Dr Michael Eyre, Clinical Research Fellow at King’s College London and Honorary Consultant Paediatric Neurologist at Evelina London Children’s Hospital, UK

“I trained in Natural Sciences at the University of Cambridge and Medicine at King’s College London, graduating in 2008. I completed subspecialty training in paediatric neurology at Great Ormond Street Hospital for Children and Evelina London Children’s Hospital. My clinical and research interests are in acute neurology, neuroimaging and neuroinflammation in children. I am funded by Action Medical Research and the British Paediatric Neurology Association to develop advanced magnetic resonance techniques for autoimmune encephalitis in children and young people. I am part of an international collaborative group developing the definitive consensus guideline for diagnosis and management of Sydenham’s chorea.”

Transcript

[Michael Eyre]
Thank you for the invitation to speak and it’s been a pleasure to listen to the other talks and that’s the nice thing about small group. You get much more free flowing discussion and I hope that will continue.

I don’t have any disclosures, I am funded in my research by two charities.

So clearly most of what we’re discussing today is how do we best care for patients with Sydenham’s chorea and as Tamsin said in her talk, I’m going to use the word “patient”. And you know she was saying she’s uncomfortable with using the word cases but it’s just too long winded to say “people affected by…” all the time. So I’m going to use the word “patient”. I hope that people find that OK.

So this pyramid is sorted by how rigorous the methods of the studies are. In the middle, you’ve got these “primary studies” of which the gold standard for evaluating treatments is the randomized control trial, and then you’ve got “secondary studies” that sum across the findings of these other studies.

And so in Sydenham’s chorea and other rare diseases, we simply don’t have very many randomized control trials, and if we do, they’re often quite small numbers of patients. We do have some well designed observational studies and those are really important. But what we’ve mostly got is lots of case reports and case series where there wasn’t a specific study design but it’s still important information and there’s lots of it.

So the first thing that I’m going to talk about today is a project where we’re trying to gather all of this data together in a meta analysis to try and extract the most information that we can from it; and then in the second part of the talk I’ll talk about how we’re trying to develop a guideline in which we’re going to sum across all of these different types of evidence.

So part one. So this is the group of people who have been working on this. We’ve done this previously in NMDA Receptor antibody encephalitis, which is a kind of autoimmune encephalitis, and other rare diseases.

So essentially the idea is that you take all of the case reports that have ever been published, and you put each patient as a separate row into a database, and then you ask some questions of that new data set.

And in this particular study, where, first of all, I’ve just trying to describe the features and what kind of management has been used in Sydenham’s chorea over the years, and then secondly, we’re looking at statistical associations within that data, which might give us clues as to what predicts the outcome for individual patients, based on their clinical features, and what treatments they’ve been given. So in particular, we were interested in what factors are associated with how long it takes for the chorea to resolve. And in the first episode we’re interested in what factors might predict whether or not you have a relapse.

Which factors are associated with your functional outcome? And I’ll explain what I mean by that in a moment. And what factors might be associated with your final psychiatric outcome?

So we do a systematic literature review. We’ve actually searched more, different databases than that (and Emmanuela’s help was really important for navigating that),  but the main articles were retrieved from Pubmed and Embase. So when you put this search in you get an awful lot of results and you need some help to read all of those articles, or at least to screen through them in order to see which ones are going to be relevant. So that was why we need this team of people helping with that. And then from these articles we extract cases of Sydenham’s chorea where they’ve given individual patient data. So they’ve described a case history for a patient with a particular age and sex and it’s clear that they’re providing data on an individual patient and then we put that into our combined dataset. But we have to rely mostly on the authors having made the correct diagnosis of Sydenham’s chorea  We did exclude a few more historical cases where it seemed that the diagnosis wasn’t correct.

Once we’ve got that dataset, we then pick out the cases which meet one of these outcome definitions, and this is what’s going to be the input to the statistical tests. So relapse is quite straightforward – if it’s a patient where they reported a relapse, then we can put them into that category. But in order to categorize them as not having had any relapses, we only select the patients who had at least 24 months of follow up.

Functional outcome we evaluate with the modified Rankin scale score, so here the green is the good outcome and red is the poor outcome. So this is how we’ve separated this. You can see that this is quite a crude scoring system, but you can’t really apply anything more detailed than this to case reports. And it’s quite weighted towards mobility. So the more severe is mostly about your ability to mobilize independently.

