Please tune in to Episode 5 at Apple Podcast, Google Podcast, Amazon Music, Spotify, and Firstory

Transcripts:

Intro:

How are real world evidences generated, collected and used for health economics assessment? What are some biostatistics related roled within life science consulting? How will the life science consulting industry potentially evolve in the future?

Tim Disher is here to answer all these questions for you! Dr. Tim Disher obtained his PhD degree in Nursing from Dalhousie University working on economic evaluation of neonatal health technologies. He is now a registered nurse and serves as senior director of Biostatistics at Eversana. His expertise includes developing and critiquing models from both statistical and clinical point of view. He is also interested in the latest tech developments and believes they can eventually assist and change the life science consultancy workflow. Let’s dive into this episode to see what Tim shared with us!

Jocelyn: Thank you, Tim, for coming to our biostatistics podcast. And it’s really great to have you with us here.

Tim: Awesome. Thanks for having me. I appreciate it.

Jocelyn: Let’s get started by you telling us a bit about your background and how you became interested in biostatistics.

Tim: Yeah. So I’m a nurse, I call myself a nurse scientist, I guess, or a decision scientist. I did my PhD in nursing and before that I did my undergrad. I worked on a mother-baby unit, which was a lot of fun. And kind of all throughout kind of the undergraduate nursing training and my time on the floor, I was really interested in kind of how research gets into practice and kind of how we use studies to come to conclusions about ways we should treat babies or families. And so that’s why my PhD, I wanted to go in and do a very like decision science focused PhD. It was actually like a health economics study, like micro simulation, everything. And so I just kind of, I like statistics as a way to just have a structured way of looking at data and looking for patterns and data, coming up with some decision rules and then acting based on what you find. So that’s really, it’s more of like, I’m a very applied person, I guess. So nobody’s coming to me to help find the like really nifty way that this nonlinear formula can actually be expressed in linear terms. That’s not me, but I’m more of a, how can we kind of approach this in a smart way? And what are the different ways that this could work out and kind of thinking more at that higher strategic level and more of like a translational level.

Jocelyn: I mean, applications are really important, right? Especially when it comes to decision making. And um, you mentioned you are doing a health economics related job. Can you tell us a bit more about what you’re doing currently?

Tim: Yeah, so I work as a senior director of biostats at Eversana, which is a global life sciences consultancy. So we work with pharmaceutical and medical device companies to help bring products to market and then to help kind of maximize their market share. So we’d say we do a lot of everything after the trial, the pivotal trial mostly. So we’ll do a lot of post hoc analysis. We’ll do some analyses to support as like additional analyses to support like an FDA or EMA submission. But the group that I work in primarily is focused on this really small piece of the pie called market access, which is where like in Canada, once you have your Health Canada approval, you then have to create a submission to CADATH, which is a health technology assessment agency. And they look at the clinical effectiveness and the cost effectiveness and a few other pieces to help make a recommendation to the provinces. All the provinces except for Quebec, Quebec has their own, but to make a recommendation to whether the provinces should list therapy or not. So we do a lot of the statistical support for that. And then we also have a team that does a lot of the health economics and I’m kind of in the middle. So I’m a lot of the like, how do we take the results of the meta-analysis or the indirect treatment comparison or like the like IPTW analysis or whatever, like real world data and turn that into a way that the modelers can use it to kind of tell the story that they need to tell. Jocelyn: I see. That’s interesting. So what would you say that the biostatisticians’ role in this kind of, I guess, industry, is?

