1 00:00:08,942 --> 00:00:12,579 But remember kind of this basic, that first bullet. 2 00:00:12,579 --> 00:00:14,981 Kind of two types of research. 3 00:00:14,981 --> 00:00:18,218 It's observational versus this kind of experimental interventional. 4 00:00:18,218 --> 00:00:19,419 So the observational, 5 00:00:19,419 --> 00:00:24,224 my goal is to observe and collect data on characteristics of interest 6 00:00:24,224 --> 00:00:28,261 without influencing the participant, the environment, or the disease course. 7 00:00:28,261 --> 00:00:29,462 I literally observing. 8 00:00:29,462 --> 00:00:33,099 I do not want to intervene in any way. 9 00:00:33,099 --> 00:00:35,101 I want to see natural. 10 00:00:35,101 --> 00:00:37,103 Experimental is when you are 11 00:00:37,103 --> 00:00:41,141 -- the researcher, are deliberately influencing the course of events, 12 00:00:41,141 --> 00:00:45,945 at least you're hoping to, and investigating the effect of the intervention 13 00:00:45,945 --> 00:00:48,782 on some carefully selected population of subjects. 14 00:00:49,049 --> 00:00:53,453 I'll say observational is usually a carefully selected set of subjects, too. 15 00:00:53,453 --> 00:00:57,857 When we do experimental subjects on humans, we call them clinical trials 16 00:00:57,857 --> 00:00:58,958 or clinal studies. 17 00:00:58,958 --> 00:01:03,396 Similarly, though, you know, a lot of this work, all of this 18 00:01:03,396 --> 00:01:07,434 applies to animals, it applies to a lot of different projects. 19 00:01:11,504 --> 00:01:15,341 So we're going to cover observational studies in detail next week, 20 00:01:15,341 --> 00:01:19,479 but the general idea here is that you may have case reports, 21 00:01:19,479 --> 00:01:21,915 which is literally the doctor writing down 22 00:01:21,915 --> 00:01:26,086 a set of information, like something looks weird, but I'm going to 23 00:01:26,086 --> 00:01:31,291 write it up in a structured manner so I can share it with other folks. 24 00:01:31,291 --> 00:01:33,726 Several case reports make a case series. 25 00:01:34,794 --> 00:01:37,130 This kind of fundamental epidemiology 101. 26 00:01:37,130 --> 00:01:41,468 It's also because a pharmacist noticed something looked odd and started 27 00:01:41,468 --> 00:01:47,340 working on a set of case series and case reports that we discovered AIDS. 28 00:01:47,340 --> 00:01:51,644 So you used to get -- CDC actually now publishes 29 00:01:51,644 --> 00:01:55,982 it more electronically, but you had Morbidity and Mortality Weekly Report. 30 00:01:55,982 --> 00:02:01,454 So when I was in school, every Friday we went to read this report, 31 00:02:01,454 --> 00:02:06,159 to see kind of what looked new and weird around the country. 32 00:02:06,159 --> 00:02:09,295 What we should have our eyes open for. 33 00:02:09,295 --> 00:02:11,764 Those are usually case series. 34 00:02:11,764 --> 00:02:16,436 You still see them published today in a lot of journals. 35 00:02:16,436 --> 00:02:19,806 Cross-sectional prevalence surveys. This is a snapshot picture. 36 00:02:19,806 --> 00:02:25,311 So this might be the National Health Interview Survey in the United States. 37 00:02:25,311 --> 00:02:26,579 Case control studies. 38 00:02:26,579 --> 00:02:32,919 We'll talk a little bit about this, but usually you get a series of disease 39 00:02:32,919 --> 00:02:39,259 cases, and try to find some match controls, and figure out what's different between them. 40 00:02:39,659 --> 00:02:46,466 If you have a really rare disease, this is a very useful type of study to do, 41 00:02:46,466 --> 00:02:52,071 to try to figure out a list of reasons that you might have disease. 42 00:02:52,071 --> 00:02:54,073 Cohort studies that are longitudinal. 43 00:02:54,073 --> 00:02:58,478 A lot of times -- so when we had major disasters, 44 00:02:58,478 --> 00:03:03,683 we will follow the healthcare workers, or the people that are cleaning up 45 00:03:03,683 --> 00:03:08,121 those disaster sites long-term to see if they have psychological issues, 46 00:03:08,121 --> 00:03:12,525 if they have respiratory-related issues, other problems that come up. 47 00:03:12,525 --> 00:03:14,127 Natural history studies. 