Biomarkers – essentials to get you started

Interview with Guillaume Desachy and Nicole Krämer

Interview with Guillaume Desachy and Nicole Krämer

Biomarkers help you in all kinds of different ways, so stay tuned while Guillaume, Nicole, and I talk about the following points:

• What are biomarkers for you?
• What makes it such an interesting topic?
• Why do you want to revitalize the biomarker SIG?
• What are your goals for the SIG?
• The richness of biomarkers requires more advanced methods. What would you foresee your SIG to do about these?
• How can people join the SIG?

Reference: SIGS Biomakers

Please send an email to Nicole or Guillaume.

Listen to this episode and share this with your friends and colleagues!

Guillaume Desachy

Since graduating from ENSAI (Biostatistics M. Sc.) 10 years ago, Guillaume has been immersing himself in precision medicine.

Data-driven, he is passionate about answering scientific questions and making sure we convey the right message to stakeholders, both internally & externally. 

He feels very fortunate to have had the chance to work with various kinds of OMICs data and leverage the power of biomarkers to strengthen drug development.

He also feels incredibly lucky to have worked in a diverse set of settings, be it in academia (UCSF, U.S.), in a biotech (Enterome, France) or in the pharmaceutical industry (BMS, Servier & AstraZeneca, France & Sweden). He now works as a Statistical Science Director for AstraZeneca in Gothenburg, Sweden.

Apart from his day job at AstraZeneca, Guillaume teaches a course about OMICs data analysis at ENSAI (www.ensai.fr), is a Board Member in the ENSAI alumni association (www.ensai.org) and is a mentor for Article 1, a non-profit organization promoting equal opportunity (https://article-1.eu/).

Whether it is to discuss statistics, choices that you are making in your early career or any other subject, you can contact him via LinkedIn (https://www.linkedin.com/in/guillaume-desachy/).

Nicole Krämer

Nicole is a senior principal statistician at Boehringer Ingelheim in Biberach, Germany. As a member of the Therapeutic Area & Methodology Statistics Group, she supports clinical development teams in terms of strategy and methodology for biomarker analysis, translational topics and early clinical development. Her interests include complex high-dimensional data, subgroup identification and model validation.

She received her PhD in machine learning in 2006. Together with Guillaume Desachy, she chairs the EFSPI/PSI Biomarker Special Interest Group. She is also a member of the EFSPI/PSI Subgroups Special Interest Group.

Transcript:

Alexander: Welcome to the effective statistician podcast, the weekly podcast with Alexander Schacht and Benjamin Piske, designed to help you reach your potential, lead great sciences and serve patients without becoming overwhelmed by work. Today, I’m talking with Nicole and Guillaume about biomarkers, so stay tuned. 

Biomarkers are really, really incredibly important topics. They can help you in all kinds of different ways and how exactly we will dive into this in the episode. We also speak about the biomarker SIG and this biomarker SIG will also be very very present at the PSI conference that is coming up.  So if you have not yet registered for the PSI conference, head over to psiweb.org and sign up there now because it’s really around the corner and maybe know you can convince your supervisors to attend it. It will be offsetting, I will be there as well in Gothenburg and you know just looking outside it probably will be an amazing event also from the weather perspective.

 I’m producing this podcast in association with PSI, who was organizing the conference so come over register there and I’m sure it will be absolutely amazing. So now let’s get into the discussion with Nicole and Guillaume.

Welcome to another episode of The Effective Statistician. Today, I’m speaking with Guillaume and Nicole about biomarkers, nice to have you on the podcast. 

Nicole: Thank you Alexander for having us. 

Guillaume: Yeah, I’m very, very happy to be here. Thank you, Alexander. 

Alexander: So maybe as a start, you can introduce yourself first, Nicole do you want to start?

Nicole: Yes, sure. So and before I start, I just want to state that the views that I’m stating today are my own. And I’m not speaking on behalf of my employer, which is Boehringer Ingelheim. Ok. So hello everybody. My name is Nicole and I’m a senior principal statistical advisor at Boehringer Ingelheim and I’m located in Germany. So being a member of the therapeutic area and methodology group, my role is mainly to support all the clinical development teams and all topics around biomarker analysis, translation strategy, and early clinical development. So these are all topics very close to my heart and topics I’ve worked on in the past. I’ve always loved high-dimensional data coming from the life sciences identifying best models and ways to analyze these data. Maybe just very briefly regarding my scientific background, I studied mathematics in Cologne and then I did my PhD in Machine learning quite a while ago in Technische, Berlin. And I’ve also spent several post-taught years in Munich and also known as a guest professor at the LMU biostatistics. So a few years ago, I moved from academia to industry working for a consultancy called Staburo for several years and now I’m in Boehringer Ingelheim last July. 

