WEBVTT NOTE This file was generated by Descript 00:00:00.000 --> 00:00:03.299 Judith: Welcome to Berry's In the Interim podcast, where we explore the 00:00:03.299 --> 00:00:07.880 cutting edge of innovative clinical trial design for the pharmaceutical and 00:00:07.880 --> 00:00:10.229 medical industries, and so much more. 00:00:10.960 --> 00:00:11.680 Let's dive in. 00:00:12.780 --> 00:00:13.530 Scott Berry: All right. 00:00:13.530 --> 00:00:14.700 Welcome everybody. 00:00:14.700 --> 00:00:20.460 Back to, in the interim where we talk about all things clinical trial science. 00:00:20.460 --> 00:00:26.610 And I have a, uh, distinguished clinical trial scientists, uh, with me today. 00:00:26.610 --> 00:00:28.200 So I have Dr. 00:00:28.200 --> 00:00:29.610 Derek Angus. 00:00:29.705 --> 00:00:35.255 Who is a distinguished professor and holds the Mitchell p Fink endowed 00:00:35.255 --> 00:00:39.815 Chair in Critical Care Medicine at the University of Pittsburgh. 00:00:40.295 --> 00:00:44.105 Uh, also secondary appointments in medicine, health policy and 00:00:44.105 --> 00:00:48.935 management, clinical Trans Clinical and translational sciences. 00:00:49.415 --> 00:00:54.230 He's been the chair of the Department of Critical Care Medicine since 2008. 00:00:55.965 --> 00:01:00.405 Is, and since 2015, he is been the director of the, uh, 00:01:00.645 --> 00:01:03.285 U-P-M-C-I-C-U Services Center. 00:01:03.765 --> 00:01:06.285 He's done everything in critical care. 00:01:06.495 --> 00:01:11.385 Uh, he's in Pittsburgh, uh, uh, in the Department of Critical Care. 00:01:11.385 --> 00:01:13.815 He's also a senior editor at jama. 00:01:13.815 --> 00:01:17.085 So Derek, welcome to in the interim. 00:01:18.085 --> 00:01:19.375 Derek Angus: Thank you, Scott. 00:01:19.375 --> 00:01:21.205 It's uh, it's my pleasure. 00:01:22.650 --> 00:01:25.440 Scott Berry: Yeah, so this is a, a, a space. 00:01:25.440 --> 00:01:30.030 We want to talk about all things clinical trials, and we, I think we have 00:01:30.030 --> 00:01:32.280 several topics we'd like to go to here. 00:01:32.690 --> 00:01:38.390 And the first topic, uh, I think is, is interesting in terms of clinical trial, 00:01:38.565 --> 00:01:44.270 trial science, and something I don't know much about, but steroids for the 00:01:44.270 --> 00:01:47.300 treatment of community acquired pneumonia. 00:01:49.065 --> 00:01:52.095 And maybe we should set this up a little bit. 00:01:52.155 --> 00:01:57.555 Um, my experience and, and Derek and I worked together on the REMAP CAP trial. 00:01:57.555 --> 00:01:59.535 We've worked together on multiple things. 00:01:59.865 --> 00:02:02.745 Uh, I, I consider Derek a very good friend. 00:02:03.075 --> 00:02:07.280 Uh, so I'll get that, that, uh, conflict of interest outta the way. 00:02:08.130 --> 00:02:10.770 But let's talk about steroids. 00:02:10.770 --> 00:02:12.780 So the history of steroids. 00:02:12.780 --> 00:02:17.910 It'd be better if you give this history of steroids from Cape Cod through Remap Cap. 00:02:17.910 --> 00:02:22.800 What do we know about using steroids in community acquired pneumonia, 00:02:22.800 --> 00:02:24.750 and why is it so controversial? 00:02:31.170 --> 00:02:35.310 Derek Angus: It kind of tires me just to think about the topic, but, um, 00:02:35.460 --> 00:02:40.500 all right, let's, let's, so first of all, steroids are incredibly 00:02:40.500 --> 00:02:42.690 powerful pleiotropic agents. 00:02:42.690 --> 00:02:46.830 They do all sorts of things, uh, that, broadly speaking, in this space of 00:02:46.830 --> 00:02:52.290 critical illness can include dampening down the immune system and can 00:02:52.290 --> 00:02:55.110 include some potentially beneficial. 00:02:55.695 --> 00:03:00.795 Uh, effects on your cardiovascular system by helping, uh, uh, retain 00:03:00.795 --> 00:03:03.075 fluids help improve resolution. 00:03:03.525 --> 00:03:06.315 But they're pleiotropic in that they do all sorts of things that 00:03:06.315 --> 00:03:08.145 are disadvantageous as well. 00:03:08.565 --> 00:03:12.675 Uh, every medical student learns about all the pros and cons of prolonged 00:03:12.675 --> 00:03:14.235 steroid therapy and so forth. 00:03:14.235 --> 00:03:18.375 So they are powerful agents that do a lot of things and they're dirt cheap. 00:03:19.485 --> 00:03:24.555 So if they were to work, uh, and you knew how to give them, uh, they 00:03:24.555 --> 00:03:29.355 could be incredibly advantageous worldwide in all sorts of settings. 00:03:30.105 --> 00:03:33.525 Um, in the space of community acquired pneumonia. 00:03:33.915 --> 00:03:37.395 So first of all, we're not necessarily talking about the pneumonia where 00:03:37.395 --> 00:03:43.635 you never come to hospital, but of all the pneumonia that gets you sick 00:03:44.025 --> 00:03:45.400 enough that you need to go to hospital. 00:03:46.080 --> 00:03:49.200 That also overlaps with getting sepsis. 00:03:49.260 --> 00:03:53.790 If you come in with a pneumonia and you get organ dysfunction, you've 00:03:53.790 --> 00:03:56.700 effectively met the criteria for sepsis. 00:03:57.210 --> 00:04:00.450 And so when you talk about steroids for pneumonia, you're also 00:04:00.450 --> 00:04:02.520 talking about steroids for sepsis. 00:04:02.700 --> 00:04:06.810 In fact, pneumonia is the most common cause of sepsis. 00:04:07.140 --> 00:04:12.870 So there's a big literature on trying steroids in both sepsis and in pneumonia. 00:04:13.395 --> 00:04:16.725 And when you do a sepsis trial, uh, of steroids, half 00:04:16.725 --> 00:04:18.765 of the patients have pneumonia. 00:04:19.305 --> 00:04:23.205 And when you do a pneumonia trial, half of the pneumonia patients have sepsis. 00:04:23.535 --> 00:04:29.685 So they're sort of, they're joined at the hip, um, back in the eighties. 00:04:30.570 --> 00:04:35.970 Uh, people were giving huge boluses of methyl prednisone, like a gram of 00:04:35.970 --> 00:04:40.890 prednisone, and people thought that would instantaneously bring people 00:04:40.890 --> 00:04:47.730 back from the brink of otherwise dying within minutes of profound septic shock. 00:04:48.090 --> 00:04:49.540 And then a couple of big trials said. 00:04:50.745 --> 00:04:52.725 Oh no, this is actually killing people. 00:04:52.725 --> 00:04:56.625 You, you might think that you get some temporary stabilizing effects 00:04:56.625 --> 00:05:00.825 on the cardiovascular instability, but actually that's such a huge dose. 00:05:00.825 --> 00:05:03.945 It probably causes secondary infections, et cetera. 