Messy epidemiology: the tale of transient global amnesia and three control groups

Clinical epidemiology is sometimes messy. The methods and data that you might want to use might not be available or just too damn expensive. Does that mean that you should throw in the towel? I do not think so.

I am currently working in a more clinical oriented setting, as the only researcher trained as a clinical epidemiologist. I could tell about being misunderstood and feeling lonely as the only who one who has seen the light, but that would just be lying. The fact is that my position is one privilege and opportunity, as I work with many different groups together on a wide variety of research questions that have the potential to influence clinical reality directly and bring small, but meaningful progress to the field.

Sometimes that work is messy: not the right methods, a difference in interpretation, a p value in table 1… you get the idea. But sometimes something pretty comes out of that mess. That is what happened with this paper, that just got published online (e-pub) in the European Journal of Neurology.  The general topic is the heart brain interaction, and more specifically to what extent damage to the heart actually has a role in transient global amnesia. Now, the idea that there might be a link is due to some previous case series, as well as the clinical experience of some of my colleagues. Next step would of course to do a formal case control-study, and if you want to estimate true measure of rate ratios, a lot effort has to go into the collection of data from a population based control group. We had neither time nor money to do so, and upon closer inspection, we also did not really need that clean control group to answer some of our questions that would progress to the field.

So instead, we chose three different control groups, perhaps better referred as reference groups, all three with some neurological disease. Yes, there are selections at play for each of these groups, but we could argue that those selections might be true for all groups. If these selection processes are similar for all groups, strong differences in patient characteristics of biomarkers suggest that other biological systems are at play. The trick is not to hide these limitations, but as a practiced judoka, leverage these weaknesses and turn them into a strengths. Be open about what you did, show the results, so that others can build on that experience.

So that is what we did. Compared patients with migraine with aura, vestibular neuritis and transient ischemic attack, patients with transient global amnesia are more likely to exhibitsigns of myocardial stress. This study was not designed – nor will if even be able to – understand the cause of this link, not do we pretend that our odds ratios are in fact estimates of rate ratios or something fancy like that. Still, even though many aspects of this study are not “by the book”, it did provide some new insights that help further thinking about and investigations of this debilitating and impactful disease.

The effort was lead by EH, and the final paper can be found here on pubmed.

Advertisements

Genetic determinants of activity and antigen levels of contact system factors

2018-11-08 12_43_09-RATIO instol zymogen.ppt [Compatibility Mode] - PowerPoint
One of my slides with a cartoon of the intrinsic coagulation system. I know, the reality is way more complicated, but still, I like the picture!
The contact system, or intrinsic coagulation system, have for a long time been an undervalued part of the thrombosis and hemostasis field. Not by me. I love FXI & FXII Not just now, since FXI is suddenly the “new kid on the block” as the new target for antithrombotic treatment through ASOs, but already since I started my PhD in 2007/2008. As any of my colleagues from back then will confirm, I couldn’t shut up about FXI and FXII as I thought that my topic was the only relevant topic in the world. Although common amongst young researcher, I do apologize for this now that I have 20/20 hindsight.

Still, it is only natural that some of the work I continues to be focused on those little bit weird coagulation proteins. Are they relevant to hemostasis? Are they relevant in pathological thrombus formation? What is their role in other biological systems? Questions that the field is only slowly getting answers to. Our latest contribution to this is the analyses of genetic variations in the genes that code for these protein, and estimate if the levels of activation and antigen are in fact -in part- genetically determinant.

This analysis was performed in the RATIO study, from which we primarily focused on the control group. That control group is relatively small for a genetic analyses, but given that we have a relative young group the hope is that the noise is not too bad to pick up some signals. Additionally, given the previous work in the RATIO study, I think this is the only dataset that has a comprehensive phenotyping of the intrinsic coagulation proteins as it includes measures of protein activity, antigen and activation.

The results, which we published in the JTH, are threefold: we were able to confirm previously reported associations between known genetic variations and phenotype. Se were also able to identify two new loci (i.e. KLKB1 rs4253243 for prekallikrein and KNG1rs5029980 for HMWK levels). Third, we did not find evidence of strong associations between variation in the studied genes and the risk of ischemic stroke or myocardial infarction. Small effects can however not be ruled, as the sample size of this study is not enough to yield very precise estimates. 

The work was spearheaded by JLR, with tons of help by HdH, and in collaboration with the thrombosis group at the LUMC.

The paper is published in the JTH, and as always, can also be found at my Mendeley profile.

Getting your life back on track after stroke: returning to work

https://goo.gl/CbNPSE

Stroke severity and incidence might be stabilizing, or even decreasing over time in western countries, but this sure is not true for other parts of the world. But here is something to think about: with increasing survival, people will suffer longer from the consequences of stroke. This is of course especially true if the stroke occured at a young age.