 We also looked at psychiatric outcome, which was simply defined as whether the patient still had ongoing psychiatric and behavioural  symptoms at the final follow up; and for the “functional and psychiatric outcome”, we’re  just looking at patients who had at least six months of follow up because obviously if you look at one month, they’re still in the acute episode and you can’t really draw any conclusions about that being an outcome.

So these are the outcome variables as well as the duration of the chorea . And then we’re interested in the effects of all these different clinical treatment factors on these outcome variables. And so for that we use regression modelling, which puts all of these predictor variables into a model together with an outcome so that we can separate out the independent effects of those factors. So these are the things that we looked at:

• age and sex,
• whether you’re from a country with a high or low rheumatic fever risk,
• presence of pre-existing neuropsychiatric disorders,
• other autoimmune disorders,
• whether there was evidence of a preceding streptococcal infection at the first episode of Sydenham’s chorea,
• some of your clinical features (trying to capture some elements of severity during the acute first episode),
• whether the patients had other rheumatic fever complications, and
• the main treatments. We are going to go through that list in a moment. I’ll show that in the results.

So we retrieved about over 300 articles which contained individual patient data. And from that we get 1.5 thousand patients. Some of these patients go all the way back to the 19th century because this actually has a really long history of being a recognizable condition.

This is the only slide where I’m going to show you data from all of that cohort, because I think it’s just quite interesting to look at these historical trends. So you can see that what hasn’t really changed over the years is that Sydenham’s  chorea tends on average to affect children of school age and median age of around 10 and there are more females than males affected, but what’s interesting with the fever, I think I’ll move on to the next bit of this slide because these two things are linked.

So what you can see is that after the Second World War, antibiotics become available and the use of antibiotics has steadily increased since then, so now almost all patients are treated with antibiotics, and you can see that what used to be more often than not a febrile illness during the acute presentation with chorea, that’s now not the case. And so I think that we would put this probably down to the antibiotics and then in these other plots you can see when other treatments became available. So we’re seeing a lot of case reports of patients being treated with haloperidol coming into the 60s and 70s and then sodium valproate in the 70s and 80s.

Then the immunotherapies more recently than that, but still actually there are quite low numbers of patients in this dataset who’ve been given immunoglobulin or plasma exchange.

I’m looking at some of the outcomes that we’re interested in. So curiously, it seems that the average duration of the chorea at the first episode is longer in recent decades. That might be because they maybe had a different threshold for when they said that chorea had resolved in the past. It’s quite difficult to be confident about those old cases.

I think the relapse rate maybe was a little higher historically than it is now and it looks like the number of cases where they have disability at the final follow up is certainly lower than it was in the past.

So for the rest of this data, I’m just showing the cases that had onset after 1945. We chose that because that was the year penicillin first became widely available, so all the rest of the data I’m going to show you exclude those very historical cases.

This is the age distribution among those cases, and this clearly aligns with what we know about this disease, that it’s mainly a disease of childhood, and mainly of school age childhood. Very few have onset in adulthood.

We collected cases from 50 countries – just showing the top 20 countries here, so Brazil, Turkey, Italy and the United States dominate this dataset.

So what were the features of Sydenham’s chorea at the first episode? Just over half had some symptoms of infection prior to the Sydenham’s chorea and as we know that mostly that was pharyngitis, on average 8 weeks before the onset of the neurological and psychiatric symptoms.

And 91% of these patients were admitted to hospital for a median stay of 21 days in hospital.

Going through some of the symptoms, and I think that these are probably underreported because it’s difficult to say in a case report if they don’t mention a symptom, you have to make a decision about whether to assume that it was truly absent or that they just didn’t mention it. And I think probably we underestimate the frequency of some of these symptoms because of that limitation.  That notwithstanding, 70% of the patients had impairment in their ability to walk including 34% couldn’t walk at all or without assistance.

2/3 had impaired object manipulation, so ability to use their hands and upper limbs.

64% of that had psychiatric symptoms and I’ve just listed on the right there the most frequent psychiatric symptoms. And I think this very much aligns with the talks that we’ve had earlier today. Although again probably an underestimate because in only a third of those patients have they mentioned that an assessment by mental health specialist in the case report.