00:05:04.365 --> 00:05:08.505 And so then steroids disappeared and for a long time people thought, 00:05:08.505 --> 00:05:12.000 yes, steroids always help in the short term, but as best as we can 00:05:12.000 --> 00:05:13.515 tell, they're not actually helpful. 00:05:14.325 --> 00:05:18.435 And then around the turn of the century, the French, uh, really led by 00:05:18.435 --> 00:05:25.095 Jali and Ann started to suggest, um, we weren't using steroids properly. 00:05:25.875 --> 00:05:28.425 Um, and he in particular started thinking 00:05:28.440 --> 00:05:34.519 Scott Berry: And, and properly, is it mean, the right patient or the right kind? 00:05:35.479 --> 00:05:36.619 Derek Angus: Yes, exactly. 00:05:37.189 --> 00:05:39.859 So he came up with you. 00:05:40.069 --> 00:05:44.174 He came up with, you need to use a different formulation in 00:05:44.269 --> 00:05:45.889 a different subset of patients. 00:05:46.999 --> 00:05:51.799 Um, so, uh, he started experimenting with. 00:05:52.504 --> 00:05:57.514 Effectively giving a week of hydrocortisone, not a day of methyl 00:05:57.514 --> 00:06:03.394 prednisone at a much lower dose of sort of total steroid dose. 00:06:03.544 --> 00:06:07.204 He also combined it with food cortisone because he argued that that would 00:06:07.204 --> 00:06:13.894 also be beneficial and he restricted it to these really sick septic shock 00:06:14.189 --> 00:06:16.564 patients that he felt had not responded. 00:06:17.089 --> 00:06:22.399 To fluids and, and, and, and who were so likely to die of the sepsis 00:06:23.029 --> 00:06:24.829 that you should give them steroids. 00:06:24.829 --> 00:06:29.119 That if you, if you used a milder or broader cohort, you might accidentally 00:06:29.119 --> 00:06:34.699 give patients who weren't going to die anyway, but now you might increase the 00:06:34.699 --> 00:06:36.499 likelihood of harm with the steroids. 00:06:37.189 --> 00:06:42.919 And he, and he particularly also said there's even a subset in whom. 00:06:43.339 --> 00:06:49.339 They're ex, we all make steroids and it, it's a stress response hormone. 00:06:49.339 --> 00:06:55.279 And so he said you should even be doing a test to try to understand whether your own 00:06:55.279 --> 00:06:58.699 endogenous stress response is compromised. 00:06:59.119 --> 00:07:05.059 Uh, and if you can't mount a good enough internal endogenous steroid response, you 00:07:05.059 --> 00:07:07.369 would be particularly likely to respond. 00:07:08.089 --> 00:07:11.749 Bottom line is he had a spectacularly successful study. 00:07:12.229 --> 00:07:16.549 Suddenly the international guidelines were saying, oh, you should give steroids. 00:07:16.909 --> 00:07:18.619 They clearly work in sepsis. 00:07:18.619 --> 00:07:21.529 And then people started playing around with broader patient 00:07:21.529 --> 00:07:23.449 populations, again, like pneumonia. 00:07:24.109 --> 00:07:27.649 Uh, they said, well, maybe the pneumonia that's particularly 00:07:27.679 --> 00:07:30.859 pro-inflammatory might benefit the most. 00:07:31.189 --> 00:07:34.879 Once you know you have antibiotics on, you should maybe also give steroids. 00:07:34.879 --> 00:07:36.559 And there were a number of. 00:07:36.994 --> 00:07:40.594 Studies that were sort of grumbling along, somewhat beneficial or not. 00:07:41.254 --> 00:07:48.544 But meanwhile, in the broader area of sepsis, no one could repeat Ali's study. 00:07:49.144 --> 00:07:52.714 The Europeans tried it in something called Corticus. 00:07:53.134 --> 00:07:57.904 Um, uh, the, the international guidelines, which were published 00:07:57.904 --> 00:08:00.064 every four years, all through the. 00:08:00.469 --> 00:08:01.699 21st century. 00:08:02.179 --> 00:08:08.059 If you pull up the steroid section, the, the literature was changing slowly, but 00:08:08.059 --> 00:08:12.289 every four years the guidelines would be written differently depending on who 00:08:12.289 --> 00:08:17.959 was in the room because this was, uh, what I would call an opinion rich data, 00:08:17.959 --> 00:08:23.899 poor environment, and, and, and people just felt, they continued to feel like. 00:08:24.589 --> 00:08:26.749 Uh, everyone had their favorite study. 00:08:26.749 --> 00:08:29.449 If you wanted to give steroids, you liked the Annane study. 00:08:29.449 --> 00:08:31.879 If you didn't wanna give steroids, you liked the Corticus study. 00:08:32.839 --> 00:08:35.389 So then people said, enough is enough. 00:08:35.389 --> 00:08:38.329 We need some bigger, more definitive trials. 00:08:38.329 --> 00:08:42.859 And two groups decided to do much larger trials in septic shock, and they're much 00:08:42.859 --> 00:08:45.139 bigger than any of the pneumonia trials. 00:08:45.139 --> 00:08:48.649 So, I know you asked about pneumonia, but I feel like in many 00:08:48.649 --> 00:08:50.539 respects, the sepsis literature. 00:08:51.544 --> 00:08:52.804 Is the more dominant. 00:08:53.254 --> 00:08:59.884 So, uh, the French did another much bigger study called APROCCHSS um, and it 00:08:59.884 --> 00:09:03.214 was published in the New England Journal of Medicine, and yet again, in their 00:09:03.214 --> 00:09:08.284 hands, they knocked it out of the park with a big improvement in mortality. 00:09:08.644 --> 00:09:12.214 Meanwhile, the Australians ran the trial called Adrenal with 00:09:12.214 --> 00:09:15.634 several thousand patients also published in the New England. 00:09:15.874 --> 00:09:18.754 No Benefit from steroids. 00:09:19.174 --> 00:09:20.014 It's like, are you 00:09:20.019 --> 00:09:21.514 Scott Berry: A, any, any harm? 00:09:21.814 --> 00:09:24.334 Any harm potential from adrenal? 00:09:24.904 --> 00:09:30.304 Derek Angus: So Adrenal didn't really show any harm overall. 00:09:30.364 --> 00:09:35.734 Um, but one obviously wonders if the French were successful and if 00:09:35.734 --> 00:09:39.454 the int criteria weren't exactly the same and they were recruiting 00:09:39.634 --> 00:09:42.904 a, a narrower subset was adrenal. 00:09:44.299 --> 00:09:48.529 Neutral because it was including the people that just looked like the 00:09:48.529 --> 00:09:53.899 French in the APROCCHSS trial, but also including other patients that 00:09:53.899 --> 00:09:57.499 in whom there was no benefit and one canceled the other, if that makes sense. 00:09:58.159 --> 00:10:01.939 Now, some of this is published, some of this is not published, but people have 00:10:01.939 --> 00:10:07.939 then tried to find within adrenal the subset of patients that look like the 00:10:07.939 --> 00:10:10.609 patients that were in APROCCHSS to see. 00:10:11.089 --> 00:10:13.519 Well, is there at least some suggestion? 