To understand the true impact of stroke, we need to look beyond increased risk of secondary events. We need to understand how the disease affects day-to-day life, especially long term in young stroke patients. The team in Helsinki (HSYR) took a look at the pattern of young stroke patients returning to work. The results:

We included a total of 769 patients, of whom 289 (37.6%) were not working at 1 year, 323 (42.0%) at 2 years, and 361 (46.9%) at 5 years from IS.

That is quite shocking! But how about the pattern? For that we used lasagna plots, something like heatmaps for longitudinal epidemiological data. The results are above: the top panel is just the data like in our database, while the lower data has some sorting to help interpret the results a bit better. 

The paper can be found here, and I am proud to say that it is open access, but you can as always just check my Mendeley profile.

Aarnio K, Rodríguez-Pardo J, Siegerink B, Hardt J, Broman J, Tulkki L, Haapaniemi E, Kaste M, Tatlisumak T, Putaala J. Return to work after ischemic stroke in young adults. Neurology 2018; 0: 1.

Cardiac troponin T and severity of cerebral white matter lesions: quantile regression to the rescue

quantile regression of high vs low troponin T and white matter lesion quantile

A new paper, this time venturing into the field of the so-called heart-brain interaction. We often see stroke patients with cardiac problems, and vice versa. And to make it even more complex, there is also a link to dementia! What to make of this? Is it a case of chicken and the egg, or just confounding by a third variable?  How do these diseases influence each other?

This paper tries to get a grip on this matter by zooming in on a marker of cardiac damage, i.e. cardiac troponin T. We looked at this marker in our stroke patients. Logically, stroke patients do not have increased levels of troponin T, yet, they do. More interestingly, the patients that exhibit high levels of this biomarker also have high level of structural changes in the brain, so called cerebral white matter lesions. 

But the problem is that patients with high levels of troponin T are different from those who have no marker of cardiac damage. They are older and have more comorbidities, so a classic case for adjustment for confounding, right? But then we realize that both troponin as well as white matter lesions are a left skewed data. Log transformation of the variables before you run linear regression, but then the interpretation of the results get a bit complex if you want clear point estimates as answers to your research question.

So we decided to go with a quantile regression, which models the quantile cut offs with all the multivariable regression benefits. The results remain interpretable and we don’t force our data into distribution where it doesn’t fit. From our paper:

In contrast to linear regression analysis, quantile regression can compare medians rather than means, which makes the results more robust to outliers [21]. This approach also allows to model different quantiles of the dependent variable, e.g. 80th percentile. That way, it is possible to investigate the association between hs-cTnT in relation to both the lower and upper parts of the WML distribution. For this study, we chose to perform a median quantile regression analysis, as well as quantile regression analysis for quintiles of WML (i.e. 20th, 40th, 60th and 80th percentile). Other than that, the regression coefficients indicate the effects of the covariate on the cut-offs of the respective quantiles of the dependent variable, adjusted for potential covariates, just like in any other regression model.

Interestingly, the result show that association between high troponin T and white matter lesions is the strongest in the higher quantiles. If you want to stretch to a causal statement that means that high troponin T has a more pronounced effect on white matter lesions in stroke patients who are already at the high end of the distribution of white matter lesions. 

But we should’t stretch it that far. This is a relative simple study, and the clinical relevance of our insights still needs to be established. For example, our unadjusted results might indicate that the association in itself might be strong enough to help predict post stroke cognitive decline. The adjusted numbers are less pronounced, but still, it might be enough to help prediction models.

The paper, led by RvR, is now published in J of Neurol, and can be found here, as well as on my mendeley profile.

 von Rennenberg R, Siegerink B, Ganeshan R, Villringer K, Doehner W, Audebert HJ, Endres M, Nolte CH, Scheitz JF. High-sensitivity cardiac troponin T and severity of cerebral white matter lesions in patients with acute ischemic stroke. J Neurol Springer Berlin Heidelberg; 2018; 0: 0.

Impact of your results: Beyond the relative risk

I wrote about this in an earlier topic: JLR and I published a paper in which we explain that a single relative risk, irrespective of its form, is jus5t not enough. Some crucial elements go missing in this dimensionless ratio. The RR could allow us to forget about the size of the denominator, the clinical context, the crude binary nature of the outcome. So we have provided some methods and ways of thinking to go beyond the RR in an tutorial published in RPTH (now in early view). The content and message are nothing new for those trained in clinical research (one would hope). Even for those without a formal training most concepts will have heard the concepts discussed in a talk or poster . But with all these concepts in one place, with an explanation why they provide a tad more insight than the RR alone, we hope that we will trigger young (and older) researchers to think whether one of these measures would be useful. Not for them, but for the readers of their papers. The paper is open access CC BY-NC-ND 4.0, and can be downloaded from the website of RPTH, or from my mendeley profile.  