Tim: Yes, it’s pretty mixed. And I think it’s kind of specializing in some ways. So I am maybe in some ways a dying breed because I kind of have my fingers in everything. I am considered more of a generalist. So I do the strategic piece, I help with the design. I still actually do a ton of the coding on the newest stuff. I do a ton of coding adaptations. So like a new paper comes out, this is a neat method that applies to our use case. So I would do a lot of the adaptation of that. And then I would kind of be supporting, just recently was supporting like a regulatory submission. I’ve done some support of our real world data stuff. So I kind of do a little bit of everything. But I’d say generally we would have people mostly focus on one of two areas, I’d say. So they’re either kind of like doing a lot of what we would call real world evidence generation or real world data, which is really just electronic health records and claims based analyses. So like routinely captured data for the most part. So a lot of people doing study design and analysis around that. And then we’d have the people that are doing more of the traditional market access, which is primarily like network meta-analysis and different ways of combining aggregate data, like study level data and individual participant data, but focusing mostly on trials. I’d say we kind of have like a third stream that’s a little bit more like data science, I’d say. So like some of the work that we’re doing there, for example, is like we have a really interesting project in a rare disease where we don’t have a lot of data on, we don’t have a lot of patients. And so we can’t do like a lot of predictive modeling in terms of like, does like X predict Y or time to event or anything like that. But we do have a lot of like serial measurements of different biomarkers. And so we’re kind of looking at can we use the bits of the data where we have a lot of it, which is these repeated measures to start kind of looking at maybe different phenotypes of trends in these biomarkers. And let’s say, okay, these are the people that need to be treated more quickly. And these are the people that kind of could be deprioritized. Or maybe it could be the opposite. It could be like these people with these kind of more lazy trends end up going untreated for longer. And so there’s actually more damage there. And so it might be an important subpopulation that needs to be treated quickly. So that’s kind of, that’s like all like, that’s more like longitudinal clustering and like K-means stuff. And so I would kind of put that more into just a more data science-y type area. And same with like, we have some like, like web-based tools with Shiny that are, I’d say that’s kind of more within like that toolbox. So taking an analysis and then making it so like a client can fiddle with the inclusion exclusion or some like sensitivities, stuff like that.

Jocelyn: I see. That’s really interesting. I’m wondering for you yourself personally, which out of these three, which is your favorite part to work on?

Tim: I really love still getting everything into the health economic model. So I’m kind of at the like end of the pipeline of, I still like the most working on those projects because I just, it’s kind of fun to take, it’s essentially like a gigantic multivariate analysis that you’re reducing down to putting through this gigantic loss function and saying like, what is the life years gained or the quality adjusted life years and kind of what are the costs associated with that? And so it’s a lot of fun to kind of build these things and make what is really just like a gigantic Frankenstein machine. I see. Lots of like fun stuff where it’s like, Oh, like if I look at this, like, it’s like, Oh, like I have survival data, so I might have hazard ratios from a model. I might have another study where all I have is like counts and exposure time. And so just like, it’s fun to be like, okay, well, yeah, if I actually just do like a complimentary log-log transformation, do a model on these, and I can combine my hazard ratios from that and my hazard ratios for the rest. Now I have this kind of neat multi-parameter evidence synthesis model. So it’s kind of like puzzle, putting together puzzles. I really like that piece.

Jocelyn: And you did mention micro simulations when it comes to health economics. I’m wondering what do you think is like the biggest challenges for micro-simulation?

Tim: So I’d say the biggest challenge right now is that prepare to be shocked if you are not like deep within health economics. But I would say, I think almost, I would say a hundred percent actually, maybe of submissions to CADA with these health economic models that there’ll be like a big mix, right? So there’s like cohort, like Markov cohort models, then there’s micro simulation models and discrete event simulation, which is kind of like a super fancy micro simulation in a way, are all developed in Excel.

Jocelyn: Oh, what? Really?

Tim: Yeah, the whole thing. Every, every, the whole micro simulation, everything about it is all written in Excel for the most part with minimal VBA, like there’s like VBA for the iterations, but a lot of it has to be in order to be transparent and to meet the requirements. So these things, you have this model that takes like, if you were, I think we have a poster going into ISFOR that we’re waiting to hear if it got accepted or not, but where we actually said, okay, like if you wanted to have like plus or minus 5% precision on like net monetary benefit for this micro simulation model. And the base case for CADA is always that you have to account for the uncertainty in all your parameters that go in. So our treatment effectiveness, they call it like a probabilistic sensitivity analysis. That’s the base case. So essentially you’re like replicating your micro simulation model, like K times. And so we did this whole piece, it was like, okay, here’s a real micro simulation model we have that has been submitted in the past, to be submitted in future submissions. What is like the optimal balance of like the size of the micro simulation population and the size of the PSA population, like those two levels of uncertainty, and then how long would it take to run? And it’s basically like, it was like two years or something like that. Oh my God, that’s insane. So I’d say the biggest challenge is like a structural one in some ways that it’s about like at the end of the day, this is going into Excel. So at least right now, I think there’s some movement towards like other software. So R is probably the closest one to people doing more submissions in. But then like once you open up the door to R, it’s like there’s people that code these in like Java, not JavaScript, but Java. There’s people that will code these in or have components of it written in C++. Because then that’s just like way faster. Now there’s Julia, which is like another statistical programming language that’s much, much faster than R and Python even. So yeah, I think in the short term, the biggest challenge with the micro simulation models is really deciding when you need one and coming up with a clever way to make it as fast as possible with the limitations that you have.