00:10:13.849 --> 00:10:20.419 Uh, but thus far, no one can really, no one can find within the adrenal trial 00:10:20.809 --> 00:10:24.979 a subset that looked like APROCCHSS who then have a similar effect to APROCCHSS 00:10:25.399 --> 00:10:28.879 So people are confused along the way. 00:10:29.884 --> 00:10:33.574 Uh, the French also did another trial called Cape Cod that looked 00:10:33.574 --> 00:10:35.614 quite like the APROCCHSS trial only. 00:10:35.614 --> 00:10:39.994 It was just in pneumonia, slightly lower, uh, severity of illness group. 00:10:40.024 --> 00:10:44.314 There again, they showed fantastic benefit. 00:10:44.914 --> 00:10:49.804 So every trial that's run with flu drug, cortisone based 00:10:49.804 --> 00:10:52.144 outta France has been positive. 00:10:52.414 --> 00:10:54.244 There's a lot of ity there. 00:10:54.664 --> 00:10:55.294 I don't know. 00:10:55.774 --> 00:10:57.304 I don't know whether it's about. 00:10:57.694 --> 00:11:02.374 Being French or if it's about giving food, cortisone, no one has tested 00:11:02.374 --> 00:11:04.204 food cortisone outside France. 00:11:04.624 --> 00:11:07.324 Uh, and every time they test it, it's been beneficial. 00:11:07.924 --> 00:11:15.034 Um, and then the French then also went back to approach, um, and looked at 00:11:15.124 --> 00:11:19.654 the patients in approach the subset of sepsis patients who had pneumonia. 00:11:20.134 --> 00:11:24.064 And there too, they found most of the signal in APROCCHSS appeared 00:11:24.064 --> 00:11:25.319 to be in the pneumonia patients. 00:11:27.124 --> 00:11:33.454 So, uh, Cape Cod was positive APROCCHSS was a positive and APROCCHSS was 00:11:33.454 --> 00:11:37.624 most positive within the patients that had sepsis and pneumonia. 00:11:37.984 --> 00:11:39.334 So this definitely. 00:11:40.549 --> 00:11:44.959 I mean, if you are in Fran, if there was only the French literature, this 00:11:44.959 --> 00:11:50.509 would all be tied up with a bow on it, which is you would give steroids 00:11:50.599 --> 00:11:54.979 broadly to hospitalized pneumonia and broadly to sepsis, and you would expect 00:11:54.979 --> 00:11:59.179 the largest benefit in the patients who had sepsis due to pneumonia. 00:11:59.974 --> 00:12:03.784 And we would sort of wipe our hands and go on to one of the many other 00:12:03.784 --> 00:12:05.854 unresolved questions in critical care. 00:12:06.454 --> 00:12:12.154 Uh, but the problem is we cannot generate an evidence base outside 00:12:12.154 --> 00:12:14.914 France that goes in the same direction. 00:12:15.304 --> 00:12:21.064 There are some other pneumonia trials with non mortal endpoints that 00:12:21.064 --> 00:12:24.334 are somewhat beneficial, but, uh. 00:12:24.979 --> 00:12:29.389 Most of the positive signal is driven by these French trials, and so even 00:12:29.389 --> 00:12:33.139 when you look at meta-analysis, I would say most of that is being 00:12:33.139 --> 00:12:34.879 driven by the French experience. 00:12:36.484 --> 00:12:42.994 The French are now leading another huge consortium, uh, to study sep 00:12:43.054 --> 00:12:45.454 uh, steroids, particularly in sepsis. 00:12:45.454 --> 00:12:49.834 But again, many patients who have sepsis due to the mor, I forget 00:12:49.834 --> 00:12:52.864 the acronym for it, but it's just been funded by the European Union 00:12:52.864 --> 00:12:55.114 and it's been led by Jali Nan. 00:12:56.314 --> 00:12:59.044 And one of the things they're going to be trying to tease out 00:12:59.224 --> 00:13:04.654 is, uh, well, let me step back. 00:13:05.584 --> 00:13:05.974 So. 00:13:06.919 --> 00:13:11.149 I've somewhat jokingly said there's the French experience versus the 00:13:11.149 --> 00:13:16.039 rest of the world experience, but, but more pertinently. 00:13:16.039 --> 00:13:23.569 I think what we can say is we don't get consistent results and across the 00:13:23.569 --> 00:13:27.859 different trials, we give the drug slightly differently and we select 00:13:27.859 --> 00:13:29.179 the patient slightly differently. 00:13:29.509 --> 00:13:32.539 We could also be giving co interventions differently as well. 00:13:32.959 --> 00:13:35.389 We also have a slightly different duration of follow up. 00:13:35.389 --> 00:13:37.489 You asked earlier on about harm. 00:13:37.939 --> 00:13:42.709 Steroids always run the possibility of being beneficial on the short end, 00:13:43.069 --> 00:13:48.859 but then giving steroids has some unwanted sequelae that manifests later. 00:13:48.859 --> 00:13:52.179 So it could also be the timing and the nature of the primary end part. 00:13:53.194 --> 00:13:57.934 So when the evidence base is consistent, consists of variable, variable 00:13:57.934 --> 00:14:02.884 drug and dose, variable patient selection and variable endpoint, 00:14:03.334 --> 00:14:08.164 and then you don't get a consistent result, you're left saying, okay, 00:14:08.524 --> 00:14:11.524 uh, what's driving the inconsistency? 00:14:12.754 --> 00:14:16.084 Now, obviously the endpoint is relatively easy to fix. 00:14:16.084 --> 00:14:19.114 In all of the different trials, you could collect multiple different 00:14:19.114 --> 00:14:23.644 endpoints and then try to measure a common endpoint across all the trials. 00:14:23.644 --> 00:14:27.454 So regardless of what was primary, if you could, if you had long 00:14:27.454 --> 00:14:31.114 enough and rich enough follow up, you could, you could standardize. 00:14:32.194 --> 00:14:39.664 Um, so then let's go back to the two other plausible domains of heterogeneity. 00:14:40.084 --> 00:14:44.044 Um, there, there, there are, I should say, before I 00:14:44.044 --> 00:14:45.094 Scott Berry: Well, we, we should wait. 00:14:45.244 --> 00:14:47.674 Let's touch on, let, let's touch on a little bit. 00:14:47.854 --> 00:14:51.844 But then there was C and Cape Cod, for example. 00:14:51.844 --> 00:14:57.574 I don't think they enrolled influenza, so they stayed away from, uh, influenza. 00:14:57.994 --> 00:15:05.584 But then during COVID recovery, REMAP CAP showed strong benefit of steroids. 00:15:06.154 --> 00:15:08.254 For the treatment of hospitalized COVID. 00:15:08.464 --> 00:15:09.694 Derek Angus: Yes, yes, yes. 00:15:09.769 --> 00:15:09.989 So. 00:15:10.909 --> 00:15:12.439 Uh, Hey, it's Friday. 00:15:12.469 --> 00:15:13.429 It's Friday afternoon. 00:15:13.429 --> 00:15:14.209 What can I say? 00:15:14.209 --> 00:15:15.919 I'm not firing on all cylinders. 00:15:16.159 --> 00:15:20.134 I should totally have brought that up that prior to COVID, I. 00:15:20.974 --> 00:15:24.694 We were a bit on again, off again with steroids and we were moving 00:15:24.694 --> 00:15:28.594 increasingly towards, well maybe in the sickest septic shock patients, in 00:15:28.594 --> 00:15:30.094 part because of the approach trial. 