How you quantify the impact of your results matters. A bit.

This is not about altmetrics. Nor is about the emails you get from colleagues or patients. It is about the impact of a certain risk factor. A single relative risk is meaningless. As it is a ratio, it is dimensionless, and without the context hidden in the numerator and denominator, it can be tricky to interpret results. Together with JLR I have a paper coming up in which we plead to use one of the many ways one could interpret the impact of your results, and just simply go beyond the simple relative risk. This will be published in RPTH, the relatively new journal of the ISTH, where I also happen to be on the editorial board.
Venn diagram illustrating the intersections of the independent predictors and poor outcome 12 months after stroke.  https://doi.org/10.1371/journal.pone.0204285.g003
One of those ways it to report the population attributable risk: the percent of cases which can be attributed to the risk factor in question. It is often said that if we had a magic wand and would use it to make the risk factor disappear X% of the patient will not develop the disease. Some interpret this as the causal fraction, which is not completely correct if you dive really deep into epidemiological theory, but still, you get the idea. In a paper based on PROSCIS data, with first author CM at the helm, we have tested several ways to calculate the PAR of five well known and established risk factors for bad outcome after stroke. Understanding what lies behind which patient gets has a bad outcome and which doesn’t is one the things we really struggle with, as many patient with well established risk factors just don’t develop a poor outcome. Quantifying the impact of risk factors, and arguably more importantly, ranking the risk factors is a good tool to help MDs, patients, researchers and public health officials to know where to focus on. However, when we compared the PARs calculated by different methods, we came to the conclusion there is quite some variation. The details are in the table below, but the bottom line is this. It is not a good sign when your results depend on the method. Similar methods should get similar results. But upon closer inspection (and somewhat reassuring) the order of magnitude as well as the rank of the 5 risk factors stays almost similar.
https://doi.org/10.1371/journal.pone.0204285.g003
So, yes, it is possible to measure the impact of your results. These measures do depend on the type of method you have used, which in itself is somewhat worrying, but given that we don’t have magic wand of which we expect to remove a fraction of the disease of up to 2 decimals precise, the PAR is a great tool to get some more grip on the context of RR. The paper was published in PLOS One and can be found on their website or on my mendeley profile PS This paper is one of the first papers with patient data in which we provided the data together with the manuscript. From the paper: “The data that support the findings of this study are available to all interested researchers at Harvard Dataverse (https://doi.org/10.7910/DVN/REBNRX).” Nice.

Teaching award from the German society for epidemiology

teaching at ESOC 2018 summer school
teaching an interactive session on study design at ESOC 2018 summer school

The German society for epidemiology has an annual teaching award, i.e. the “Preis für exzellente Lehre in der Epidemiologie”. From their website:

Mit der Auszeichnung sollen herausragende Leistungen oder überdurchschnittliches Engagement in der Lehre der Epidemiologie gewürdigt werden. (…)  Preiswürdig sind innovative, originelle oder nachhaltige Angebote, ebenso wie ein besonders hoher persönlicher Einsatz für die Lehre.”

In short, anything goes in terms of format, innovation, personal commitment etc. However, there is a trick: only students can nominate you. So what happened? My students nominated me for my “overall teaching concept”. Naturally, the DGEpi wondered what that teaching concept actually was and asked me to provide some more information. So I took that opportunity and actually described what and why I teach, to see what the actual concept behind this all is. Here is the result.

The bottom line is simple: I think you learn the best not only by reading a book, but that you learn by doing, help in the organization and help teach in various epi related activities. You need to get exposed in several formats with different people. So I have helped set up a plethora of activities for the young student to learn epidemiology in different ways on different levels: read classics, discuss in weekly journal clubs, use popular scientific books  in book clubs, but also organize platforms for discussion, interaction and inspiration (yes, I am talking about BEMC). The most important thing might be that students should learn the basics for epidemiology, even though they might not need that for that own research projects. This is especially true for medical students who want to learn about clinical research.

Last week I learned that the award in the end was awarded to me. Of course I am honored on a personal level, and this honors needs to be extended to my mentors. But I also take this award as an indication that the recent and increasing Berlin based epi-activities I helped to organize together with epi enthusiast at the IPH, iBIKE and QUEST did not go unnoticed by the German epidemiological community.

I will pick up the price in Bremen at the yearly conference of the DGEPI. See you there?