driven by the health economic implications for it. So obviously if you've got no benefit and you're spending more to do it, it's very unlikely to come out in the wash as being a thing that would be recommended to be done.


I think we could do the study. There's no two ways about it. It needs big numbers. It needs about 22 and a half thousand patients on the basis of the effect sizes we've seen in the study. So that does go with kind of, you know, the only way that really feasibly we're going to deliver this with the funding available to us for doing studies is in an efficient embedded trial design. That does mean that we have to be very confident that the kind of mechanism that we're using to collect the data such as registry recently collected healthcare data is the outcome that you're really interested in. 

So for something like [00:25:00] revision,  for the indications, I think we could be reasonably confident. But even then, even if we had the most efficient trial design ever equipoise may well be a challenge. So I think it needs a very decent scoping work upfront to find out actually, what proportion of surgeons would be happy. Would they be selective in who they were willing to include in the study? Could you use something like the cluster design where actually you randomize at a unit level rather than the surgeon level to stop those kinds of crossovers and the complexities. So I think it'd be desirable and doable. I think it would be tricky and expensive. 

Well, Mike that's a good place for us to finish up on. So thanks so much for taking the time to join us and congratulations on an excellent study that is without doubt, an invaluable addition to the literature, and I'm sure will lead to much discussion  and ongoing thought for us all. So thanks for joining us, Mike. That was great. 

No problem. Thank you, Andrew.

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 and a like. Feel free to post or tweet about anything we've discussed here today. And thanks again for joining us. Stay [00:26:00] safe, everyone.