00:15:30.394 --> 00:15:34.414 But we'd got away from giving it more broadly in pneumonia 'cause it 00:15:34.414 --> 00:15:35.794 just wasn't a strong enough signal. 00:15:35.794 --> 00:15:36.124 Cape. 00:15:36.634 --> 00:15:40.144 The, the non COVID Cape Cod hadn't yet been published in so on. 00:15:40.684 --> 00:15:46.144 However, then during COVID, which is bad pneumonia, suddenly it looked 00:15:46.144 --> 00:15:47.824 like steroids were beneficial. 00:15:48.184 --> 00:15:52.954 And in the Dexamethasone experience and recovery, the, the benefit 00:15:52.954 --> 00:15:54.694 was all in the sicker patients. 00:15:55.174 --> 00:16:00.304 Uh, which again, that just almost reinforces that it's not 00:16:00.304 --> 00:16:02.104 just in bacterial pneumonia. 00:16:02.134 --> 00:16:06.394 It may be in general that, that they, this sort of. 00:16:06.889 --> 00:16:12.169 The severe end of this syndromic experience of getting threatened lung 00:16:12.169 --> 00:16:19.969 function from some invading body that then could be complicated by multiple organ 00:16:19.969 --> 00:16:25.669 failure sepsis, regardless of whether that's driven by virus or by bacteria. 00:16:25.669 --> 00:16:25.849 It seems. 00:16:27.349 --> 00:16:29.179 Steroids are really doing something. 00:16:29.179 --> 00:16:34.279 People also thought one of the advantages of COVID is that it's, it's a much more 00:16:34.279 --> 00:16:39.919 homogenous insult that in some of the other trials you can't, sometimes you 00:16:39.919 --> 00:16:41.989 can't even work out, is it pneumonia? 00:16:42.019 --> 00:16:45.079 Like a ton of pneumonia, classic pneumonia trials. 00:16:45.079 --> 00:16:50.029 Many patients have no identified organism, and so one of the longstanding things 00:16:50.029 --> 00:16:54.109 was, well, you didn't even put the right patients in the trial, whereas. 00:16:54.499 --> 00:17:00.199 The COVID trials people said, well now we, these patients are in hospital 00:17:00.379 --> 00:17:02.929 for a pathogen driven pneumonia. 00:17:03.469 --> 00:17:06.259 They're not just in hospital for a pneumonia complex. 00:17:06.259 --> 00:17:09.859 That might sometimes actually just be heart failure that looks on 00:17:09.859 --> 00:17:12.529 the chest X-ray, like pneumonia. 00:17:12.529 --> 00:17:16.609 And so that, and so then people sort of said, well, you definitely 00:17:16.609 --> 00:17:21.079 want to give it in COID and what's more, when we have a true. 00:17:22.199 --> 00:17:26.494 A classically selected population in whom we have very high confidence that 00:17:26.494 --> 00:17:31.084 they definitely have an infectious pneumonia and especially when they're 00:17:31.084 --> 00:17:35.614 severely ill and getting altered, you know, distal organ dysfunction, 00:17:35.614 --> 00:17:37.984 IE sepsis due to the pneumonia. 00:17:38.464 --> 00:17:44.224 We seem to get this pretty strong homogenous sense of benefit from steroids. 00:17:44.644 --> 00:17:45.934 That's one argument. 00:17:45.964 --> 00:17:47.039 Uh, there are definitely. 00:17:48.364 --> 00:17:49.684 Critics of these trials. 00:17:49.684 --> 00:17:53.734 But for those who like them, they would say, not only should you give steroids in 00:17:53.734 --> 00:17:59.044 COVID, but while we're at it, it's making us increasingly think maybe steroids. 00:17:59.434 --> 00:18:01.414 You know, if you like peanuts, you'll love peanut butter. 00:18:01.474 --> 00:18:06.694 And so now the guidelines are, they've totally swung to be very positive. 00:18:07.549 --> 00:18:08.839 Current guidelines. 00:18:09.789 --> 00:18:10.199 They all. 00:18:10.879 --> 00:18:15.709 Differ slightly in who they say you should give it to, and even which kind of drug 00:18:15.709 --> 00:18:18.769 you should give, whether you should give hydrocortisone or prednisone, et cetera. 00:18:19.189 --> 00:18:23.059 But I would say in general, guidelines from the different infectious disease 00:18:23.059 --> 00:18:27.769 and cial care societies, the WHO, et cetera, have all started to become much 00:18:27.769 --> 00:18:32.389 more, they're, they're all down on, they're all pretty positive on steroids. 00:18:36.319 --> 00:18:41.719 Scott Berry: so we then we were involved in remap cap in non COVI, 00:18:42.469 --> 00:18:49.459 where we hydrocortisone and we were looking at, uh, stratifications of 00:18:49.459 --> 00:18:52.309 patients by shock, by influenza. 00:18:52.459 --> 00:18:52.879 Yes. 00:18:52.879 --> 00:18:53.389 No. 00:18:54.019 --> 00:18:56.299 And that result showed. 00:18:56.824 --> 00:18:59.074 Reasonable probability of harm. 00:18:59.224 --> 00:19:02.704 It showed worse mortality in all four of those groups. 00:19:02.704 --> 00:19:06.154 We didn't have a whole lot of influenza, but, uh, it, it was 00:19:06.154 --> 00:19:13.474 pretty convincingly negative, 90% probability of harm in, in cap, uh, 00:19:13.504 --> 00:19:20.284 shock or not, uh, patience, again, a differential result to all of that. 00:19:23.514 --> 00:19:27.204 Derek Angus: Yeah, I hated that result. 00:19:29.454 --> 00:19:32.514 Um, that was incredibly disappointing. 00:19:32.814 --> 00:19:34.674 Uh, it, 00:19:37.259 --> 00:19:37.479 so 00:19:41.119 --> 00:19:42.714 I, I think we would agree that. 00:19:43.519 --> 00:19:47.419 That trial didn't run quite as well as we would've hoped. 00:19:47.809 --> 00:19:55.579 Um, it was an open label trial and we had worked with site who said that they 00:19:55.579 --> 00:19:59.299 have enough equip poison that if the patient is randomized to control, they 00:19:59.299 --> 00:20:01.219 would be happy not to give steroids. 00:20:01.789 --> 00:20:06.079 But there was bleeding of steroid use into the control arm. 00:20:06.709 --> 00:20:10.219 It wasn't massive, but you never know whether. 00:20:10.819 --> 00:20:13.489 You know, if it, if it's, if it's random bleeding, it 00:20:13.489 --> 00:20:15.109 probably doesn't matter too much. 00:20:15.109 --> 00:20:17.719 It's just slightly weakening any efficacy signal. 00:20:18.079 --> 00:20:22.789 But if the physicians have some magic inside track on knowing just exactly who 00:20:22.789 --> 00:20:27.469 they should treat, then you could imagine that any signal you would've found is 00:20:27.469 --> 00:20:30.644 getting wiped out by, uh, contamination. 00:20:30.864 --> 00:20:34.729 Having said that, I don't think contaminate, so I 00:20:34.729 --> 00:20:36.469 wish there hadn't been that. 00:20:37.309 --> 00:20:40.729 Steroid use in the control arm, but the steroid use in the 00:20:40.729 --> 00:20:44.929 control arm wouldn't explain the, almost hitting the harm signal. 