Listen to Andrew Duckworth, Liam Yapp, Chloe Scott and Nick Clement discuss the paper 'Long-term mortality rates and associated risk factors following primary and revision knee arthroplasty: 107,121 patients from the Scottish Arthroplasty Project' published in the January 2022 issue of The Bone & Joint Journal.
Click here to read the paper.
Welcome everyone to our first BJJ podcast of 2022 and for the month of January, I am Andrew Duckworth and a happy new year and warm welcome from your team here at the Bone and Joint Journal. We hope you've all had an enjoyable festive period with your family and friends and as we return to some normality over the next year.
As always, we'd also like to thank you all for your continued comments and support as well as a big, thanks to our many awesome colleagues who take part in our series. Along with our monthly podcasts, highlighting some of the papers published here at the journal, we will also be continuing our special edition podcasts with the specialty editors here at the journal, as well as some further episodes from our insights from the U S series.
So we do encourage you to look out for these over the upcoming months. So moving to today, I have the pleasure of being joined by three of my colleagues from here in Edinburgh to discuss their paper entitled 'Long-term mortality rates and associated risk factors, following primary and revision knee arthroplasty', which has been published in the January edition of the BJJ.
Firstly I would like to welcome one of our superstar trainees here in Edinburgh, Liam Yapp, who is the lead author on the paper. Welcome Liam and a big thank you for taking the time to join us today. Hi, Andrew, thanks [00:01:00] very much for the invite to speak today. Liam is joined by two of his senior authors for the paper.
And two of my exceptional friends and colleagues here at Edinburgh, Chloe Scott, and Nick Clement. Chloe, It's great to have you back with us. No thank you very much for inviting me again.
And Nick. It's always, always great to speak to you. It's always nice to be invited back after being late last time. Thank you. So Chloe, if I, if I could maybe start with you and the, and the study, the aim of your study, which utilized data on over a 107,000 patients from the Scottish Arthroscopy Project. Also determined the longterm term mortality rate, following primary and revision knee arthroplasty as well as to identify factors associated with this.
So if you could maybe give us a brief introduction to the paper and some background as to why it's so important to consider the risk of death for these patients in terms of the influence on the risk of revision. Of course thanks. What a great question to start with. So one of the key metrics of success in orthopedics, and then arthroplasty in particular is obviously implant survival.
And we typically assess this [00:02:00] using the Kaplan-Meier method. So this method examines the cumulative survival of an implant, but it only considers a single event, which in the context of knee arthroplasty is implant revision. So in essence, it estimates the probability of a revision at a certain time point, assuming that the patients cannot die.
However, when we perform long-term survival analysis, there are actually four which contribute to our results and how we interpret them, including intact implants in alive patients, patients who are lost to follow up, implant failures, obviously, which is what we're looking for or patient death. So alternatively speaking, the death of the patient is a competing risk [audio out] the implant and Kaplan-Meier analysis deals with this by treating either lost or deceased patients for so-called censor data.
And by doing that, it assumes that the risk of implant [00:03:00] revision is independent of the risk of, of patient death. So when it does that, it theoretically overestimates the cumulative risk of revision when there are competing risks. And actually this has been demonstrated nicely in a recent systematic review, comparing survivorship from Kaplan-Meier analysis to an alternative competing risks model of survival.
So we're all used to looking at implants viable data from the National Joint Registry, for example, and from this [audio out] using NJR data, we know that the risk of revision of knee arthroplasty in particular is higher in younger patients. But again, the NJR doesn't include mortality data, and many older patients will unfortunately have died before their implant could fail.
Conversely, patients expected to live several decades after their arthroplasty [audio out] shorter lifespans. So the [00:04:00] longer your life expectancy is the greater your lifetime risk of revision is, but interestingly, the previous estimates of lifetime risk of revision, have used mortality rates from the general population, assuming these to be the same as the arthroplasty population.
And this is actually quite a big assumption which our study that we're talking about today goes on to show is probably incorrect. Yeah, thanks for, that's a really nice overview of actually, of, of what was sort of, sort of really the aim of the paper and why the, the, I suppose the paper was sort of set up isn't it.
And Nick, if I can come to you next and before, maybe we move on to some of the details of the study and sort of following on nicely from what Chloe said, can you give the listeners an idea a bit more either about the current literature in the area? And how and how, you know, as Chloe says we've used general population, mortality rates in the past and how that affects the data we see on things like lifetime risk of revision for knee replacements and how it's been considered before.