Jocelyn: So when you’re talking about coding, you’re mainly talking about coding in VBA?

Tim: That would be like on the health tech side, the biostatisticians don’t do any VBA. We work exclusively in R really. We have some, we do keep like some SAS and Stata licenses because every once in a while you need to reproduce an analysis that was submitted for regulatory approval. Some of the algorithms are slightly different between R and SAS. And so you have to kind of go down the roof to figure out like, am I really getting a different result or have I just like not understood some data processing piece and that’s what’s coming through. I see. Oh, that’s very interesting. SAS and R have different default methods for quantile estimation. So like even something like making a summary table, it’ll be different or it can be slightly different and it all can come down to the software. I guess when it comes to regulation, like with the submission, SAS is usually used in the pharma industry.

Jocelyn: So would you say, what would you say is the standard requirement when you do health economic related submissions?

Tim: So for health economics submissions, it’s like really like the kind of like my core software set that I use when I’m supporting that from like the statistical side, my mindset is R and it could be any program, but R because that’s what we all work in and Winbugs or JAGs. So a lot of the, a lot of the HTA, like health economics submissions are, the sentences portion will be from a Bayesian standpoint for the most part. So having some like Winbugs JAGs experience is always really good. Other than that, that seems to be where people are more or less coalescing. There are some like status shops out there, but yeah, I’d say like the industry statisticians mostly work in SAS still. There’s some that are moving to R, so I think GSK and Roche both seem to be doing a lot of stuff with the folks that, what used to be RStudio.

Jocelyn: Yeah, sounds like so. Yeah. You mentioned you work on, a lot of these are done in the Bayesian framework. I’m wondering why, like, does it have to be done using a Bayesian framework or is it just because it’s like very flexible?

Tim: I think it’s mostly because it’s very flexible. So like, like I think it’s a bit of like a momentum thing. So when the very first, so like pretty much every CADA submission or every like HTA submission or like Haltech for like market access viewpoint will be in an area where there’s like many different comparators and you need to have one analysis that accounts for like the correlation and all of them so that you get like a single consistent comparative effectiveness assessment and everything kind of like, like all the uncertainty is properly accounted for as you go through. And so the way that we do that from like the efficacy endpoints to input is with a network meta-analysis, which is just a meta-analysis that’s generalized to multiple comparators. And all of the original code for that was written in Bugs because if you can write down the math for something, you can usually write down the Bugs model for it, which is what’s like what makes it so fun and so great and so easy to work with and flexible, like you mentioned. And then, so now there are frequentist alternatives, but like the, it’s kind of like the same thing with like SAS and industry, like all of our code is based on the Bugs code, all the bits and pieces that are like modular that we reuse over projects are all written, assuming that like the output is coming from like a Bugs model that we control every aspect of it. So we know what the names are for things.

Jocelyn: I see. But even, I guess, so the computation, for example, the improvement on the computational efficiency using, sorry, frequentist methods is not, like they don’t really take into consideration because everything is done in Bayesian.

Tim: Yeah, so here’s the thing. So for the computation piece, yeah, like you’d be right. Like if you were actually refitting the model every time, then it would make way more sense to do like, to use like a frequentist piece because it’s like more or less instantaneous. But what the Excel models are actually using is one of two things. So for the uncertainty and the comparative effectiveness piece, so like the treatment effects or odd ratios or hazard ratios or like absolute responses, however, whatever ends up getting put into the model, that comes in one of two ways. Either you give them the actual, like all the posterior samples from Bugs, and then they just sample one row at a time. Right. And that’s kind of nice because then there’s no distributional assumptions. They don’t need to say like all these parameters are multivariate normal or something like that. Or the other option, you can do it in either a frequentist or a Bayesian link, is to give them all of the treatment effects and the variance covariance matrix for them. And then they will do like a hand coded Cholesky decomp in Excel to sample from that multivariate normal. So either way, like at the end of the day, the health tech model is Bayesian in a way. It’s basically like an MCMC type thing for the parameter uncertainty. So no matter what, even if however we fit the comparative effectiveness support from the model, the bottleneck is always on the health tech side.