00:20:45.979 --> 00:20:51.979 Like, uh, steroid use in the control arm would, if, if steroids would've 00:20:51.979 --> 00:20:56.419 normally worked, all it would've done is reduce the size of the benefit. 00:20:56.449 --> 00:20:59.359 It wouldn't have flipped it in the wrong direction. 00:20:59.359 --> 00:21:03.229 Now obviously you, as you start to get smaller sample size, 00:21:03.229 --> 00:21:04.279 some of this is just sort of. 00:21:04.654 --> 00:21:11.254 Random error and so forth, but that, you know, it made me slightly uncomfortable. 00:21:11.254 --> 00:21:14.289 We also, as you know, we had a. 00:21:15.244 --> 00:21:20.524 Uh, there was a transposition error by one of our data vendors early 00:21:20.524 --> 00:21:24.484 on in the trial that sent us a data set that had been mislabeled. 00:21:24.964 --> 00:21:29.164 We, we, as soon as we found that out, we sort of fixed it all and they're no 00:21:29.164 --> 00:21:31.744 longer a data vendor even in the trial. 00:21:32.254 --> 00:21:37.294 But, um, that did mean that there was a period when we were probably randomizing. 00:21:38.254 --> 00:21:42.664 Uh, with a more distorted sort of response, adaptive the response. 00:21:42.664 --> 00:21:47.314 Adaptive randomization triggered a more aggressive randomization 00:21:47.314 --> 00:21:49.444 away from what would've been ideal. 00:21:49.774 --> 00:21:54.244 And again, the main consequence there was probably, you know, we had several 00:21:54.244 --> 00:21:57.124 hundred patients in this analysis, but they weren't evenly distributed. 00:21:58.179 --> 00:22:02.284 We would've had more power if it was more of a one-to-one distribution. 00:22:02.944 --> 00:22:06.724 Um, and so that again, just introduces another level of uncertainty. 00:22:07.174 --> 00:22:13.234 Uh, but, but what I don't want to lose in any of that is while none 00:22:13.234 --> 00:22:19.384 of these things were ideal, we were a long way from looking anything 00:22:19.384 --> 00:22:21.634 like the Cape Cod experience. 00:22:22.759 --> 00:22:23.089 Scott Berry: Yeah. 00:22:23.479 --> 00:22:27.754 Derek Angus: In, in other words, no matter how much you subset the trial, 00:22:27.754 --> 00:22:31.294 no matter how much you tried to, is, for example, in the early part of the 00:22:31.294 --> 00:22:36.244 trial before there was any response, adaptive randomization whatsoever, when 00:22:36.244 --> 00:22:40.504 this was just like a straight head to head, traditional one-to-one trial. 00:22:40.744 --> 00:22:47.974 Even there, the trend that, uh, the patient's getting steroids, their outcome, 00:22:48.559 --> 00:22:51.094 the, the observed outcome rate was worse. 00:22:51.544 --> 00:22:51.664 Um. 00:22:52.474 --> 00:22:55.234 So is it all just the play of chance? 00:22:56.344 --> 00:23:06.274 Uh, possible, but, but to your point, um, uh, uh, the, the probability 00:23:06.274 --> 00:23:12.424 statement nonetheless would say that even, even accounting for chance, it's, 00:23:12.424 --> 00:23:18.574 it's just highly unlikely that, um, that steroids were strongly beneficial. 00:23:19.144 --> 00:23:20.344 Um, and. 00:23:20.734 --> 00:23:25.744 I wanna remind people that the adrenal trial with 4,000 00:23:25.744 --> 00:23:28.474 patients found no benefit. 00:23:28.474 --> 00:23:32.644 So it's not as if remap cap was an outlier. 00:23:32.914 --> 00:23:39.964 Um, remap cap in many ways was one more trial consistent with the many trials 00:23:40.114 --> 00:23:42.394 that have not actually found a benefit. 00:23:43.054 --> 00:23:46.024 Uh, and earlier on you had asked about harm. 00:23:46.564 --> 00:23:49.894 Um, there was. 00:23:49.954 --> 00:23:54.034 It's like the old one, the big European Cardus trial. 00:23:54.454 --> 00:23:58.774 Um, they had thought that patients appeared to get slightly 00:23:58.774 --> 00:24:00.934 faster resolution of shock. 00:24:01.654 --> 00:24:06.544 And by the way, in remap cap, it looked like the observed rate of cardiovascular 00:24:06.544 --> 00:24:08.974 dysfunction appeared to resolve sooner. 00:24:08.974 --> 00:24:12.394 And that was also what the adrenal trial found. 00:24:12.454 --> 00:24:15.154 That, that the early short term cardiovascular 00:24:15.154 --> 00:24:16.924 instability resolved sooner. 00:24:17.959 --> 00:24:21.709 But in Cardus, where they had done follow up on secondary infections, 00:24:21.709 --> 00:24:27.619 secondary infections, which, you know, steroids can cause, uh, were higher. 00:24:27.679 --> 00:24:32.119 There was a higher nosocomial infection rate, uh, in the steroid arm. 00:24:32.629 --> 00:24:35.719 And I, it is possible that. 00:24:36.559 --> 00:24:40.729 Uh, something was going on in Remap Cap where we might have actually been getting 00:24:40.729 --> 00:24:45.679 some of the early cardiovascular benefits, but the net effect for the, for the cohort 00:24:45.679 --> 00:24:48.349 in general was that steroids were worse. 00:24:48.379 --> 00:24:50.059 And so now we get back to, 00:24:51.064 --> 00:24:51.354 Scott Berry: Yeah. 00:24:52.939 --> 00:24:54.919 Derek Angus: uh, can you better 00:24:55.004 --> 00:24:55.294 Scott Berry: Okay. 00:24:55.309 --> 00:24:57.529 So let's let Larry, yeah. 00:24:58.534 --> 00:24:58.954 Right. 00:24:59.134 --> 00:25:03.334 So one of the things that was sort of striking in the remap cap trial is 00:25:03.334 --> 00:25:09.874 that mortality was against steroids, but organ support free days, which 00:25:09.874 --> 00:25:13.594 includes mortality, but it includes time to resolution of organ support, 00:25:13.804 --> 00:25:15.934 actually was better on the steroid arm. 00:25:16.194 --> 00:25:21.564 Despite the fact that mortality was worse, it, it, it has the earmark of 00:25:21.594 --> 00:25:25.794 another particular disease that was somewhat striking is the treatment of 00:25:25.794 --> 00:25:30.054 acute stroke, where early on in the treatment of acute stroke and the use 00:25:30.054 --> 00:25:36.324 of endovascular therapy, they saw this non-proportional effect that they, they, 00:25:36.324 --> 00:25:39.564 they certainly improved the number of. 00:25:39.729 --> 00:25:41.289 Really good outcomes. 00:25:41.439 --> 00:25:44.439 They certainly increase that, but they increased the number of deaths 00:25:44.529 --> 00:25:48.579 and it was a sign that there's differential heterogeneous treatment. 00:25:48.579 --> 00:25:52.149 In fact, they were probably treating people they should not treat, 00:25:52.149 --> 00:25:55.839 and they were causing harm on those, and they were treating some 00:25:55.839 --> 00:25:57.399 that they absolutely benefited. 