Thanks, Dr. Kaiser. Thanks to everyone for being here. My name's Raymond. I'm one of the categorical residents at the Children's Hospital of Philadelphia. And today, I will be presenting work mentored by Matt Kirshen in the critical care division. I won't read the title again. I have no disclosures. Firstly, let's review pediatric arrest epidemiology for the audience. There are approximately 20,000 arrests per year in children in the United States, with mortality figures ranging from roughly 5% to 30% survival in the last 10 years, based on arrest location in hospital versus out of hospital. Among those who survive to discharge, neurologic injury is, of course, a common and very feared morbidity, with a recent analysis of the THACA trial showing about a 30% to 65% prevalence in arrest survivors of at least moderate neurologic injury at discharge. It's been well established in both adults and kids that hypotension is an important contributor to secondary brain injury during the post-arrest period. And for this reason, the AHA recommends using arterial line monitoring and targeting a systolic blood pressure greater than the 5th percentile for age to prevent hypotension. Looking at some of the existing literature that's been done in this space, an important consideration to take into account regarding the study design of existing studies looking at blood pressure management in kids after arrest is that the vast majority of studies done in this space define hypotension via the presence or absence of a single blood pressure measured non-invasively, usually less than the 5th percentile for age, during a pre-specified time period after ROSC. What this means is that hypotension in these studies is characterized as a binary variable. Either a patient has hypotension or they don't. And it does not capture what could potentially be significant differences in the duration or the magnitude of hypotension experienced from patient to patient. We sought to answer this question because the knowledge gap here is that very little evidence exists looking at the duration or the magnitude of hypotension and whether that specifically is associated with outcomes more so than just the presence or absence of hypotension alone. Our strategy here was to tackle this question using arterial line data, which provides much more spatial and temporal resolution than non-invasive blood pressure measurements do. So for our first aim, we calculated a numeric score, which from here on we'll call the burden of hypotension. And this is a numerically calculated score that incorporates both the magnitude and the duration of hypotension that a patient experiences. And we sought to determine the association between burden of hypotension less than the 5th percentile for age in the first 24 hours of ROSC and any association with unfavorable neurologic outcome at discharge. Next, we sought to determine the variability in association between burden of hypotension and neurologic outcome as we adjusted the threshold mean arterial pressure we're using incrementally up from the 10th to the 50th percentile for age. Finally, we asked the question of whether association between burden of hypotension and outcome would differ between patients who received less than or greater than 10 minutes of CPR. To get into more details about the methods, this was a single center retrospective study. It enrolled all patients who were admitted to the CHOP-PICU for post-arrest care or who had arrests as part of their PICU course, extending all the way back to October of 2018. They had to have an arterial line in place and they had to be connected to the Moberg multimodality neuromonitoring system. We had no exclusion criteria and our primary outcome was favorable neurologic outcome at discharge as defined in accordance with previous literature as a PCPC of one or two or no change in PCPC from pre-arrest baseline. For each patient, we set our time equals zero at ROSC and from that point on, collected up to 24 hours of mean arterial pressure data sampled at a frequency of 100 hertz. You can see an example trace up there in green with eight hours of recording data. Next, we used a spike detection algorithm to remove noise and artifacts in an unbiased and automated way. We populated our normative age-matched percentile map values based on published data, looking at an inpatient pediatric cohort. Let's get into our definitions now. So first, for the duration of hypotension, we're defining that intuitively as the percent of time, total recording time that a patient map was below the threshold map. In this example, you can see an eight hour recording with the last two hours or so being hypotensive for a duration of hypotension of two over eight or about 25%. Then we define the magnitude of hypotension. For this, we took all times during which the patient was hypotensive, and during those times took the median difference between the patient map and the threshold map. Again, referring back to the same example, taking the difference between the red and the blue line, anytime the red line is below the blue line, and then taking the median of those differences. And then finally, burden of hypotension. We define this as the area under the curve of the patient mean arterial pressure waveform below the threshold mean arterial pressure. This of course, by definition, includes both duration in its x-axis as well as magnitude in its y-axis. Notably, we did normalize this number by the total recording time, such that two patients who have different lengths of recording for whatever reason, but who are similarly hypotensive over time would have similar calculated burden values. Final conceptual point I wanna make before we get into the data is the effect of changing the threshold mean arterial pressure on the calculated burden. You can see in