Jocelyn: I see. I’m wondering, because there are programs in universities called health economics, right? How is that differentiated from what biosciences are doing in terms of health economics?

Tim: So I would say like, the people that I work with that have done health economics degrees are usually focused on actually doing the more standard approaches to everything. And then their fancy bits are around thinking about how to build their, whether this is a partition survival model, or whether they’re going to do this as a continuous time markup, or whether they’re going to do it with tunnel states. They end up being really strong there and really strong in the logic and understanding where all the different bits and pieces should come from. Whereas the biostats people, when they get involved in the health tech, it’s usually because there’s some complex piece, complex input that’s not easy to find. And so they’re extending it in some way. So they’re doing maybe fitting a slightly more complex model, or they’re changing, rewriting WinBUGS code, et cetera. But they can overlap quite a bit. I’d say the one thing that I love about industry and consulting is at the end of the day, it’s just like, as long as you can do the work, that’s all who cares about. Credentials are useful for like getting you in the door, and they’re useful for giving you the tool set that you work from and like a theoretical framework that you work from and like how you have learned to approach these problems. But then from that point forward, it’s just like, we need someone who can figure out how to get this estimate that we need for the model. And if it’s a biostatistician who can figure it out, and that’s great. But if it’s like someone who did like an biotech, who can figure it out, because they found like some paper from the 70s that has this like neat application and they’re like, yeah, I have to do is like apply this formula and everyone’s done it. And yeah, it’s an approximation, but it’s also a better approximation than nothing. So yeah, so I’d say like, there’s no clear delineation in the way that there might be in like, real like pharma industry where you have like, statistical programmers, and statisticians and the statisticians have like a regulatory like role. And they’re more, they’re not really doing coding for the most part. They’re really more just like planning, writing and interpreting. And then like doing the final QC and sign off and all that sort of stuff. But right, or about everyone does whatever, does we try to keep people kind of within their specialty, but at the same time, it’s whoever can get the job done at the end of the day.

Jocelyn: I guess speaking of credentials, for students, or people who want to switch careers, who want to do something that’s related to what you’re doing now, what advice do you want to give them? Or what do you wish that you knew when you first started this whole gig?

Tim: I’d say like, understanding, like really understanding trial design, and trials has been like a thing that has been a huge help to me. So I did a lot of work in clinical trials, I still work on some like, national clinical trial work, again, and anyways, and I like that, I can always go back to like, understanding so much about industry and so much about evidence and where things need to come from, because everything I learned from the trial piece, so like really having a good understanding of like experimental design, I find is like a bit of a superpower. There’s not, there’s a lot of people who are really good at like, doing the analysis once the design is made, but like really understanding the design and like, what if I need to figure out like, like, this slightly different version of this research question, like, how does the design have to change? Like, those are the things that stump people, I find. And then the other big thing is just like, I don’t know if it’s like a valid statistical concept anymore, because that field seems to kind of throw everything away every 20 years. But when I was going through my undergrad, there was like this concept of grit, that was maybe like Angela Duckworth, I think, when I was in first year of junior high, she might still be doing research. And it was just the ability to, like, continually plug through even when you are in this like, overwhelming, uncomfortable situation, being stretched beyond your knowledge, and really having to like, persevere and grow and find things like the people who have that, I think are the people who I see succeed a lot. To the hardest thing about consulting work is that for as much as some aspects of the job become modular, like every project is very, very different and new. And you could be working in oncology today, and you could be working in some rare disease tomorrow. And the expectation is that you’re going to be able to kind of get up to date on the medical side, and like biology and everything. And also kind of everything that’s been done from a statistics standpoint on that. And so that can throw people a lot. So you don’t really have templates that you can just do over and over and over again. Like some like CRO type work, you end up more just like working in the same area over and over again. And you can kind of build up, like you really become comfortable. But consulting, I think, is defined by being uncomfortable. That’s really good advice. I guess. You have to like, you have to really like that. Right. It’s probably applicable for anything that you want to succeed in to constantly grow.

Jocelyn: But I guess when you talk about consultancy, is this statistical consultancy? Does it cross any path like the regular consultancy when you face the client?