Alexander: Awesome, very, very good. Guillaume, how about you?

Guillaume: Same as Nicole, before we get started, what I’m gonna be expressing today is my personal views, my personal opinions, not speaking up on behalf of us as AstraZeneca. So these being said, my name is Guillaume and I work as a statistical science director for AstraZeneca in Sweden. And I graduated from ENSAI, which is a French National School of Statistics about 10 years ago and I ended up having my internship at the University of California, Francisco. And over there, I was working on aiming and finding genetic variants that could be linked to autism. And I think I got hooked, I loved it. I loved the fact that we were in uncharted territories. I love the fact that when we are speaking with data providers, they are actually telling us, what you are trying to do, simply not possible, because that’s not how we designed the cheap. So I got hit and since then I’ve been working with biomarkers. When I relocated to Europe a couple of years back, I started to make the transition to the private sector and I wrote for a French Pharma, working with biomarkers in oncology. After that, I spent a bit of time in a biotech which was working on metagenomes and using the microbiome to develop new jobs. And now I’m back in Pharma and I work for AstraZeneca and I work for statistics design, but still biomarkers are very close to my heart. 

Alexander: Do you work in a specific therapeutic area or do you work across your therapeutic areas? 

Guillaume: So, at the moment, I work in respiratory.

Alexander: Okay, very good. So speaking about biomarkers, what actually is that? 

Guillaume: Well actually I think that’s a very good question but if we think about it, there is somehow an official definition from the NIH. And this official definition dates back to the late 90s and I wrote it down and I’m going to be the data because I think it’s very interesting. So this definition states that a biomarker is a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes or pharmacologic responses to a therapeutic intervention. So once you’ve got these definitions in mind, you can fit many things under the term biomarker. It can be genomics, can be transcriptomics, can be proteomics, can be metagenomics, it can mean many things. It can even be digitorum longus which are biogenomics measured by a digital device, that’s the NIH definition. But actually, what I think really modeled with biomarkers, is the role that they can be playing. And usually, we talk about three main roles with the saying that a biomarker is, is prognostic, that is predictive, or that is pharmacodynamic. 

Alexander:  Prognostic would mean that it gives you kind of value in terms of whether the disease will improve, or deteriorate, whether side effects will occur or not, predictive means that it actually gives you an indicator of, you know, which treatment will respond better or not. And what about the last one, the pharmacodynamic?

Guillaume:  You’re telling the truth regarding prognostic and predictive. And regarding pharmacodynamic biomarkers, these are biomarkers that actually allow to witness the pharmacodynamics of a treatment without always a link with treatment response and that’s the difference between the predictive and pharmacodynamic biomarkers. 

Alexander: Into pharmacodynamic may mean you have it tells you kind of you get two different plasma levels but different plasma levels don’t necessarily directly employ different treatment responses. 

Guillaume: Exactly. And it’s a good kind of showing some kind of targeting response. If you’ve got a given by, if you know it is going down again that way, then you might be interested in having a look at this biomarker. And this biomarker, a good pharmacodynamic indicator because if it varies, then it’s a good indicator that you’re hitting your target, you’re engaged in the target. Then we’ve got these three roles and obviously the whole good is predictive biomarkers because it allows you to find your biostatisticians. But they are very hard to find and that’s why everyone is whole. 

Alexander: Yeah, yeah, yeah. I spent a lot of time on this topic. So biomarkers obviously is a really, really interesting topic and it’s great that you cannot explain these three roles for that. What else about that? 

