7:44 PM, May 21, 2026
Edited 9:08 PM, May 21, 2026
Hello all! Apparently more people read this blog than I thought, as more than one person I’ve known in real life pointed out the rather personal nature of the blog post made around this time last year. I do not regret it, but I have deleted the overly personal portion of it. Now onwards to more exciting things, the future! There are two sort of life updates that I plan to share, but first I feel that it is very important that I brag about myself given that this is after all my blog, and so I am here letting the world know that at UPSTAT last month I was awarded the gold medal for my poster about clustering using a co-association matrix! This is a project that I have been working on in an on-and-off state for the last couple of years with my friend Luke Duttweiler, who now works for himself as a statistical consultant. I would like to follow in his footsteps someday, but in a sort of different way. I would like to follow in his footsteps after having worked in Academia or Industry for ten or twenty years, and I’d like the consulting work that I do to be directly related specifically to Clinical Trials, while the work that Luke does is much broader and covers a lot of areas. Anyway, here’s a picture of me with the winning poster.
Next, I am doing a data science internship in Hartford Connecticut this summer! While I wouldn’t quite consider myself a data scientist per se, my degree at RIT was in “Statistics with Data Science” and my work at Excellus was very data-science-y, so hopefully I won’t feel like I’m in waters that are too foreign, I’ll just have to brush up somewhat on my Python. The department that I’m working for is the Bond Department, so the only preparation I’ve done is watching Goldfinger, although I was also thinking of watching Casino Royale and Tomorrow Never Dies so that I am exposed to what I’ve heard (anecdotally) to be the three best Bonds. Kidding! Anyway, I’m not entirely sure what financial bond-related work I’ll be doing, but I’m sure it will be rather new and interesting, and could potentially provide an exciting alternative career path if I don’t end up working as a Biostatistician. Although, I heard recently they were planning to cast Christian Grey (???) as the next 007, so becoming Bond could be a long shot. I’m also very much looking forward to the fact that I’ll be working alongside dozens of other interns, perhaps more than than, since the size and scope of the internships and of Traveler’s hiring seems to be quite large. The company is putting us all up at the Hartford Residence Inn, which is only a short walk from the Traveler’s building, and the summer weather I imagine will be beautiful! Safe to say I’m very excited about it.
As for that career path in Biostatistics, I applied not very long ago for an F31 grant through the National Cancer Institute upon encouragement from my PI on the Environmental Health Training Grant, Sally Thurston, and from my advisor Dr. Tanzy Love. When proposing an F31, what you are essentially doing is outlining what you would like to do as your dissertation project! In order to explain the research that I am doing, or I ought to say the research that I will do when I return from my internship – which starts in less than two weeks (it really snuck up on me!) – some background information is in order. First, the personal motivation behind the project is fairly simple: I’d like to work in Oncology trial design since it seems to be the biggest and most impactful subdivision of trial design, and yet it seems to be applied enough to require more creativity, collaboration, and knowledge than strictly mathematical ability, which I don’t honestly think I have enough of to make any revolutionary contributions on the purely theoretical side.
After reading a lot of papers about Oncology Clinical Trial Design I learned the following: the typical setup of drug approval has three phases: phase I (finding a safe dose in a small number of patients), phase II (giving that safe dose to a larger number of patients to ensure the drug is statistically significantly better than some fixed level of efficacy) and phase III (a large trial comparing the treatment to control). For phase I, what is typically done is the 3+3 design, where dose is escalated across a number of dose-levels, 3 patients at a time, until a toxicity is seen, and then 3 more patients are assigned at that level. It is a simple-to-implement way to determine the Maximum Tolerated Dose or MTD. Other Phase I designs exist that are more efficient than the 3+3, but can be more complicated to implement. The problem is that the MTD is the best target for Chemotherapy, where you want to give patients as high of a dose as you can safely, but not necessarily for other Oncology drugs. For immunotherapies and targeted therapies, the MTD is less likely to be of primary interest: dose may plateau or be less effective with the more that is given. As a solution to this problem, a number of designs have been proposed, typically putting together phases I and II so that both efficacy and toxicity can both play a role in determining the optimal dose from the beginning, called a seamless Phase I-II design. These types of designs are typically Bayesian and (based on the literature I’ve read) can get relatively complicated – at least complicated enough that the clinicians themselves are not likely to fully understand them. One aspect of the models used in these types of designs is that there must be some way to model the correlation between efficacy and toxicity, and essentially my proposal is that instead of modeling this with a “gaussian copula” (which assumes there is some underlying normal distribution that is correlated between toxicity and efficiacy) as is done in many of the papers I’ve read, it might be more efficient/more accurate to model the pharmacokinetic/pharmacodynamic relationships between drug dose, efficacy, and toxicity. Pharmacodynamics and pharmacokinetics are sort of fields in themselves that are typically not integrated directly in dose escalation, although one paper I came across did incorporate a pharmacokinetic model in a phase I. This is all a very cursory explanation that I hope to explain in more detail in future blog posts.
Anyway, this is the current track of research that I am on right now, but regardless my overall goal for my dissertation (and perhaps my future career) is to find better ways to adjust dose in early phase oncology trials. Another aspect that my research seeks to improve is the use of continuous biomarkers in the dose adjustment, since using only ordinal or binary indicators of the treatment effectiveness (eg. in cancer you often have Complete Response, Partial Response, and Stable Disease) may take a long time to become available, so it might make sense to use continuous biomarkers that might be more quickly available (such as the count of cancerous B-cells measured by a blood test after the use of an immunotherapy for Lymphoma) that gives an idea of how well the dose is creating the response of interest in the body. I’ll find out over the summer how well my grant application did, and if it doesn’t get funded I will have the opportunity to resubmit in December when I’ve made more progress on the project. It’s all very exciting and I can’t wait to start the main part of my dissertation work when I get back to campus in the fall. No more classes! The last class that I had to complete for my course requirements was Longitudinal Data Analysis, which I passed in the previous semester. Honestly, as much as I’ve loved taking classes throughout the time that I’ve been a student at the University of Rochester and as an Undergraduate at RIT, the relief of not having to do coursework anymore is a huge load off of my back.
In future blog posts I plan to talk more about trial design stuff that I’m learning about in Cancer Research, but that’s it for now. Good night and good luck!
EDIT: The actor for Christian Grey, who is apparently named Jamie Dornan, was cast as the next Aragorn (big shoes to fill!), not the next 007. 007 is still up for grabs!