Tim: Yeah, so we kind of have two, like two general teams, people that are more technical that do like minimal client facing and then people like myself who are kind of in the middle. I would. So I’m responsible for business development. I do like a lot of contracting work. I do a lot of sales work. I do like presentation to clients, I consult with clients in terms of like, what is your problem? And how can we help you solve it? Presenting results and troubleshooting results and kind of responding to things live. And we do a lot of training and stuff like that as well, for clients primarily. So those are the two like mainstreams, I’d say, the technical people will still be client facing, like they’ll still present results at times, but generally, it’s kind of like the results are more or less kind of passed through the chain, I’d say. I see. Does Eversana mostly face Canadian clients? Or is it the whole North America or even just global? Yeah, so Eversana, the company is like huge. So I’m only talking about like, so there’s like Eversana, this gigantic, like, there’s like a whole data analytics team that we haven’t talked to it at all. It’s all like claims and like market research and a lot of EHR and like really interesting things that they’re doing that’s kind of outside of our group. And then I’m in this group called professional services. And then within that group, you come even smaller. And our, like our, our group is Valiant Evidence, and we do like a lot of market access. So we work with people globally, I’d say like a pretty good mix of Canada, US, UK, I’ve done some APAC work, so like, Australia and Asia for market access support, and like some European, some small like German and French stuff as well, Italian, Spain. So we kind of work everywhere. And then kind of Eversana as a whole has like offices in India, offices in Europe, offices in the US and Canada. And so it’s really kind of a, like, it really is a global consultancy on the on the huge scale. And then even on our small scale, it’s still kind of global. But I’d say most of the people in our group are based in Canada. I see. Yeah. There’s about 115 of us now, I think. That’s a lot. I’m in Burlington, and then I’m in Halifax. And there’s a few people up in Sydney, Nova Scotia.

Jocelyn: I see. Do you guys go to travel a lot to talk to clients?

Tim: There is the option to travel a lot. A lot of people have opted not to travel much. Especially like there was a pretty big change pre post COVID. For sure, there’s a lot more travel early, but like, we’ll send, like, at least a couple people to conferences. So all the bigger conferences we will go to, we’ll go to the Cata Symposium, that’ll be in Ottawa this year in May. ISPOR, which this year is in Boston, I think also in May. And I think we sent a team to ISPOR Europe. So that was in… Why am I blanking? I was in Europe this year. I can’t remember exactly where.

Jocelyn: I see. So mainly the travel and rent is to the conferences.

Tim: Yeah, mostly the conferences. We used to do some client stuff. So like a couple of my friends that I work with, like went and like presented to offices in like the US mostly. I went to like a product launch in Denmark once.

Jocelyn: I see. That’s pretty cool.

Tim: So it’s neat. If you like travel, I’d say it’s not as much travel as management consulting would be. For sure. That’s defined by travel. But there’s like the ability to get some anyway.

Jocelyn: I see. I guess it’s a really good combination of like the technical things statisticians do and then some bit of travel to either clients or conferences. Sounds really cool.

Tim: Yeah.

Jocelyn: So how do you think this industry will evolve in the coming years compared to, I guess we’re not talking about pharmaceutical company, pharmaceutical industry, maybe just the consultancy industry?

Tim: Yeah, I’d say it’s like, I don’t think of the major ways. So I mean, like, from a day to day, I think it’s pretty clear that the big large language models are going to probably change a lot of people’s workflows. So like, chat GPT, very famous, everyone’s talking about it right now. There’s like Google bard, those are only continue to grow and like, their ability to take something like a lot, like something that a lot of statisticians struggle with, for example, or at least the ones that I’ve worked with, is taking the very specific estimate that you are providing an estimate for research question and the model you did it and then explain it to a room of people that is like a mix of executives, non technical people, right, but still have enough of the technical piece of the technical people in the audience don’t think that you have done a bad job. And like, it’s pretty undeniable that chat GPT can take something that’s very technical and turn it into something that’s a lot better. It’s a really good first pass at something for a lay audience. And so I think things like that are going to become pretty integrated and workflows. Interesting. Does a good job of like creating structure for different types of reports. For like doing some grammar checking for for helping to like think through in some ways, it can be better than than the internet for finding like a really specific, like, you need to know like in our how can I take this multivariate distribution, where I have my variance covariance matrix and a vector of means I want to express it in terms of like a series of conditional distributions, like how can I do that? Can be really difficult to find a straight answer for that, like, exchange, or like to find the right textbook that has like what you’re looking for. But chat GPT can get you moving and get you in the right direction. It’s not always going to give you the right answer, but it’ll get you kind of understanding where you need to look next, I find. So I think that’s really going to change the way that people approach a lot of problems. So I guess companion, when you’re stuck coding something like, I mean, I’ve used to use it to be like, how can I make this for loop more efficient? How can I make this like mapping function more efficient? And it can give you some, some neat ideas and some estimates on what the time savings would be. And so it’s kind of cool. I think it will, I think that’ll end up changing. I don’t think it’s gonna replace anybody. I don’t think it’s gonna do a lot for our workforce. But I think it’s gonna be a non judgmental companion to ask some questions that you might be stuck on. I guess it’s eventually helping with all of the industries a lot.