Nicole: Yes. So I completely agree with everything Guillaume said, so there’s an official definition that kind of covers everything. It also covers, of course, I don’t know HBA1C in diabetes, of course, it’s a biomarker, it’s a physiological measurement, but on the other end it’s a very established end point, the analysis is crystal clear. So there’s probably not much to discuss. So to me biomarkers always have this additional notion of complexity, maybe of multidimensionality and also maybe not an established way to analyze them. And just what you mentioned in terms of dynamic effects that you don’t yet have an established link to clinical endpoint and also on top of just analyzing the data one important field is also how to qualify or validate these biomarkers how to say, okay, these are necessary to show a clinical response. So, if you, for example, have established this, you could use them for early decision making because if you don’t see anything in a biomarker, then it’s also clear. You will never see a clinical response. So these are kind of interesting topics so that not everything is already laid out and established and then how to help change this. And one point, I also want to stress I think that’s sometimes a misconception that because biomarkers are often termed exploratory, because they’re not yet established primary endpoints for regulatory purposes. People misunderstand this in terms of that kind of analysis aimless that we just measure them. And then let’s just have a look and see if we signal pops up. And I think this is not true. So we always have these purposes that you mentioned, you won’t have a prognostic marker, a predictive marker were so that you can help select patients who respond best to treatment or this pharmacodynamic effect or early endpoints. And to me it’s very important and close to my heart that we get away from this notion. It’s just exploratory. We will never learn. We can have a look and maybe visualize the data and we will never learn anything from this. So thanks very important to have, also some statistical rigor and the correct models to analyze these types of data. So, this is really at the core of what we also want to achieve with the biomarker analysis. 

Alexander: You mentioned this notion of validating a biomarker. What does that mean actually?

Nicole: I think it’s a very wide field. It starts kind of at the technical and analytical validation to make sure that you can reproduce it. If you measure three times that you get similar results, maybe I don’t know if you base something on a tumor biopsy or if you freeze it and think again that you can still measure these kinds of measurements. So that’s kind of more on the technical analytical side, which is very important. But also from a statistical or regulatory point of view, this may be the first step having maybe a correlation to a clinical end point when it comes to pharmacodynamic effect. And that I think also the holy grail or the top would be to have it established as a surrogate biomarker. And for this, of course, we have to do much more, we have to show that you have a causal relationship between treatment effects in the biomarker translated to treatment effects and the clinical end point. And then also when it comes to predictive biomarkers that you can really select the right patient population, there’s also a lot of things to do. You have to show that patients have a treatment benefit, that it also has some clinical utility if you use it in the clinic, that actually people also benefit from it and many many different topics are everything, very exciting and open active research fields. 

Alexander: Okay. Okay, very good. So you kicked off the biomarker SIG again. The special interest group, which is kind of joined forces of FSPI and PSI and that SIG has been kind of dormant for quite some time. Why do you revitalize it now? 

Guillaume: We all face the same challenges when we acquire market data. And if you think about it, we all face the challenge of handling data below the lower limit of quantification and its impact cycle analysis. So, that’s one thing. One favorite topic of mine is how to find optimal biomarkers. And we all face this, it’s not just, medium or unique, or we all face the same issue. And this topic has been discussed for years and years, but there is still no consensus. So that was our starting point for Nicola and I where we all face the same issues so maybe we can get together, maybe we can gather a diverse group of people coming from academia, coming from the pharmaceutical industry, from biotechnology area, we can get together, we can discuss these challenges. And maybe we cannot serve as a kind of bridge, the virtual boundaries that we see between organizations. And maybe we can push the biomarker topic a bit further. So those are the starting points of the idea of getting together,sharing lessons learned, sharing challenges and seeing where we can go.

Alexander: In terms of organizational boundaries, what are the main issues you see there? 

Guillaume: I mean, obviously there’s one is data sharing, so that’s one but to clear this off the table as part of DC, the one being kind of data sharing, they won’t be kind of private data sharing. So we’ll be having a look at the various methods, playing with table data sets and seeing what we can do and what the different methods are doing on target. So this is the need to do. So what can you say Nicole, what is your opinion? 

Nicole: Yes. So I would rather have a different spin on this, a positive spin for this. So I think we all know that decisions work in the pharmaceutical industry, that PSI and FSPI have lots of tools to break these boundaries. So I have to say that when I transitioned from academia to industry, I was really amazed how openly topics are discussed within the industries, across companies at the beginning, I couldn’t believe it at university people wouldn’t probably, of course, they would share their results, but they wouldn’t openly discuss scientific topics, but maybe also operational topics, or I don’t know, just sharing experience, how things are handled in their companies. I find this actually makes me proud to work, as a statistician, in the pharmaceutical industry. And I kept telling colleagues from other departments how we interact and tend to open this is and I think it’s a great opportunity to have this PSI, special interest group to discuss these topics across companies and to find common problems, common solutions and to share them also with the outside world. So I’m really very optimistic about the interaction that we’re going to have in the special interest group. 