I think probably the best example is kind of the mark off models that are often [00:05:00] used to predict cost-effective analysis and along those lines, but just to touch on what Chloe said earlier on, it wasn't until just very recently, I started to read the NJR in a bit more depth, then realized all the figures thats produced is risk of revision.
So their Kaplan-Meier curves are sometimes censored it for lost to followup, but they're not censored for, for death. So, so all the papers, like I've just done one quite recently with the Newcastle group, and so it's a probability of revision that's what's presented. So potentially you might have a probability revision of 10% at 10 years, but actually you've lost 25% of your population as Liam has just shown.
So actually you [audio out] your survivorship is far less. You've only got 75 patients at 10 years or 75% at 10 years. It will be 10 divided by 75, which obviously is far a different percentage as opposed to 10 divided by 100. Yeah. And that's, the penny has just dropped with me with the NJR [00:06:00] data just recently, but, but just to go back your original question, I'm so sorry, Andrew.
I got sidetracked there. I think one of the big things for me in that, that the things I've done in the past with mark off models and, and randomized trials, try and do that. Now after two years of randomized trial, robotic versus something else. So one knee versus another, and some health economist comes along and models data for the next, until the patient dies.
And the average life is 15 years after, after a hip or knee replacement for the average seventy year old. So relation data and as Liam's shown, for the first 10 years is far less. So actually, if you're doing the the cost economic model, actually your bang for your bucks even more because you're going to get a better outcome.
Also your revision rate might go up ever so slightly because the population surviving might go up. So that would have to say, I think there's great data for that. And certainly for, for me going forward, maybe doing similar sort of mark off models and studies like that, I think this data is essential. That's really interesting.
And I think it's a really, that's [00:07:00] really great from both of you just explaining, you know, what it really puts into context, the reason why this study is important and why, why you all three of you decided to do it. So if we move on to the study design and I'll probably come to you Liam next, if that's okay,
this was a big data retrospective study. And utilized data from the SAP. As we said, it's 107,121 patients, and they all underwent primary or revision, knee replacement. And it was over a period from the 1st of January, 1998 to the end of December, 2019. So Liam, if you could maybe give us a brief overview for for listeners who are not aware of it, a brief overview of the SAP, you know, an idea about the robustness of the data, I suppose it collects and, and how the data for this study was
sourced. So the Scottish Arthoplasty Project' s a national audit of all arthroplasty procedures undertaken in Scotland. It was set up in [audio out] it's produced annual reports sinceabout 2001. It's overseen by a steering committee, comprised of surgeons, anaethetists, data analysts, managers, and patient representatives.
And it's [00:08:00] wholly funded by the Scottish government and is overseen by NHS Public Health, Scotland. The data that it receives is primarily obtained from two sources. One is the Scottish Morbidity Records 01, which is an electronic health record. And it has a permanent linkage to the national records of Scotland, NHS Central Register, which has a record of all births and deaths which occur in Scotland.
And they are able to achieve this through utilizing the unique identifier, which is assigned to every patient in Scotland, from birth, which is called the Community Health Index number or the Chi number. So in essence, any patient who undergoes a total joint replacement in Scotland is able to be accurately followed up for the duration of their lifespan.
And in terms of data quality, well, routinely collected data sources are not perfect. I think thats well established and a fair, fair thing to say. So NHS Public Health, Scotland actually regularly perform internal audit and quality assurance [00:09:00] processes for all the data sets, which they hou es and the most recent audit data that we have
suggests that diagnostic accuracy for common comorbidities we're interested in is approximately 96% and procedural coding such as primary or revision procedures of about 94%. So it's not perfect, but it's, it's pretty good for what, for what it is. Yeah, no, I agree with you Liam. I think that's really good.
It sort of highlights, you know, the robustness of the data there. And I think like, as you say, highlighting the uniqueness of the CHI and what that provides us with, there's not many places that really in the world that could, that can do that in terms of the uniqueness of the CHI and linking those, those data sets.
So if we move on to the study just briefly, you know, what were your inclusion criteria for this study and sort of how you did you deal with patients who had gone the gone? Cause they were quite long study periods, maybe multiple procedures. So the inclusion criteria was any patient who had unilateral or bilateral simultaneous, or the first knee of staged bilateral [00:10:00] procedures.