62% of the patients had impairments in speech, including the 3% who were completely unable to speak, make any speech sounds, and around 1/4 had impairments in chewing and swallowing, including 16% who needed nasogastric feeding.

Moving on to the treatments. 86% of patients received antibiotics, so remember, these are cases going all way back to 1945, but the more recent years, it’s closer to 100%.

26% of the patients received immune therapies, by far the most frequent being corticosteroids.

And symptomatic medications, so medications to treat the chorea or the psychiatric symptoms, rather than immunotherapies and antibiotics, and by far the most frequent were Haloperidol and sodium valproate.

So what this figure shows is the data we collected on what the person reporting the case had said that they felt was the benefit of the medication. So this is very much anecdotal and relying on we don’t know what criteria they use to determine whether these medications were effective or not. But if the reporter says that they, for example, sodium channel blockers (mostly carbamazepine), “we gave carbamazepine with good effect” then that’s what is shown in the green element of each bar. If it wasn’t clear from the case report whether there was a benefit or not then that’s in the gray section. And if they said we gave carbamazepine and there was no benefit, then that’s the red section of these bars. And then what the asterisks show is where there are significant differences between different classes of medications.

So you can see that sodium channel blockers and antihistamines were most frequently associated with an impression of benefit in the case reports, although this is quite small numbers of patients from a few papers compared to sodium valproate and Haloperidol, which have been given to many more patients.

Looking at some of the descriptive data on the outcomes that we’re interested in, 86% of the patients had a full resolution of chorea after the first episode – and I think the caveat here is that in many cases the follow up perhaps wasn’t long enough to get to the point of full resolution.

The average duration of the chorea was three months, and that’s what this graph here shows. So you can see that when you get down to 50% and if you sort of draw a line horizontally from 50%, you get down to three months. If you want to know how many patients still have chorea at 12 months, for example, you go left and you see that 20% or less. Still having persistent chorea at 12 months, I think this is probably an overestimate actually because the patients who have persisting chorea are more likely to be retained in follow up.

If we look at relapses, 1/3 of these patients were reported to have had at least one relapse, and quite a small number of patients have many relapses.

And at final follow up the median Mrs score was 0, so that means “no disability”, no symptoms, and the maximum was four. So there are no patients who were still bed bound or had died at final follow up, out of 1200 patients, so I think that’s quite a reassuring message.

And 6% of had ongoing psychiatric and behavioural symptoms at final follow-up where that was at least six months after the onset. And 3% had cognitive school performance problems. I think this is a huge underestimate. And again, I think these things just aren’t mentioned in in case reports. And so that’s the limitation of this kind of data.

OK, so finally I’m going to show you the results of the multivariable models, so here I’ll go back to this questions that we’re interested in. First one was we want to know which clinical treatment factors were associated with the duration of chorea at the first episode.

So without wanting to get too much into the details of this, the line down the middle is the kind of line of equipoise. So that variable doesn’t have any impact on your outcome if you’re on that zero line. Then the square is the central estimate of the effect of that variable and the arms are the 95% confidence interval. And I’m only showing the variables that are significant at that 95% confidence threshold. So although we entered many variables into the model, these are the ones that are significant with an effect on this outcome.

And so what you can see is that immunotherapy was associated with Increased resolution of chorea during the treatment with immunotherapy [log hazard ratio 0.35], and other treatments weren’t significant for that; and the presence of carditis/valvulitis was associated with the longer duration of chorea [log hazard ratio 0.4].

The next question we were interested in was what might predict relapse.

So again, it’s the same kind of plot where 0 means that it doesn’t affect the outcome either way and red is bad and green is good. So patients who had arthritis/arthralgia (joint involvement) in their rheumatic fever were more likely to have relapses, and patients who had received haloperidol were more likely to have relapses.

On the good side, strangely, patients who had impaired speech during the first episode were less likely to have relapses. Patients who’ve been treated with steroids, sodium valproate, antibiotics were less likely to have relapses.

One thing I should say here is that, as I mentioned IVIG [immunoglobulin] and plasma exchange were used in very small numbers of patients, and lots of the symptomatic treatments were used in very small numbers of patients, so that means that they often just won’t achieve statistical significance. And so they’re less likely to come up in the significant results when they’ve only been used in small numbers of patients.

The third thing we’re interested in was the functional outcomes. This was the modified Rankin scale score.