Alexander: Yes, that is a cool thing. You know, we were incentivized by helping patients. We are not incentivized by producing publications. And so I think that creates a very, very different environment, yeah, which I really love and enjoy working and I think special interest group is an awesome platform for this type of exchange of participated in different six over the last 10 years and was always really, really great fun to see how everybody contributes, and everybody also shares, you know, the inside but also the challenges and when you see, ah, they all have the same challenge, I don’t feel as alone as its kind  of when you look sit at your desk and you think like why is that so hard? And so you see it’s the same for everybody. So speaking about the goals of your SIG, you have talked to put a little bit on this already. Do you have any kind of specific field you would like to tackle first? 

Nicole: So we’ve just, maybe let’s start kind of with the general goals that we have. So on the one hand, we want to explore all the hot topics at the moment that come about biomarkers. These can be really everything I don’t know about the multi-omics that you mentioned, the biomarker thresholds. Maybe the digital biomarkers that we haven’t talked about yet. And we are also looking for more colleagues to join, a really very wide range of topics that we can tackle. So that’s one thing really on topic level to discuss but also to increase the interaction with the other disciplines. So we always have a strong interaction with medicine with the translation of biomarker colleagues, biology academic research also to establish this in a broader range. And I think from all the topics that we already mentioned, it’s clear that biomarkers have really interaction with many statistical topics. I don’t know, high-dimensional data analysis. Modeling data over time, estimands visualization. I can probably go on for weeks and weeks and we really also want to connect with other special interest groups to have kind of the interaction with subgroup identification and predictive biomarkers with visualization and how we can use this for biomarkers. So these are kind of the three overarching goals I would think. So you want to add anything to that? 

Guillaume: I think that’s very good. 

Alexander: So if people want to join, who would be the good person to join the SIG? And I hope to move forward with this. 

Guillaume: People Google, PSI biomarkers SIG, the first link that is going to pop up is the webpage of the PSI biomarkers SIG and over there there’s only contact info, otherwise people can reach out to us, either Nicolai on LinkedIn and on what we also want to say that in terms of ways of working, while going to be meeting about once a month and all are welcome. And I think it is very important to us. We think about the diversity, we talk about the biomarkers definition, that it’s a broad topic, that is the broad definition, and then we’ve got the broad statistical analysis and about topics. So that’s why we think it’s important to have a diverse group of members. Because the more diverse that set of members, the more ideas would flow. And as Nicole saids,  all weak links at the moment of blind spots, you can say so are digital biomarkers and we think it’s a very hot topic and it’s going to be in my mind a very hot topic in the coming years. So if you have this, if you work with digital biomarkers, please feel free to reach out to us. Yeah, we would love to have you on board. 

Alexander: So speaking about digital biomarkers, these would be SIGs that come from for example labs. 

Guillaume: Yeah, biomarkers, we did not have in mind the past few years ago. So, usually when we talk about biomarkers, the first thing that comes to mind is genomics. So you’ve got the blood sample, for example, and then measure DNA. And that’s incredibly one biomarker. But then, when you talk about digital biomarkers, you can be very creative. So you can be wearables, but you could also be the modulation of the voice and picking up signaling pods and then, you know, it’s a new kind of biographies and obviously comes a new statistical analysis and you are branching towards so Nicole requiring the collaboration with other fields but if she thought into these kind of data, maybe you’re branching towards new Sigs as well. And exploring again, very uncharted territories. So that’s going to be pretty cool in the next few years. 

Alexander: Yeah, I’ve heard about a colleague who was working in that and she looked into different variables, to assess different things, and had an intern, and she was testing I think four different variables and four different looks, what shows on her veracity to kind of check how they work and what are the challenges kind of, you know, for example that say didn’t work when the wifi was gone or these kind of things, yeah. And so this, you can have quite a lot of fun with these kinds of things to find out how it was like, yeah, Nicole?

