Welcome to the Focus group Causality blog

This blog focuses on sharing, using and developping methods to understand the causal mechanisms underlying disease.

Membership of the focusgroup is available for members of the Dutch society of epidemiology (VVE, http://www.epidemiologie.nl) but this blog is open for everybody. Update on the blog wil be sent automaticaly to members of the focus group.  Please check regularly or subscribe by email or  with your RSS reader.

JC: multi-causality; use the two hit, treshold or component cause model?

So, we all understand the sufficient cause model, but lets try to focus on other models that underly the causal mechanisms leading to disease. In this journal club suggestion, several of these models are applied to an interesting disease, being transfusion related acute lung injury. I asked one of the authors to introduce the article, and he wrote:

We need to appreciate and understand the multi-causal nature of diseases to be able to fully understand their etiology and identify the most effective and convenient targets for (preventive) intervention. In this paper we discuss two models commonly used to appreciate the multi-causality of diseases: the threshold model and the sufficient cause model. We apply both models to the example of transfusion-related acute lung injury and discuss the differences, overlap, limitations and strengths of both models.

So, why wait? You can either view the article on pubmed or just here (pdf). Enoy!

Book review: Enigmas of Health and disease (Morabia)

cover Enigmas of health and disease (amazon.com)

It has been a while since our last post. There is no real excuse, except standard academia ‘busy busy busy’. And while being busy it is very relaxes to just take a moment off, kick back, pour a nice glass of wine and read AlfredoMorabias new book “Enigma’s of Health and Disease”. This book is not really targeted at students of practitioners of epidemiology, but more for interested people such as your little cousin who is in high school and is fascinated by medical research. The book, focused on the concept of comparing groups, uses light-non technical language to describe how epidemiology evolved from the ‘numerical method’ (chapter 5)  applied in French hospitals in 1800′ to the trial methodology now used to decide whether a new treatment should be incorporated into clinical practice or not. Without ever getting to technical,Morabia describes all fundamentals of epidemiology, i.e. comparing groups without counting not enough, not counting the right thing, or mistaking one aspect for the other.

There is one major thing missing: prediction. Morabia does discuss the concept of causation (there is even an appendix on interaction of causes and a nice description of confounding, or ‘ when group comparisons fail” (poppers as the cause of Kaposi’s Sarcoma, page 172), and since this is blog on causal inference we are very glad that a clear-cut description of causality is indeed given. However, epidemiology is more than just our little nook of the universe, and a good description of the historical development of prediction research and prediction models in clinical practice would  have fitted in this book. Perhaps Morabia decided to leave the concept of prediction (the word isn’t even found in the index), for the sake of simplicity, but this is a missed opportunity: A section juxtaposing prediction and causal inference does provide strong insight in the work does epidemiologist do.
And it is this what this book is all about: explaining how epidemiologist contribute to medical knowledge through group comparisons and population thinking. By doing so, he hopes to promote health knowledge amongst the general public. In the last chapters he also describes his wish, which is to teach epidemiology, or least the concepts of group comparison and population thinking, to high school students. And I believe that this is indeed a good step in educating the public on medical matters. Of course, most of epidemiological methods are to complex for this group, but the concepts can be taught! And although this book is, according to Morabia, “not a textbook”, it does provide lots of material from both ancient and recent history to make his wish come true.

Fall Meeting in Utrecht on november 14th

We are looking back to a great preconference meeting during the WEON 2014. But why stop there? Why not continue our series of get togethers with a fall meeting? To stimulate interaction between researchers on causality, the causal focus group of the Dutch Epidemiological Society will organise a meeting where causal researchers from different universities will present their causal work. The meeting is on November 14 at the UMC Utrecht, in the Dompleinzaal, from 2-5 pm. No big overviews this time, but Focusgroup members and their current work.
The program is below.
VvE focus group Causality
Date: November 14, 2014
Time: 2 pm – 5 pm
Location: UMC Utrecht, Dompleinzaal
2:00 S. LeCessie / R.H.H. Groenwold,
2:10 L. Smits, Maastricht University
Collider stratification bias
2:40 C.M. Hazelbag, UMCU
Marginal structural modeling of continuous exposures
3:10 A.G.C. Boef, LUMC
Instrumental variable analysis
3:40 Tea break
4:00 G. Swaen, Maastricht University
Weight of evidence
4:30 W. Schuller, VUmc
Causality and evidence based medicine
5:00 S. LeCessie / R.H.H. Groenwold
 There are no registration fees, but please register by sending an email to
We hope to see you all !