the upper graph and the lower graph representing the same patient and the same mean arterial pressure tracing. As we move in the upper graph from the first percentile mapped for age set as a threshold down to the 50th percentile, we can see that even though the mean arterial pressure doesn't move because of the change in definition of what we consider to be hypotensive, we have a much higher calculated burden. And with all that out of the way, let's get onto our data. So we analyzed 108 patients altogether. Of those, 66% had an unfavorable outcome. The median delay to recording onset was 90 minutes from ROSC, and the median total duration of monitoring was 21 hours. There was no significant difference in sex, mean age, race, or ethnicity between our favorable and unfavorable groups. There was roughly equal distribution between in-hospital and out-of-hospital arrests in our cohort. Again, no difference between unfavorable and favorable groups. And then finally, as expected, a significant difference between the median CPR time, the first lactate after ROSC, and the first pH after ROSC between favorable and unfavorable patients. Our first aim was to look at burden, duration, and magnitude between favorable and unfavorable patients. We can see the three graphs below with box and whisker plots, and MANT-WINI tests were used to differentiate between the two groups. We can see that for burden of hypotension alone, but not for duration and for magnitude, there is a significant difference in the calculated score between favorable and unfavorable groups, with unfavorable patients on average having higher median burdens than favorable patients. With that result, we next sought to extend our analysis by replicating the same calculation, but modifying our threshold mean arterial pressure. We initially tried the 10th percentile, and then increased incrementally up to as high as the 50th percentile for age. I'll draw your attention first to the leftmost graph. You can see that, number one, as we increase the percentile from the 10th to the 50th percentile, our calculated burden score increases as expected, but that there's no difference between this increase between favorable and unfavorable groups anywhere from the 10th to the 50th percentiles. Of note, there is a significant difference between favorable and unfavorable groups when we look at median magnitude alone, but that only applies for the 10th and the 30th percentiles, and these data are not adjusted for multiple comparisons. With this in mind, we next hypothesized that potentially the effects of burden of hypotension on outcome could be modified by the severity of the initial arrest, or that is to say, potentially for patients with a more severe initial arrest, it may be more important to control blood pressure to affect a good outcome at the end of the day. To test this hypothesis, we dichotomized our patients based on CPR time, which is a known surrogate for the severity of initial arrest, and we chose a cutoff time of 10 minutes as this would divide our cohort evenly into two equal halves. So looking at the stratified data, this is the same graph as I've shown before, just with two separate cohorts, those with CPR time greater than 10 minutes and those with CPR time less than 10 minutes. I'll draw your attention to the right subgraph, which is the greater than 10 minutes duration of CPR, folks. For these patients, there's a significant difference in the calculated burden of hypotension at all percentiles tested from the 5th all the way up to the 50th percentile for age. Among patients with less than 10 minutes or equal to 10 minutes of CPR, there was no association seen at any percentile we looked at. There are many limitations to the study. First of all, this was a single-center study with a retrospective study design. Our institution also happens to have a very robust post-arrest blood pressure management pathway, and a likely sequela of this is that the overall prevalence of hypotension, less than the 5th percentile in our total data set, was quite low, less than 1% of all the time points we considered. So while it's gratifying to know that we're adhering very well to our pathway targets, it means that we don't have as much signal as we would have liked to to conduct this analysis. Next, the median time to recording onset was 90 minutes, and so any early-onset hypotension that occurred right after ROSC would not necessarily be captured by this study. Finally, a significant portion of our patients did have a poor outcome despite a low-calculated burden of hypotension. This isn't surprising to us, given the diversity of pathophysiologies of patients who arrest in RPICU. Some of these patients likely had poor outcome from arrest-independent factors, while others may have had confounding reasons to have a non-hypotensive blood pressure, such as being on steroids or having increased ICP. So to conclude using high-resolution arterial blood pressure data in our single-center cohort, we found that the burden of hypotension less than the 5th percentile for age in the first 24 hours after ROSC is associated with unfavorable neurologic outcome. And then separately for patients specifically with a CPR time greater than 10 minutes, that extends past the 5th percentile up to as high as the 50th percentile for age. I have many people to thank for this work. Couldn't list them all on one slide, but specifically I have to call out Matt, Alexis, and Monique who provided unconditional mentorship and put a lot of faith in a humble GenPeds trainee. Tommy, who helped really build this project up from the ground, especially from the software side and the residency and division for sponsoring this work. Take any questions. Thanks so much.

hypotension

neurologic outcomes

pediatric patients

cardiac arrest

arterial line data