And we found that people who had adult onset of Sydenham’s chorea were more likely to have a poor outcome, and there was an association of having been given haloperidol with having a poor outcome. And females had a better outcome.

And finally, we’re interested in the psychiatric outcome. So this was, “do you have ongoing psychiatric symptoms at final follow up?” And the associations with ongoing psychiatric symptoms were having received chlorpromazine during the acute SC episode, and, unsurprisingly, having had psychiatric/behavioural symptoms during the acute episode, and being from a country or a community with a high rheumatic fever risk.

And the association with a good psychiatric outcome was receiving steroids.

OK, so this is just summarizing those findings again, but sort of treatment wise, so immunotherapy was associated with the shorter duration of chorea and steroids were associated with reduced odds of relapse and increased odds of a good psychiatric outcome. Here, for those who aren’t academics, the P value is the probability that this result could have occurred just by chance and P value is less than 0.05, we generally regard as significant.

Antibiotics were associated with the reduced odds of relapse, and sodium valproate was associated with reduced relapse as well.

Haloperidol was associated with increased relapse and increased odds of poor functional outcome and chlorpromazine was associated with increased odds of poor psychiatric outcome.

So our main conclusion is that this provides some support for steroids, antibiotics and sodium valproate in Sydenham’s chorea.

So this is kind of the good, the bad and the ugly, and the ugly is really important. The ugly is the limitations, and mainly it’s about the type of data that we’ve collected this dataset from. So we know about publication bias – the case reports might not be representative of the whole population of patients who have had Sydenham’s chorea.  And our assessment of outcome is quite coarse because we can’t retrospectively apply a psychometric scale, so that’s probably why we’ve got quite small numbers of patients who actually seem to have a poor outcome.

And I think it’s really important also to recognize that we’re not capturing all of the subtle differences between patients, which might co-vary with their treatments. So patients who receive haloperidol probably had much more severe psychiatric symptoms than patients who didn’t. And maybe it’s that association which is driving the results that we’re seeing. So those caveats are important.

So I’ll just move on to the other element of the talk, which is to talk about our process for developing a consensus guideline. So that data I’ve just shown will be one of the inputs to that. But just as important as all the other types of published evidence which are already out there.

And of course, also when we’re practicing evidence based medicine, it’s not just about looking at the published data. It’s about integrating that with expert clinical judgment and with Sydenham’s chorea community and patients, who tell us what’s important for them.

And the process that we have to guide this is called a Delphi process. So in brief, you have a steering committee and that’s the names at the top of the screen there. And we’ve invited a rating committee of experts. So there’ll be 20-something experts on Sydenham’s chorea from all continents of the world. Except Antarctica, I think.

And we will define what we think the problems are and develop a questionnaire and then we send that questionnaire, asking everyone in the Expert Committee to vote on their agreement with different statements. And then you look at the initial results and then you refine the statements and repeat that process. And then at the end you have an in-person (or perhaps a virtual meeting) to get to reach a final consensus, and then you produce a report and a guideline based on that. So this is a process that we’ve done for NMDA receptor Antibody encephalitis, and I think, I hope that the output from that is quite useful for patients and families, because you get quite concrete statements about how the condition should be managed.

And you can see the percent of experts that agree with each of those statements, and this can be quite useful to show, to give to people who might not be so expert in the disease, to show, look, this is what’s expected. This is the gold standard for this disease.

So far we’ve got this far on the Delphi process, we’ve developed the first set of questions, but they haven’t yet been put to the experts. So for that reason. I’m not able to show you the specific items, but I’m going to show you some of the themes, the sections that are going to be on that questionnaire, and the overall goal of this.

So – the “problem definition” part of this is:

• to affirm key statements on the natural history and clinical features of Sydenham’s chorea,
• to affirm the key insights from the systematic review that I’ve just shown you, alongside the other kind of evidence that we have for treatments,
• to give guidance on the best approach to assess Sydenham’s chorea and follow up of children with Sydenham’s chorea,
• to identify areas where we have a particular need for future research.

OK, so there’s I think 12 sections. So the first section is just about establishing agreement on what Sydenham’s chorea is. And in general, the context in which the diagnosis is reached.

Second section will address which professionals should be involved in the care of a patient with Sydenham’s chorea, and so, you know, not just paediatric neurologists and paediatricians, but perhaps also cardiologists and psychologists and psychiatrists.