We considered that the index procedure. Specifically to avoid counting death and same patient twice, because we knew that that would negatively influence the estimated risk of death in those patients. Yeah. And, and obviously I always ask people what their primary outcome measure is, but it's obviously mortality for this, but how you sort of define that.
And maybe just before we move on to the results, in terms of, you know, revision and re and how that's defined in the SAP, what can you just give the listeners an idea about there? So our primary outcome as you state was the patient mortality at any time point, following primary revision and knee arthroplasty.
And I think it was important to mention the definition of revision that the SAP uses. So they state that a revision procedure should be considered to be a permanent removal or exchange of knee arthroplasty components. So specifically we did not consider a secondary patellar re urfacing as a revision for the study, and that's obviously quite different to how [00:11:00] other registries, like the NJR would consider that procedure.
So I think it is kind of is important to note that. I absolutely, I totally agree. And before we sort of, that will lead us on into the results, but just before we do that, Nick, if I could just briefly come back to yourself. what, if you just, cause obviously it's something that you talk about in this study is about something called the standardized mortality ratio.
Could you just give our listeners a brief overview of that? And then what analyses were performed in relation to the outcomes that you looked at? So standardized mortality ratio is used quite a lot. Most people I'm sure you've probably come across it in like fractured neck of femurs, probably as opposed to within, arthroplasty.
So standardised mortality shows, you need to know what it's standardized to of course. So that will be in your methods somewhere that people have to sift that out, most will be age and sex. But it can also be ethnicity as well. And even some of the cardiac studies and diabetic studies, you can actually match for that as well, if you're really going for it.
So for, for example, the average 80 year old in Scotland or 10% of [00:12:00] 80 year olds. If you get 100, 10% are dead every year. So as opposed to a fractured neck of femur, where 30% are dead every year. So 30 divided by 10 that's that's three, so a standardized mortality ratio of three. And that's generally what fractured neck of femurs are approximately, humerous.
i roundabout 2, and distal radius is about one and a half. So that's been done before. From Edinburgh, actually, Mike Robinson's study. So that's standardised mortality ratio. It's just, so you have some kind, that just because somebody gives you a mortality rate, you don't know whether it's high or low. So just gives you a ratio.
And obviously if it's, if it's greater than one it's it's high and if it's less than one its low, and Liam used mortality figures from the Scottish office on survivorship or predicted survival according to age and sex. And he does some wonderful stuff that I have no idea how to do and came up with some lovely figures and tables.
I could not agree more. They are lovely figures and tables. So if we move on to the results then Liam, and [00:13:00] that leads us in nicely. So in the study there were just over 98,000 patients who underwent a primary knee replacement, just over 8,000 underwent a revision. So the median age at surgery was 68 years and the median followup for the group was just over seven years.
So Liam, if we come back to you again, what do we find in terms of your primary outcome in terms of the primary knee replacement cohorts, and what factors did you find that were associated with an increased risk of mortality in that patient group? So for the primary knee cohort, we had 27,474 deaths that occurred within the period of followup.
So that was approximately 27% of the cohort. And when we performed K plan survivorship estimates. We found that about 73% of the cohort were actually alive at 10 years. And of those who had followup up to 20 years, that dropped down to 30%. So quite a significant drop-off. And that's where we get the figure of about a third or a quarter of patients have [00:14:00] died by 10 years.
And so that gave us an overall SMR of about 0.74. So as Nick already alluded to SMR over one would suggest an increased risk of mortality compared to the general population. Whereas below one would suggest a survival advantage or, or, or benefits. So these patients are living longer than expected compared to their
age and sex matched peers. And we found that that actually persisted for about 12 years following primary knee arthroplasty, before settling to rates consistent with the general population. And when we looked at what factors may be additionally associated with mortality within the first 10 years, we found obviously increasing age at the time of surgery, which was
perhaps not that surprising, male sex, a diagnosis of inflammatory polyarthropathy, greater socioeconomic deprivation, revision for infections and greater number of medical comorbidities. Interestingly, we didn't find any [00:15:00] association with aseptic revision, which I guess we can talk about later on.
Yeah, no, absolutely. And in terms of looking at, is that's for the primary cohort, what did you find for the revision cohort? How was that similar or different? So the findings for the revision knee cohort were very similar frankly. So there was 2,611 deaths. So approximately 31% of the patients died within that time period.
And again, the Kaplan-Meier estimates up to 10 years or 68 8% up to 20 years, about 31%. So very similar in terms of drop-off in both cohorts. And the overall SMR was 0.83. So again, that suggests an overall survival advantage, which persists in that group up to eight years, actually, before normalizing to ra t es consistent with the general population.