Jocelyn: And I guess, especially in the healthcare consultancy will help more with the communication side of it. And resource gathering.

Tim: Yeah, yeah, definitely. Like on communication and just relieving some of the cognitive load, like one of the things that I find a lot of. So like, we do everything right. So like when like our statisticians are also like writing SAPs, and they’re contributing to reports, and they’re making slides and everything else. And I find that like, we can be really efficient on the analysis side, but sometimes we’ll get stuck on like, the background for a disease area. And you can just like throw a few abstracts of papers together, say like, get a background for this disease area, then you have a starting piece that you can then edit and build out and like, go and like reference check and everything else. But at least you’ve like taken that cognitive load of like, what’s the first thing I need to talk about about this disease? A lot of people just like freeze at their keyboard.

Jocelyn: That’s so relatable. That’s how I feel when I write my first manuscript. I had no idea what I should put in the paper. Yeah, that sounds interesting. And, thank you for all the advice and information I share. And we’ll conclude this podcast episode by this question. So what is one question that you wish I asked? And how would you have answered it? Oh, or just anything that I haven’t asked about, and you want to share about your work or your insights?

Tim: Yeah, I guess maybe would have been good to talk a little bit about like, what things are kind of not so great about the job sometimes, I think on some of the things that maybe that people don’t, that people typically don’t like. So like that, like really being uncomfortable and always really having to be ready to think on your feet. And every project is just different enough that you feel like you’re starting over a little bit. But I guess the only other thing like someone coming into consulting probably have to be prepared for is like, and I kind of gets tied into that is that there isn’t quite as much like, especially if you’re coming from an academic, there’s not so much like formal training where it’s like, you’re going to go to a workshop to learn how to do this analysis. Or like, you’re gonna go and do like a two day, like causal inference workshop. For the most part, it’s going to be like, here are the key resources. And you have to get like 80% of the way there and you need to get there and like, within a timeframe that works for the people that are working on the project. So I think something that would be great would be finding a way to, I think back when we were all, I mean, I was always remote, but like, I think when people were working more like in an office, you’d have a lot of informal conversations and training and things. And people kind of mentoring you on the spot, like if you get stuck, I think we’ve struggled to find a way to replace that sort of thing. And I find that people are now getting a little bit more stuck on the training piece. And trying to find a solution to that. I don’t know what it is at the moment. But yeah, I’d say the biggest challenge for for people to watch out for is like if you don’t like to really teach yourself and learn kind of on the fly and be pointed in the right direction and kind of say like, that’s where we need to go over there. And then we need you to help us find how to get there. Then I can, people can really struggle.

Jocelyn: I see. That’s a really good point to think about. So yeah, does it sometimes bring stress to work?

Tim: Yeah, well, I mean, I don’t find that stress. I like to like self taught myself everything. Like my, my, my, all my stats training, everything was all just within like self directed courses, primarily my PhD. So like, I kind of grew up in that. So I get it. But a lot of people do not like that at all. So audience for who, for whom that want to be in this industry eventually make sure you’re in the industry. Eventually, make sure to be prepared mentally.

Jocelyn: I guess that’s all my questions for today. And thank you for joining us on this episode of the biostatistics podcast. And maybe we’ll invite you back some other time to talk about some more new projects that you work on.

Tim: Yeah, absolutely. That’d be a lot of fun. Thanks for having me. I appreciate it. It’s good to have to think of all these things, instead of just doing all the time. So thank you. Appreciate it.