00:25:57.639 --> 00:26:00.909 And it took a while before they kind of figured the right patients. 00:26:00.909 --> 00:26:03.129 And then they saw huge benefits in that. 00:26:03.489 --> 00:26:07.209 So, uh, the, the steroids has this hallmark of. 00:26:08.059 --> 00:26:10.939 Heterogeneous effect across the population. 00:26:10.999 --> 00:26:17.029 And I know you, you touched on, maybe it's the infection, maybe it's the host, 00:26:17.299 --> 00:26:21.649 maybe it's the timing of the disease, maybe it's the steroid themselves, 00:26:21.889 --> 00:26:27.169 dexamethasone, prednisone, that, the type of that there's so much here. 00:26:27.529 --> 00:26:30.949 So it's got the earmarks of heterogeneity. 00:26:31.654 --> 00:26:35.824 You know, should we be running trials that we enroll a thousand patients and 00:26:35.824 --> 00:26:38.554 do a single analysis of is it beneficial? 00:26:39.154 --> 00:26:40.714 Do we need better trials? 00:26:40.714 --> 00:26:47.734 So let's sort of turn it to where do we go and if you could design the right trial, 00:26:47.914 --> 00:26:53.254 you seem, you seem to not know which patients, which treatments, which time. 00:26:53.404 --> 00:26:57.664 Shock based therapy, uh, a seven day dosing. 00:26:58.119 --> 00:27:02.259 You know, where would we go if we could fund the right trial? 00:27:02.259 --> 00:27:03.909 What's the right trial design? 00:27:04.199 --> 00:27:04.399 Derek Angus: Yeah. 00:27:04.999 --> 00:27:10.129 So right before, so I would love to talk about that, but can I just ask, 00:27:10.659 --> 00:27:11.109 Scott Berry: Okay. 00:27:11.299 --> 00:27:13.489 Derek Angus: just make a clarifying, so. 00:27:14.719 --> 00:27:19.609 Uh, when, when you have the organ failure free days, if you died, any effect on 00:27:19.609 --> 00:27:23.359 your organ dysfunction is sort of subsumed within the fact fact that you died. 00:27:23.719 --> 00:27:28.009 But you did allude to this all being explainable by the right patient. 00:27:28.009 --> 00:27:31.399 But it is possible with some of these early versus late outcomes. 00:27:31.939 --> 00:27:35.569 That's just the timing of the measurement, even within an individual patient. 00:27:36.184 --> 00:27:40.354 That that wouldn't be the case when it's organ failure free days 00:27:40.384 --> 00:27:41.824 because if you died you're dead. 00:27:42.184 --> 00:27:52.174 But, um, uh, we know in critical care, for example, um, low tidal volumes, which, um, 00:27:52.264 --> 00:27:57.784 is a strategy of ventilating the patient in a way very differently from the way 00:27:57.784 --> 00:28:03.394 that we did for decades, um, would almost certainly make the patient look bluer. 00:28:03.919 --> 00:28:08.899 More hypoxic in the short run because greater expansion of the lungs 00:28:08.899 --> 00:28:10.729 actually facilitates gas exchange. 00:28:10.729 --> 00:28:16.429 You get better oxygenation, but it turned out that the low tidal volumes absolutely 00:28:16.429 --> 00:28:19.879 protected the lung and improved survival. 00:28:20.419 --> 00:28:24.649 And so high tidal volumes makes you look better in the long, better in 00:28:24.649 --> 00:28:27.229 the short run, but then kills you and. 00:28:28.294 --> 00:28:28.774 Scott Berry: Okay. 00:28:28.999 --> 00:28:32.509 Derek Angus: and so, and, and with all of these interventions when we 00:28:32.509 --> 00:28:37.699 move into these combined endpoints, do so at our peril because they 00:28:37.699 --> 00:28:42.739 do tend to rely on conceptual models that could be wrong about. 00:28:43.099 --> 00:28:46.519 Whether the intermediate outcome is on the path to greatness. 00:28:46.999 --> 00:28:51.949 You know, you, you would think a short term measure to improve organ dysfunction 00:28:51.979 --> 00:28:54.559 is surely helping you get out alive. 00:28:54.589 --> 00:28:57.709 But the answer is not so fast. 00:28:58.189 --> 00:28:59.149 So that's one issue. 00:28:59.149 --> 00:29:03.259 But, but, but, but assuming we've got the outcome right, and the problem 00:29:03.259 --> 00:29:07.129 is heterogeneous patients and you don't know which ones they are. 00:29:07.729 --> 00:29:10.549 Um, I mean, this is a classic issue. 00:29:10.549 --> 00:29:11.479 If, if you, if. 00:29:12.614 --> 00:29:19.609 Uh, uh, so I feel, I feel like, um, uh, there were always what I would call the, 00:29:19.669 --> 00:29:24.439 the, the hope and pray models, which is what everyone did before, which is you 00:29:24.439 --> 00:29:28.729 just enroll everyone and hope that you got the right group, or that if there's 00:29:28.729 --> 00:29:32.749 a bad group, it's not so bad, it doesn't drag it down and all the rest of it. 00:29:33.139 --> 00:29:38.179 And then we wanted to start having smarter trials. 00:29:39.004 --> 00:29:45.484 If you have the luxury of doing many sequential trials, you could always do 00:29:45.484 --> 00:29:49.624 one large trial, see a group in whom you thought there was a benefit, and then 00:29:49.624 --> 00:29:52.234 do a follow up trial in just that group. 00:29:52.624 --> 00:29:55.864 Even there, you don't necessarily prove that there's heterogeneity. 00:29:56.374 --> 00:30:02.554 You know, that was, I, I remember, I think it was the editorialist, very astutely 00:30:02.554 --> 00:30:07.489 pointed out that that huge Jupiter trial in the New England, um, when they were. 00:30:08.329 --> 00:30:11.659 Uh, treating acute myocardial infarction patients with some 00:30:11.659 --> 00:30:15.349 sort of anti-inflammatory strategy just in those with a high CRP. 00:30:15.829 --> 00:30:20.299 The authors then said, oh, look, we, we improved outcome showing the 00:30:20.299 --> 00:30:24.859 inflammation hypothesis and so forth in myocardial infarction and, and, and, 00:30:24.889 --> 00:30:27.499 and the edit in the editorial that said. 00:30:28.414 --> 00:30:32.434 For that to be true, you would've had to have enrolled both the inflamed 00:30:32.434 --> 00:30:36.334 and the non-inflamed and showed that the drug worked differentially. 00:30:36.334 --> 00:30:39.814 You only showed that in the inflamed that the drug worked, but you haven't 00:30:39.814 --> 00:30:43.444 ruled out that the drug could have been working in the other group. 00:30:44.194 --> 00:30:44.524 Uh, 00:30:47.229 --> 00:30:53.494 if, if, if we don't really have the luxury to do lots of sequential trials, 00:30:53.884 --> 00:30:56.134 but want to be learning within one trial. 00:30:57.304 --> 00:31:02.074 Then, uh, we, we can do a couple of things. 