And again, when we looked at the factors associated with mortality in that group, they were essentially the same. So older age at surgery, male sex, diagnosis of inflammatory [00:16:00] polyarthropathy. Increasing socio economic deprivation, re revision for periprosthetic infection, a greater number of medical comorbidities, but not aseptic revision.
Okay, that's great, Liam. I think that's a really nice overview of how, like you say, how the findings of the two groups are relatively similar. So if you move on to the implication, I want to take a bit of time sort of discussing it. You know, I mean the strengths of the study, you know, without question, in terms of the size of the data, you know, it's using nationally linked data, very robust analysis performed without question.
And it's shown that the SMR for patients undergoing primary revision knee replacement was lower than that, of the general population and remained so for several years, like you say although, you know, one in four undergoing primary joint replacement died within 10 years, and that was one in three for revision. So Chloe, if I could maybe come back to yourself what do you feel are really the key take home clinical messages of the study, considering any potential limitations of, of the data.
So I think for me, the key take home message is that for us to accurately [00:17:00] interpret implant survival or [audio out] we really need to know the life expectancy of our patients and our specific arthoplasty patients, as opposed to the general population. And using this robust national dataset, we've shown that mortality is significantly less for knee arthroplasty patients than for the general population.
17 to 26% less. In fact, so this isn't one of these small differences that is statistically significant simply because it's a big data paper. They appear to be significant differences. These patients have significantly lower mortality. Despite this, however by 10 years, a quarter of patients after.
[audio out ] revision have died. And this is associated with inflammatory arthropathy infection and deprivation which we found to all be important associations with increased mortality, in addition to what you would expect which was, you know, it would be expected that mortality would be associated with increasing age [00:18:00] and with increasing numbers of co-morbidities.
But I think there's certainly new and novel information there. Now, obviously any retrospective study has limitations. In that we can only comment on association and not causation. So we don't know why the mortality rates are lower in the knee arthroplasty patients. We can hypothesize. And unfortunately as much as I would like to, we can't say that knee replacements make you live longer.
Unfortunately, we don't have the same data on non-operatively managed degenerative, knee patients. So we don't really, we don't know how having knee osteoarthritis for example, affects your, your mortality. We simply know the association between mortality and patients who are fit enough to undergo joint replacement surgery.
And one of the main limitations, I think, of the SAP dataset is unfortunately we don't have BMI data which would be particularly useful and interesting, especially with a study like this, because. [00:19:00] Studies from other populations. So in Australia they found that BMI significantly affects life expectancy across the general population, not across the arthroplasty population, but you would expect it to do so among the arthroplasty patients as well.
Although other big data studies have shown that interestingly in BMI [audio out] revision but it would be really good to, to have that data. And that's something that we're that we're missing. Yeah, no, absolutely. I think it's a really nice overview Chloe, of like you said, the findings and just sort of the caveat ing, I'm glad you brought up the point about knee replacement potentially making us live longer.
But I mean, on a serious point, I think I'll maybe ask you, and then I'll I'll ask Nick as well, you know you know, In terms of the hypothesis that, you know, there is an argument there, you're reversing the effects of knee arthritis in terms of pain. You're improving quality of life.
You're probably stopping progression of frailty and other factors. And then I suppose the flip side that you're arguing maybe we are picking the more fitter patients in [00:20:00] that patient group to operate on But there is something potentially there, and maybe, you know, has bigger implications for the fact that, of the state that we're a lot of us are in that where we're doing so little or theres such a long wait to get to joint replacements, do you not think?
I mean, I'm sure that the, the the effects that it has on mobility and frailty, you know, some patients that remain untreated end up being house bound and they have a general decline in their, in their overall health as a consequence of that. And I'm sure that frailty plays a part. I'm sure that also, in addition to the fitter patients being selected to be referred and also the fitter patients being able to do more and therefore be more limited by their degenerative joint disease.
I think, well, obviously when you're considering someone for arthoplasty, they're basically getting a full general health check in their, in their sixties and seventies. And it's not unusual to pick up things like cardiac arrythmia on a, on an [00:21:00] ECG that were asymptomatic before that maybe you get highlighted in a pre-assessment clinic and therefore treated before the patient has their arthroplasty or cancers that have picked up because of
anemias is that, that are identified. So I'm sure that only plays a small role, but I think generally having a health check at that, at that age where you get these investigations done probably does pick up some comorbidity that would otherwise go unrecognized until it was more severe. No, absolutely.