COURSE: advanced epidemiology

This will be the second year the advanced epi-course organised by the dept of clinical epidemiology wil be organised. See below for a more detailed description, or click here for the website or here for the pdf. (course will be held in Dutch)


Datum & Locatie

Donderdag 6 & vrijdag 7 november, donderdag 11 & vrijdag 12 december 2014

Congreshotel “Kasteel Oud Poelgeest” te Oegstgeest



Deze cursus geeft een overzicht van recente methodologische ontwikkelingen in  epidemiologisch onderzoek. We behandelen klassieke en moderne methoden om te corrigeren voor confounding, zoals standaardisatie, regressiemodellen, propensity scores, inverse probabilityweighting en instrumentele variabele analyse. We besteden aandacht aan grafische methoden om causale relaties weer te geven door middel van Directed Acyclic Graphs (DAGs). Deze DAGs zijn een handig hulpmiddel om confounding en selectiebias op te sporen, en om te bepalen hoe op een correcte manier met bias en confounding omgegaan kan worden in de data analyse.  Andere onderwerpen die aan de orde komen zijn: mediatie-analyse: het bepalen via welke paden de relatie tussen expositie en uitkomst loopt; competing risks: het omgaan met overlevingsduurgegevens waarbij er meerdere concurrerende uitkomsten zijn, en het omgaan met ontbrekende gegevens in de data-analyse. Gezien het intensieve karakter van de cursus kunnen slechts een beperkt aantal deelnemers (± 30 cursisten) worden toegelaten.



De cursus is bestemd voor promovendi die in opleiding zijn tot epidemioloog, voor epidemiologen die zich willen bijscholen in moderne epidemiologische methoden en voor andere onderzoekers die betrokken zijn bij de opzet en analyse van epidemiologisch wetenschappelijk onderzoek.



Dr. J.G. van der Bom, dr. S.C. Cannegieter, dr. S. le Cessie, dr. O.M Dekkers, dr. R.A. Middelburg.



€ 1.000 euro incl. lunches en diner op donderdagen, exclusief overnachting.  Tegen een meerprijs van € 250 is een hotelkamer beschikbaar. Van de procedure voor aanvraag van een hotelkamer worden geselecteerde deelnemers via e-mail op de hoogte gesteld.


Deadline inschrijving: dinsdag 1 juli 2014. Gedetailleerde informatie over de cursus kunt u lezen in bijgevoegde folder “Gevorderde Epidemiologische Methoden 2014”. U vindt tevens in bijlage het inschrijfformulier.

Zowel folder als inschrijfformulier zijn ook beschikbaar via onze website.

correlation != causation: Spurious Correlations

Tyler Vigen is a cool guy. He runs the spurious correlations website where he allows visitors to check out all the correlations between all kinds of statistics. This is not the most important reason why Tyler is a cool guy… It is the fact after that his website went viral, his contribution to understanding statistics for the public is quite important. And all that for somebody without a statistics background. In his own words

I’m not a math or statistics major (and there are better ways to calculate correlation than I do here), but I do have a love for science and discovery and that’s all anyone should need. Presently I am working on my J.D. at Harvard Law School.

if you hav the time, check his video in which his fondness of science is quite visible. The classic case of smoking and lung cancer is also discussed.

JC suggestion: Dormant Mendelian Randomisation studies in RCTs? Yes or No?

Just some weeks ago we’ve sent out a tweet asking for tips about digital repositories of journal clubs or anything similar, that are publically available. One of the suggestions was the following link:

Not really a coincidence since the participants of this reading club do overlap with the members of the focus group. So is there something from that journal club that can be used for this blog? Yes! For example, CS participants discussed an article on dormant Mendelian randomization studies embedded within RCTs by Schooling et al in the AJE. Although there is merit for MR analyses, it is questionable whether RCT should have dormant MR studies to resolve problems of unclear causal mechanisms, according to three CS members who wrote a response to this article. As scientific discourse allows, there is also a response by the authors. All in all, a nice combination of three articles to discuss during your own JC.

The authors of the original article also responded to message. See the links below to read the all three articles.

Disclosure – the author of this post is also also of the LTTE –