00:31:02.164 --> 00:31:05.764 As you know, in remap cap, we've tended to have these a priority 00:31:05.764 --> 00:31:08.824 subgroups, um, that, that you could. 00:31:10.094 --> 00:31:15.139 If, if you think you have a rough sense of where the domains of heterogeneity 00:31:15.139 --> 00:31:20.269 may lie, uh, across both the axis and where the cut point in the axis 00:31:20.269 --> 00:31:26.749 might be, then you could start a trial with some embedded rules to then 00:31:26.989 --> 00:31:32.629 effectively, um, uh, enrich over time. 00:31:33.184 --> 00:31:35.554 Uh, uh, you can also do it in the other direction. 00:31:35.554 --> 00:31:36.844 You could deen enrich you. 00:31:36.844 --> 00:31:41.374 You could start with the subgroup that you think it works best in, and 00:31:41.374 --> 00:31:45.814 as long as you're getting a benefit, then you could then widen your entry 00:31:45.814 --> 00:31:48.484 criteria for a broader set later. 00:31:48.904 --> 00:31:55.549 I, I think those kind of trials still rely on confident science or hubris 00:31:57.044 --> 00:31:57.334 Scott Berry: Yeah. 00:31:57.429 --> 00:31:58.564 Derek Angus: whether you have. 00:31:58.939 --> 00:32:02.929 About whether you actually understand the axes of heterogeneity. 00:32:03.499 --> 00:32:06.589 Um, and so, and, and so 00:32:06.814 --> 00:32:08.884 Scott Berry: Well, well, doesn't, doesn't this now. 00:32:08.884 --> 00:32:09.124 Yeah. 00:32:09.124 --> 00:32:13.984 Doesn't this now demand a, a, a, a different approach where the, the 00:32:13.984 --> 00:32:18.064 amount of people you're caring for in critical care that have sepsis cap, 00:32:18.364 --> 00:32:24.184 moderate disease, severe disease, that we enroll a very large, try, a very 00:32:24.184 --> 00:32:28.294 large population and try to learn Yes. 00:32:28.294 --> 00:32:29.049 No in whom? 00:32:30.079 --> 00:32:36.709 Derek Angus: Yes, so, so this is where you, so let's step back for a second. 00:32:36.949 --> 00:32:37.159 So. 00:32:38.179 --> 00:32:43.129 We're, first of all, you would say we are committed to cause and effect. 00:32:43.789 --> 00:32:46.309 We're not really interested in association. 00:32:46.549 --> 00:32:51.829 Uh, we want to know if something works, but what we've done is 00:32:51.829 --> 00:32:56.869 we've begin and RCT tells you if something works, but it doesn't. 00:32:57.169 --> 00:33:01.219 But it's like causality with a small C, causality with a 00:33:01.219 --> 00:33:04.279 big C is, why does it work? 00:33:05.239 --> 00:33:10.219 And in a way, heterogeneity of treatment effect is more 00:33:10.909 --> 00:33:13.519 juda perel esque, if you like. 00:33:13.789 --> 00:33:17.509 You know, it's more getting at not just explaining if the drug 00:33:17.509 --> 00:33:19.489 worked, but why the drug worked. 00:33:19.849 --> 00:33:25.274 In fact, I think it was one of Judah Pearl's, uh, postdocs, um, Barne who, 00:33:25.879 --> 00:33:31.789 Elias Barnum I think, who started using Judah Pearl's sort of, um. 00:33:32.119 --> 00:33:38.329 Causal inference language to try to understand generalizability in RCTs. 00:33:38.359 --> 00:33:45.199 If you understood why, uh, effectively, if you understood all of the moderating 00:33:45.199 --> 00:33:49.879 and median effects between the covariate structure of who was enrolled and 00:33:49.879 --> 00:33:55.819 what the intervention was, then in a way from one trial you could predict. 00:33:56.389 --> 00:34:00.049 Exactly what the outcome would be in another trial, even if it was a 00:34:00.049 --> 00:34:01.669 slightly different patient population. 00:34:01.669 --> 00:34:07.489 You tell me the admixture of patients, and I will tell you the likely mean effect. 00:34:07.489 --> 00:34:12.319 What, what you're effectively doing is you're saying, I understand the 00:34:12.319 --> 00:34:18.769 relationship between the baseline variables and the, and the randomly 00:34:18.769 --> 00:34:21.829 exposed intervention, and I'll tell you what the effect will be. 00:34:22.399 --> 00:34:22.729 Um. 00:34:26.224 --> 00:34:33.364 The problem here is you quite quickly run outta sample size. 00:34:33.784 --> 00:34:34.174 Um, 00:34:36.544 --> 00:34:41.074 and, and I mean, there's no, there's no real magic around. 00:34:42.319 --> 00:34:42.559 I mean, 00:34:46.619 --> 00:34:53.044 in the ideal world, you could imagine every time. 00:34:53.374 --> 00:34:56.824 You were faced with a clinical decision around which there was 00:34:56.824 --> 00:35:02.974 reasonable uncertainty, then you would almost want to be, uh, 00:35:03.064 --> 00:35:05.914 randomizing among the best options. 00:35:06.724 --> 00:35:15.124 Uh, so if the entire trial experience was like a huge reinforcement learning 00:35:15.124 --> 00:35:22.624 problem where you were sitting in some markoff state where you had a current. 00:35:23.464 --> 00:35:27.934 Of the world where when a patient came into a particular state, based on the 00:35:27.934 --> 00:35:34.054 cova structure among all the different actions that could be taken, you 00:35:34.054 --> 00:35:36.604 would have, you know, a probability. 00:35:36.994 --> 00:35:41.494 Uh, of the eventual best outcome rank, order for all the different options 00:35:41.494 --> 00:35:45.454 and of the 30 treatments you could have given, it turns out that there's two 00:35:45.454 --> 00:35:49.834 or three that all look reasonable, but none of, but they're all quite close. 00:35:50.374 --> 00:35:54.604 And then you could say, okay, every time I see this, it's, it's reasonable 00:35:54.604 --> 00:35:57.844 to randomize to any of those two or three options, but it's unreasonable 00:35:57.844 --> 00:35:59.554 to randomize to the other 27. 00:36:00.589 --> 00:36:05.599 Uh, and then you could imagine if hundreds and hundreds and hundreds of ICUs, for 00:36:05.599 --> 00:36:10.549 example, were all contributing to this ongoing reinforcement learning model, 00:36:10.879 --> 00:36:17.119 that this state could be a known state for the, the agent, the agent being the, the 00:36:17.119 --> 00:36:21.979 collective body watching the accumulation of knowledge until say, several 00:36:21.979 --> 00:36:22.969 hundred patients have gone through. 00:36:23.274 --> 00:36:29.844 And, and then at then at that point you could update the, the probabilities and 00:36:29.844 --> 00:36:33.114 it would turn out, oh, remember how we said there were three choices that looked 00:36:33.114 --> 00:36:34.854 reasonable when you were in this state? 