The next section will cover the typical clinical features of Sydenham’s chorea and the typical symptoms.

What assessment tools are recommended for use in patients with Sydenham’s chorea and what other kinds of evaluation like cardiac screening?

In the 5th section, we will seek consensus on the approach to investigation, which I think just have to be quite nuanced. Not everybody needs a lumbar puncture in this condition and so that that will be an interesting one to try and get agreement on.

The next section will cover statements on the general approach to management. So what are the goals of management, what kind of information and education needs to be provided for families?

What kind of adaptations need to be made to schooling? What kind of access should be available for rehabilitation? And, you know, multidisciplinary therapies.

And the next section will cover antibiotic treatment. I expect this is going to be the easiest to achieve consensus on. That’s not too controversial at all.

Immunotherapy is much more contested and that will be a more interesting section to try and get consensus on.

We’ll also provide recommendations on other medications, symptomatic medications.  Something I haven’t talked about actually in the data earlier – I was just showing you data about efficacy, but I wasn’t talking at all about risks, and so of course, in clinical practice, we put these two things together.

Management of psychiatric symptoms. Importantly, we’re going to make a statement about psychological interventions before a statement about pharmacological approaches.

We have a section on management of relapses. What constitutes a relapse? And when to offer treatment for a relapse?

And finally identify some priorities for future research.

OK, so in conclusion I’ve shown you some preliminary conclusions from our metanalysis, which seem to support the use of steroids, antibiotics, and in particular among the symptomatic treatments, sodium valproate.

So we’re writing that up for publication at the moment, and there’s lots more that we can interrogate within that dataset. So I’ve got a student who’s going to look at some of the relationships between the psychiatric and neurological symptoms.

The timeline estimated for producing this guideline is hopefully by the end of the year and we’re going to soon put the first round of statements to the expert panel.

Thank you for all the people who’ve been working on this together, it’s been very collaborative and thank you to the people that fund my time to do this work.

[Nadine]
Thank you, Mike. That that was wonderful. It’s very, very exciting work. And so any questions from the participants please?

[Michael]
I didn’t want to be first, but I will.

I was interested in the data about the origin of the papers, and you’ve basically got three big areas: the United States, Turkey and Brazil, which produce a lot of the information.

Have you thought about trying to look at the analysis without one or other of those countries? Because I suppose there’s a weighting there to particular health systems, and particular ways of thinking, and I imagine the three countries have quite different health systems and we were talking earlier about certainly psychiatric issues being picked up more or less according to what people see. So I’d be interested to know whether it’s possible or worthwhile cutting out one country and seeing what difference it makes.

[Michael Eyre]
Yeah, I’d certainly take the point, that would be interesting to do. The problem with regression is always stopping yourself from doing it too many times, because then there’s a temptation to just choose your favourite result.

But it would be interesting to look at that.

[Nadine]
I found it quite sad when we did the systematic review because in Brazil and in Israel it was, it was more to do with, if you could afford the treatment, you got it. And I do think that is quite sad. So we had some papers where there was like 360 people with Sydenham’s, but they’d all paid privately for that care. And it is very interesting to look throughout the world at the different case studies and treatments.

[Tamsin]
I don’t have any tricky questions. Actually, I was just said I was more reflecting on the huge volume of work that’s gone into to what you’ve done.

[Michael Eyre]
It wasn’t just me, it was a team.

[Tamsin]
I’m hugely impressed. Well, you and the team, but yeah.

[Michael]
It’s very refreshing today to hear the patient story from Callum, having just spent so long staring at patients as rows in the spreadsheet, it’s actually really nice.

[Karljowan]

We have a question about to confirm the diagnosis of Sydenham’s chorea, how important and how do you see the streptococcal titres, in the serology, how important is that to the diagnosis?

I’ve seen in the literature the percentage of positive titres differs from 7 to 80% and how is your view on that, that there might be around 20% that don’t show up positive titres.  What’s your opinion and experience of on that issue?

[Michael Eyre]
My personal view, and I’m interested in other people’s views, is that it’s not essential, you don’t have to have the titres positive to make the diagnosis. It’s supportive and it probably means that you can do less other tests, but you can definitely have patients who are negative for any evidence of streptococcal infection. But you’ve got completely clear clinical picture and other tests which have excluded other diagnosis. So it’s not essential at all.