I agree. Nick, anything you sort of add to that at all. The second paragraph of the discussion that Liam wrote summarizes that really well kind of, it kind of highlights the 37% increased mortality just associated with knee osteoarthritis, which is probably due to dysfunction, increasing frailty, cardiac problems. Just because you're not stressing the heart as much as you should maybe not getting out.
Now you can just imagine it can't you, everything just deteriorates, just because you're struggling. And then potentially if you give somebody a knee [00:22:00] replacement, you reverse all that. And, and, and, and, and, and. In the same boat as Chloe, and I'd love to say knee replacements, save lives and make us live longer, but it may well be in this patient group that that might be what's happening, that it might well be that we are reversing this kind of de-functioning of the patient.
And we would probably say that over the next couple of years, I'm sure we'll have a two year waiting lists and worsening quality of life. Preoperative. And more peri-operative [audio out]. So I think that's one of the [audio out] potentially during those first years, your, your, your mortality issue. I wonder whether, if you shift those first years up, because obviously they're not getting the knee replacement when they should, whether whether their overall benefit came from a mortality point of view may well come closer to the normal population because
they're not getting it when they should. how would you interpret the findings associated with, you know, the factors associated with increased mortality and particularly [00:23:00] things like periprosthetic joint infection or revision for that?
I think all those are common sense things that you'd probably guess in an MCQ wouldn't you, if your mortality was bound to be raised. If you having, if your having a revision for infection it might be a two-stage defunction in between. You've got an organism on board that may well kill you at some point in time.
And there is certainly plenty of literature from an inflammatory point of view, even outwith how many joint replacements given to people with rheumatoid arthritis, for example, that died earlier, or certainly used to, I don't know what it's like with DMARDs now, but certainly their mortality is increased, but things like they increase in social deprivation.
I think thats relatively new. I can't remember anything else in the discussion or any other literature off the top of my head that actually showed that, certainly a paper that I did previously with the Fife group. We, we showed that BMI increased mortality, but we didn't look at social deprivation. It was a smaller cohort and we actually showed the exact opposite that actually you died earlier.
If you had a knee replacement [audio out].
And this was far bigger more reliable. But from the deprivation point of view, I think that's that, that [00:24:00] that's something we can't answer. As Chloe said earlier, that this is just an association as opposed to causation. And is it because those patients have to wait longer? There's quite a bit of evidence that if you come from a socially deprived background, it takes you longer to see the GP.
You don't know whether you should present or not. If you do present, it takes you longer to get referred. Yeah. And maybe it's a bit of that. Certainly social deprivation is associated with earlier onset of osteoarthritis. That they are presenting late. This may be what we could see in the future for everybody because they're having to wait longer for the joint replacement.
No, absolutely. And as you know, we've done work in trauma related deprivation and it shows a similar thing in the poor outcome these patients get. But I think the thing we've all found hard is that deprivation is so multifactorial. It's difficult to know what part of that deprivation or maybe all of it is contributing to that poor outcome or poor access to service.
So maybe just to finish up, maybe Liam, if I come back to you this is obviously a lot of your hard work, as we know, and you know, what, what are your thoughts maybe on the implications moving forward, or maybe what, what, what [00:25:00] work you think needs to be done next in relation to this?
Well, I, I think the implications have kind of already been stated, I suppose. I mean, I think care needs to be taken, not to overstate these findings to kind of view them within, you know, the study design to accept that there is uncertainty, but associations exist. I think we have to acknowledge the competing risk of death when we think about implant revision rates.
And I think as Chloe said, patients they have different you know, these, to those who are 80. And be aware of that when we're discussing these things with them. And, and then just finally, again, just to kind of mention periprosthetic joint infection is, you know, is a significantly associated factor, which increases [audio out]
and it just adds to a body of work, which is [00:26:00] building, which supports that, that this, you know, it's a serious condition which can shorten a patient's life span. Yeah, no, absolutely. I think that's, that's, that's a very good point to highlight as as we finish up Liam. That is a really important take home message. Well, well, well, everyone I'm afraid
that's all we have time for today. So thank you so much for taking the time to join us and congratulations to you all on a really excellent study that is a real valuable addition to the literature in the area. It was great to have you all with us. 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 the like, feel free to tweet or post about anything we've discussed here today.
And thanks again for joining us.