The first abstract is going to be presented by Dan Vail. I guess I can click on that. I get to click on it. Clicking is fun. Assessing the Promise of Endoscopic Pituitary Surgery, a Matched Analysis of Clinical and Socioeconomic Outcomes. Hello, my name is Daniel. I'm a medical student working with Dr. Ratliff and Dr. Lee on this project. I'll be talking about two different approaches to accessing pituitary tumors and basically looking at their post-operative outcomes. The motivation for our study is that there are these two major approaches, like broad categories of approaches to accessing pituitary tumors. Traditionally, the microscopic transphenoidal approach, and then the more recent, starting in the 90s, to increase in prevalence, the endoscopic approach. In 1992, there was a paper published that described a method for accessing three pituitary adenomas endoscopically. That wasn't the first time it had been done, but since that period, there's been this marked shift towards seeing an increased number of pituitary tumor removals done via this endoscopic approach. There have been a number of studies that have attempted to quantify the post-operative effects or comparing them between these two approaches. Some of them have found associations between various individual complications and the endoscopic approach or the microscopic approach. But overall, there have not been large findings on overall complication rates. Many of the studies that have been done are single institution studies that come to slightly conflicting conclusions. When you put them all together, you do a meta-analysis, and you largely don't find any consistent associations with the endoscopic approach and post-op complications. There have been some evidence that endoscopy is associated with different length of stay or long-term costs as well, but they also don't really pop up in the meta-analysis level. Our objective is to use a large multi-institution database, an admin database, to see if there's any association between either of the approaches and post-op complication rates or cost and quality indicators that we might care about, as well as the extent of tumor resection. Specifically, the outcomes we're looking at are 3D complication rates, mortality, length of stay, readmission, revision, cost of care, and likelihood of post-resection radiation therapy, which is what we use as a proxy for extent of tumor removal because it's an admin database. We don't have data on, like, the anatomy of the tumors, just sort of like a shortcoming of using these types of databases in general. We have our covariance that we control for, so demographics and comorbidities, as well as time, simply because there's this existing temporal trend in a shift towards endoscopy that might bias your results, and geographic region because, like some regions and some institutions will have institutional preferences for one approach over the other. And we run matched and unmatched models, which is if you, both of those include all the covariates, but if you run matched and unmatched models, you can then look for discordance between them, which will tell you something about kind of, like, the generalizability of your findings to the larger dataset. So a nice thing about running a unmatched regression, just like a straight regression of the covariates, without matching in the first place is that it maximizes the amount of your sample that you can use because you're not pruning people who, like, can't be matched. But the downside of it is that you're kind of, like, imputing linear relationships between people that might have, like, fairly different covariate values. And the benefit to matching is that you're quite confident that the people you're actually comparing are comparable to each other in terms of the covariates that you've specified, but there's some concern that you're going to prune out enough of your sample that you won't be able to extrapolate to, like, the actual average person that's in this population of people receiving treatment. The type of matching we use is called coarse and exact matching, which is sort of spiritually similar to propensity score matching, but it actually is, like, a dominant strategy in that it's more efficient. And a huge benefit of it is that you get to specify your covariate balance prior to running your analysis, whereas in propensity score matching, you take your vector of covariates, you collapse them down into a scalar that's your propensity score, you match, and then you have to check afterwards whether or not you've actually successfully balanced your sample. And with coarse and matching, you specify your balance, prune everyone else, and then you reweight the existing observations until you have comparable numbers of people in each categorical value that you care about. The cohort we use is drawn from the Truven market scan database. So we just identify people based on CPTU codes for either microscopic or endoscopic approach for these cellular lesions and end up with about 10,000 people in our full cohort, and then we trim that down to 6,500 after matching. For results, overall, we see the same trend within our data, which is drawn from years 2007 to 2014, that is described in the other paper, which is, yes, there has been this shift towards the endoscopic approach. The thing that kind of stands out to this is that you don't see, like, a big increase in the number of operations that are happening each year, at least within this database. So these surgeries that are being done endoscopically involve some substitution away from microscopic approaches. It's not just the case that, like, new people are coming on board that only practice this approach. Similarly, this is to, like, kind of get in front of one criticism you could make of the results, which is that it's possible that the selection of procedure type is associated with something about the patient population or about, like, something about the tumor itself. And unless you think there's, like, something substantial changing about the underlying conditions that are being treated in this time period, then that bias can actually account for the overall shift towards the endoscopic approach. What we find is that 30-day complication rates are actually higher in the endoscopic approach overall, and so the odds ratio is about 1.3 on this, and this is driven largely by neurologic and renal complications. This is a little bit discordant with some of the single institution studies that have been done overall, and so it's worth wondering, like, why that might be. One reason could be that many of the single institution studies that have been done focus on individual high-volume centers that presumably have more experienced physicians, and this is just every procedure that's being done in the Truven database, and it might well include people who have, like, recently shifted over to performing this different type of procedure. Unfortunately, Truven stopped including physician identifiers after a certain year, so we can't actually identify whether or not individual doctors within this sample are, like, shifting their practice year to year. Kind of notably, even though we see this higher overall complication rate with the endoscopic approach, we don't see that carry over to the quality indicators, like the really top-level ones that we would care about in the hospital, so this does not translate to, like, higher mortality, for example. It also doesn't translate to higher readmissions or rates of surgical revision. It doesn't translate to increased length of stay, which is different than some of the existing studies that have been done in smaller samples. Also, kind of notably, some studies, including simulation studies, had predicted that you would see higher long-term costs with the endoscopic approach, and we don't actually see any evidence of that. The initial inpatient admission is, like, $3,000 more expensive, so that's about 10% of the, you know, 10% more expensive than the microscopic approach. But at the 90-day and 180-day check-ins, the results are basically identical for costs. And then our final big outcome that we're looking at is we see that in the endoscopic sample, fewer people, or people are less likely to get post-resection radiation. And again, this is our proxy within this admin database for extent of tumor removal, since we don't actually have those, like, tumor-level variables. So this suggests to us that perhaps the endoscopic approach is associated with less residual tumor afterwards, maybe a more complete resection. You could argue that perhaps what this is indicating is that the people who are getting the endoscopic approach are involved in more, like, thorough but perhaps more invasive surgery as well that might go hand-in-hand with complication rates. So, yeah, overall, the conclusions are we see about 30% higher likelihood of complications with the endoscopic approach, which does conflict with some previous single-institution studies that were done in high-volume samples. So this could be confounded to some extent by, like, surgeon volume and training. That's not necessarily – it's hard to say, like, who that confounding effect would fall on, like the single-institution or multi-institution data set, in that this is probably a fuller picture of how the procedures are actually done and that this is every single one of these procedures that's intruven, but maybe it's not a fair picture of how the complications rates that you would see in an endoscopic approach done by, like, a more trained surgeon. We see some evidence that the endoscopic approach is associated with, like, greater resection of the tumor. We don't find differences on the quality indicator side, and we see no differences in long-term costs. So, as I mentioned before, a major limitation of just working with an admin database in general is that you don't have variables at the level of tumor anatomies, but that's just inherent in using this type of data. And for further work, if you had another admin database that had surgeon identifiers, a really helpful thing to do would be to look at transition over time and see if complication rates are associated with, basically, surgeon training time. That's it. Thank you. Thank you.

endoscopic pituitary surgery

microscopic transphenoidal approach

complication rates

cost of care

post-resection radiation therapy