Nicole: Maybe culture to the digital biomarker field,I think they are also becoming more popular in the central nervous system. So I’m also working mainly on oncology but also in CNS and I think there is a huge potential also to define new endpoints and the two things I want to stress. So I think it’s not only about the kind of analysis point of view that you think, Oh, you have such high dimensional data. How can I reduce it? I think it really needs a good collaboration between statisticians and subject matter experts, I think we will not just solve this problem by, I don’t know, running a standard high-dimensional statistical model on this. So I think it’s very interesting, but it also has implications also on how we run trials. If we can measure the end points based on some speech analysis and people don’t have to come to the clinic to do long evaluations, I think that can really make a difference also. So it also, of course, has lots of implications, you need to get the data in house, you need to have an idea how to analyze them on time. So I’m really looking forward to talking a bit more about these kinds of other implications, not just the statistics behind biomarkers but also everything from data generation relationship to estimands. What do you do if people, if the wifi likes, that’s an interesting intercurrent event, so Wi-Fi breaks down. So I’m not sure if we can handle this. And there’s also special interest groups on these types of topics and estimates. So I think we’re going to have lots of interesting operations in SIGs.

Guillaume: Earlier, you talking about the validation of biomarkers, so when were,you know, I’m going to get something in those back person in the days, when we were shipping samples and then when we had them and I say provider doing the actual measurement, it had to be validated and then we have this kind of measurement that was it. Now if we tap it to digital biomarkers and Alexander you were referring to what’s this, then what does it mean? If we think of digital biomarkers that could be the step to come then does it mean that we need to develop and by way I’m referring to the pharmaceutical industry. Does it mean that we need to develop a device or can we tap into existing devices? And if we tap into existing devices, what does it mean validated? These kinds of things are going to be very fun because it’s very new and anyone is not shipping samples anymore. Every single person and every single patient is getting shot where the devices are but you always say.

Alexander: Yeah, yeah, yeah. That is really, really interesting. There was a point when you were talking about speech and variables, one other thing is pictures. I’ve worked for a couple of years in dermatology. Yeah, and of course, kind of changes on the skin, potentially you know can make a picture instead of you know going to the physician which I anyway already pretty overloaded with work. Yeah. So and then can you kind of, you know, assess the lesion or whatever you’re working on based on the picture? And I think there’s a lot of improvement in terms of what can be done in that phase. Especially also because of telemedicine and these kinds of things. I’ve worked with physicians in both Canada and Australia. Yeah. And sometimes distances are really a big hot topic. Yeah. Not everybody works, you know, lives in a crowded space like, you know, who is in the middle of Germany or and you know, such the spaces where, you know, you look around and you always see the next house. But you know, the next house might be, you know, lots of kilometers away, living alone to the next dermatologist.

Nicole: Yeah, yeah. I completely agreed with it. It also helps clinical trials or clinical development to be more inclusive because you can have measures may be locally like one topic may be going back from the digital biomarkers, maybe two. I think in general, especially specifically, if we want to measure biomarkers over time in a college we want to, I think people want to get away from biopsy base biomarkers to something that you can measure more easily and what you just mentioned, maybe imaging and something that you could do at your local hospital. And another very interesting open research field is blood-based biomarkers, we think of CT DNA, you can measure invitational patterns on blood and if this is kind of validated and it works, then you can just go to your local hospital or to your doctor, get the measurement done, and then you can have really monitor your disease over time without the need to get constant biopsies. So that generates lots of interesting data, kind of this high dimensional data, with all these mutations coming or showing up or increasing in frequency and so it’s an interesting statistical topic but it also has lots of implications and hopefully making patients life much easier and leading to new treatments and decisions. 

Alexander: Yeah, I think also not just within clinical trials but also beyond. If you also think about observational data, lots of these digital biomarkers that you’ve said are, you know, intrusive and so it’s a much lower hurdle to include it in observational research. Yeah and then this observational research can be much more inclusive in terms of patients who are also old. So that’s another opportunity. Very very good. If someone now  listens to this and says, Ah, I’d really like to be part of this adventure because that sounds really, really cool. What kind of qualification would you look for, is it, you know, just being curious? Is that enough to join the SIG?

Guillaume: That’s interesting because people always ask what marketing obviously has a data science background because what are you going to conduct that’s a single  method, handling biomarkers data. So people having a stats or data science background, that’s the first thing and obviously, having an interest in biomarkers. If you work with biomarkers data, that’s even better. And at the moment, the members of this SIG, of people who have a root model closely with biomarkers data. But if you just have a general interest in biomarkers and that’s another thing, you know, if you think I’m having an interest, but I don’t feel like I’m qualified enough in biomarkers. We can send you a DM in the newsletter. So, feel free to reach out to us. Whether you want to include the SIG or just want you to hear about, you know, what, we’re going to be working out. 