00:36:35.184 --> 00:36:38.964 Turns out one of them is a dud and now we're down to two. 00:36:38.964 --> 00:36:42.384 So you would, you would essentially, this is now you'd move into a new 00:36:42.384 --> 00:36:46.644 state where you had sort of updated probabilities and you would run that, 00:36:46.644 --> 00:36:50.154 and you could imagine that that's really. 00:36:51.709 --> 00:36:55.999 You, you know, this notion of these freestanding RCTs 00:36:56.554 --> 00:36:57.004 Scott Berry: Yeah. 00:36:57.409 --> 00:37:03.559 Derek Angus: that give these single estimates of a patient population 00:37:03.559 --> 00:37:08.509 with an average effect, uh, for a single drug at a single dose that 00:37:08.689 --> 00:37:16.249 seems is a total misrepresentation of the degree of uncertainty in which we 00:37:16.249 --> 00:37:19.549 exist, and it's failing to leverage. 00:37:20.224 --> 00:37:25.114 All of the uncertain clinical decisions that are made in regular care. 00:37:25.174 --> 00:37:32.314 And so in a perfect world then I am massively leaping over all sorts of 00:37:32.314 --> 00:37:37.834 issues about research versus clinical care, the Belmont Report, et cetera. 00:37:38.224 --> 00:37:43.504 But, but a, assuming you had the right ethical framework for this, uh, assuming 00:37:43.504 --> 00:37:48.514 that everyone was read in on this, that, that they all approved, you could imagine. 00:37:49.129 --> 00:37:55.069 A large federated set of learning health systems where all of their electronic 00:37:55.069 --> 00:37:59.299 records are all sort of wired together and you have sort of some sort of 00:37:59.329 --> 00:38:05.959 asynchronous API that could sort of read, write commands that was able to stream 00:38:05.959 --> 00:38:11.989 the information so that just at the moment that the clinician was encountering 00:38:11.989 --> 00:38:14.779 the patient and faced with uncertainty. 00:38:15.619 --> 00:38:17.839 You could then pull from the data. 00:38:17.839 --> 00:38:18.379 Aha. 00:38:18.409 --> 00:38:23.539 I'll tell what this moment looks like is it's a patient with the following 00:38:23.539 --> 00:38:28.789 baseline characteristics about whom we have uncertainty on the right option. 00:38:29.539 --> 00:38:31.699 And so that's like a mini part. 00:38:31.759 --> 00:38:34.639 It's like a mini randomization question. 00:38:35.389 --> 00:38:39.139 Um, and then if people understood. 00:38:39.574 --> 00:38:43.354 Yes, there's legitimate uncertainty and it's, it's within the bounds of 00:38:43.354 --> 00:38:48.244 reasonableness to randomly choose, uh, what treatment they take or even 00:38:48.244 --> 00:38:50.284 to randomly make a recommendation. 00:38:50.344 --> 00:38:54.784 I mean, you could still, if you want, you could still say the 00:38:54.784 --> 00:38:58.024 patient and the physician are allowed to override the recommendation. 00:38:58.939 --> 00:39:02.899 Um, and if they did that a lot, then the trial is getting contaminated. 00:39:02.899 --> 00:39:05.239 But if it turns out that they say, oh, that's, that's not 00:39:05.239 --> 00:39:06.409 an unreasonable suggestion. 00:39:06.409 --> 00:39:07.639 We're both happy with that. 00:39:08.029 --> 00:39:12.349 If that was happening 99% of the time, then effectively you're 00:39:12.349 --> 00:39:17.149 running these huge, large, embedded, ongoing randomizations. 00:39:17.149 --> 00:39:21.649 And if it turns out that among all of pneumonia and sepsis, there are 00:39:22.429 --> 00:39:25.039 a thousand different phenotypes. 00:39:25.729 --> 00:39:32.689 Um, all with d and, and some of them cluster into net benefit, some of the 00:39:32.689 --> 00:39:35.179 net, you know, net harm, et cetera. 00:39:35.689 --> 00:39:40.279 Um, we could be learning about them over time if you, I mean, 00:39:40.279 --> 00:39:43.999 these diseases are common anyways. 00:39:43.999 --> 00:39:46.849 Have I got way too far out over my skis? 00:39:47.329 --> 00:39:47.569 Uh. 00:39:47.644 --> 00:39:49.414 Scott Berry: well, well, no. 00:39:49.414 --> 00:39:51.634 Let's, uh, let's see where you got, you got that. 00:39:51.784 --> 00:39:55.084 Currently, right now, we're not learning from any of these patients. 00:39:55.414 --> 00:39:56.494 They're being treated. 00:39:56.494 --> 00:40:01.534 They're, they're, they're, and so now we're gonna tie this all together. 00:40:01.534 --> 00:40:03.904 We're going to embed randomization. 00:40:04.309 --> 00:40:07.369 And we're going to favor randomization of treatments that are working. 00:40:07.369 --> 00:40:09.259 So we're gonna care for those patients better. 00:40:09.619 --> 00:40:14.059 But it would give us the possibility of learning about heterogeneous treatment 00:40:14.059 --> 00:40:18.989 effects, the, the signatures that benefit, uh, uh, this we're gonna enroll. 00:40:19.629 --> 00:40:22.119 Thousands of patients in this common disease. 00:40:22.479 --> 00:40:26.439 You, you've described a learning healthcare system, and maybe in the 00:40:26.439 --> 00:40:32.349 world of ai, this maybe the thought is, okay, AI is now deciding what to give. 00:40:32.589 --> 00:40:35.079 It's learning from it, it's reinforcing it. 00:40:35.349 --> 00:40:39.849 I mean, this is, this is the future and we kind of wonder why it's not now, but 00:40:39.849 --> 00:40:43.089 the, uh, I, I love where you've gotten. 00:40:43.319 --> 00:40:48.059 Uh, so we've gotten from, uh, these randomized trials that are showing 00:40:48.059 --> 00:40:51.479 differential effect from steroids to a learning healthcare system. 00:40:51.809 --> 00:40:54.719 And I want to come back and talk about that because I think 00:40:54.719 --> 00:40:57.569 that's more than just a idea. 00:40:57.869 --> 00:40:59.879 Uh, I think we, we have some really. 00:41:00.154 --> 00:41:05.074 Uh, great hopes of embedded randomization learning healthcare 00:41:05.074 --> 00:41:08.404 system, call it ai, call it modeling. 00:41:08.464 --> 00:41:11.074 Uh, underneath that is, is pretty exciting. 00:41:11.074 --> 00:41:15.334 So, Derek, I, I, I'm, I'm thrilled that you were able to 00:41:15.334 --> 00:41:16.954 join us here in the interim. 00:41:17.254 --> 00:41:20.344 Um, we do lots of interim analysis here. 00:41:20.404 --> 00:41:22.439 Uh, so we're gonna, we're gonna have you back. 00:41:23.179 --> 00:41:26.779 Uh, to talk about this, to talk about learning healthcare systems 00:41:27.079 --> 00:41:31.699 and maybe we'll have resolution about steroids in sepsis and cap 00:41:33.259 --> 00:41:34.279 Derek Angus: We can only hope. 00:41:35.629 --> 00:41:36.049 Scott Berry: we go. 00:41:37.969 --> 00:41:41.989 Alright, so thank you Derek for joining us here in the interim. 00:41:42.484 --> 00:41:43.939 Derek Angus: thank you so much, Scott.