Alexander: Yeah. And I think what you’re going to be working on will be it’s a quite sexy in a way because when I hear you talk about, you know, this is high dimensional data and you know, there’s glowing between mobile data and collect data from all kinds of different sources, you know, as you called it multi-omics, which is a great great term. That means you need to, you know, you just can’t get away with let’s say a t-test or linear regression here, you probably need to be much more sophisticated. So can you talk a little bit in that direction, what you would foresee kind of topics to be? 

Nicole: Alexander, so I think most people would think of machine learning artificial intelligence because of all this high dimensional data and in a certain sense I would also, I agree so we need the kind of models that can handle this high dimensional data. I think there’s also more to it like when we talk about these pharmacodynamic effects, I’m thinking about maybe joining models to correlate biomarker changes to respond. So I think there’s not only machine learning out there. I think when it comes to surrogacy we need sophisticated models, existing models to apply them. I think there’s also a lot of insights in terms of latent class models that are also used in other areas. Sometimes, I think, biomarkers to me often, I see them as a kind of manifestation of something latent like we want to measure pathway modulation whatever this is so. And in a sense I feel it’s kind of similar to when we have patient reported outcomes when we also have all these questions that don’t imagine some latent concept and I kind of would also think that there’s an overlap to do this not really established models methods but methods come seemingly from something completely different field and clinical development. So I’m really looking forward to this and maybe, do you want to talk a bit more about how we’re going to identify the interesting topics that we want to talk about next month. 

Guillaume: So as Nicole said we’ve identified a couple hot topics or topics about what I heard, but we think that there are certain needs to come from the group. We cannot just be saying, oh, this is a cool two pictures, let’s explore it. So it has to go into groups, this is currently being gone. We are collaboratively identifying these hot topics or the topics that are of interest to us. And it’s short, you will pick a couple of these topics of interest and would spend a bit of time working on them, maybe doing a bit of literature review and then getting back together within the SIG and in terms of mid term goals. Yeah, I guess you are organizing a webinar by an interior and to talk about a couple of these hot topics and share with the broader audience what we found and what is the current state of the art.

Alexander: Good. Yeah. Yeah. And Nicole, when you were talking about a few roles that were on my mind earlier already because of this term of validation. Yeah, you have exactly the same topic with the questionnaires, is it stable? Yeah, across different physicians. Is it stable across different time points? How sensitive is to treatment, these types of things. And I’m pretty sure there’s a lot of, you know, similar statistical concepts. And in terms of validation, awesome. Thanks so much. That was a wonderful discussion about a really exciting topic. And when I hear that, I think I’d really love to come. 

Nicole: You’re welcome. So we meet every month. So just drop us an email to me or to Guillaume. And another opportunity to meet at least Guillaume or myself, is also the PSI in Gothenburg in June. So we’ll have a poster to advertise the biomarker SIG a bit more. And so come see our poster and then talk to us and everybody please if you want to join, or if you first want to get to learn a bit more, just look at the PSI homepage, we will make an effort to also summarize our meeting so that everybody knows what’s going on even if he or she cannot join at the moment.

Alexander: And we’ll put all the links to what we have just discussed including the links to Guillaume and Nicole’s LinkedIn pages into the show notes so you can find everything there. Just head over theeffectivestatistician.com. Thanks so much. That wasn a really great discussion and I’m really looking forward to seeing you all in Gothenburg later this year. For Guillaume basically home play given that he works.

Guillaume: And I’m like yeah I’m probably gonna bike there to tell you how far it is going to be.

Alexander: Speaking of that, what is the, you know, what’s so special about Gothenburg that we need to come to go.

Guillaume: Many things, I would say, it’s pretty incredible to be in the second biggest city of Sweden but still being so close to nature. I mean, it is, it is just insane. I used to live in Paris and, you know, the nature is not as close, obviously Paris is the capital city, but like Gothenburg, yeah, it is insane. Like, you’re in a big city. So you’ve got all the facilities of a big city, but at the same time, if you are an outdoorsy person, that’s a lot of fun. You can go kayaking, you can go rock climbing, you can go hiking within a very, very short walking distance. So yeah, a pretty good city. 

Alexander: Of course, it’s a sea which will also be really, really, really nice. Thanks so much for this great discussion and see you all in Gothenburg. 

Nicole: Thank you. 

Guillaume: Take care. 

Alexander: This show was created in association with PSI, thanks to Reine and Kacey who help the show in the background, and thank you for listening. Reach your potential, lead great sciences and serve patients. Just be an effective statistician.