48 00:03:14,127 --> 00:03:16,930 You may have a group of patients, 49 00:03:16,930 --> 00:03:21,734 and you're going to follow them, and see how they actually age. 50 00:03:22,168 --> 00:03:24,571 You may see how their disease progresses. 51 00:03:24,571 --> 00:03:28,708 The NIH at the Clinical Center does quite a few of these. 52 00:03:28,708 --> 00:03:30,443 And then the ecological studies. 53 00:03:30,443 --> 00:03:34,581 This is data that's on a population rather than an individual level. 54 00:03:34,581 --> 00:03:38,351 So like I said, we'll talk more about these next week. 55 00:03:40,687 --> 00:03:41,521 Then we 56 00:03:41,521 --> 00:03:46,125 have these kind of -- some groups call them quasi-experimental studies. 57 00:03:46,125 --> 00:03:50,697 These are those one or single arm, nonrandomized, interventional studies. Dr. 58 00:03:50,697 --> 00:03:56,502 Gallin talked about several of these actually if you think about his historical lecture. 59 00:03:56,502 --> 00:03:59,005 You don't have a control group. 60 00:03:59,005 --> 00:04:02,342 They tend to be early in the investigation. 61 00:04:02,342 --> 00:04:09,015 Sometimes you may have a concurrent control group, so I may decide I'm going to bathe 62 00:04:09,015 --> 00:04:16,089 one side of the hall in my hospital, but not bathe the other side of the hall. 63 00:04:16,356 --> 00:04:21,728 I can do some weird, interesting things, but I'm not randomly choosing it. 64 00:04:21,728 --> 00:04:24,230 I just kind of allocate it. 65 00:04:25,064 --> 00:04:28,768 Then you sometimes have things called historically controlled studies. 66 00:04:28,768 --> 00:04:34,540 So pediatric oncology we used to do these where patients that -- you know, 67 00:04:34,540 --> 00:04:39,512 we basically only had enough patients to put everybody on the therapy. 68 00:04:39,512 --> 00:04:45,718 So we said, well, we'll use old patient information as kind of our control group. 69 00:04:46,552 --> 00:04:50,256 So that's kind of that early intervention-based research spectrum. 70 00:04:50,256 --> 00:04:54,727 So I talk about the sometimes it's interventional, sometimes not. 71 00:04:54,727 --> 00:04:56,095 But that quasi-experimental, 72 00:04:56,095 --> 00:05:00,099 those pre-clinical studies, phase 0, those are early studies. 73 00:05:00,099 --> 00:05:05,471 But all of this is setting the foundation for trying to do 74 00:05:05,471 --> 00:05:10,843 what's a phase 1 study or those dose finding studies many times. 75 00:05:10,843 --> 00:05:14,414 You know, in this patient population, what's tolerable? 76 00:05:14,414 --> 00:05:20,987 You know, and what might be sometimes we also look to see early efficacy there, 77 00:05:20,987 --> 00:05:25,892 or at least some change that says we might think we'll have efficacy down the line. 78 00:05:25,892 --> 00:05:30,763 Dig into these early and late phase 2 studies, again, we're looking a lot at safety 79 00:05:30,763 --> 00:05:35,968 like we are in phase 1, but we're starting to get a better idea of the dose. 80 00:05:35,968 --> 00:05:38,104 We're starting to get a better idea 81 00:05:38,104 --> 00:05:42,709 of how we should deliver a medication, or some type of medical product, or therapy. 82 00:05:43,276 --> 00:05:49,515 We're trying to get an idea of who should be in these studies and not. 83 00:05:49,515 --> 00:05:53,252 Phase 3, or what we typically call pivotal trials. 84 00:05:53,252 --> 00:05:56,155 These are your major, large efficacy studies. 85 00:05:56,155 --> 00:06:00,727 Phase 4, for me in the FDA world, is post market. 86 00:06:00,727 --> 00:06:06,132 So we've kind of decided there's efficacy, but if I put this out 87 00:06:06,132 --> 00:06:10,703 in the general population, do I still see safety and effectiveness? 88 00:06:12,338 --> 00:06:16,209 You also get into these dissemination and implementation studies. 89 00:06:16,209 --> 00:06:16,642 Great. 90 00:06:16,642 --> 00:06:23,049 You think that if you make this change your hospital process that you will improve, 91 00:06:23,049 --> 00:06:28,221 you know, let's say it's rates of some type of hospital-acquired infection. 92 00:06:28,221 --> 00:06:32,058 You've done this at your very rigorous, focused hospital. 93 00:06:32,058 --> 00:06:36,796 Is that going to work at the middle of nowhere hospital? 94 00:06:36,796 --> 00:06:41,501 Is it going to work in a really busy public hospital? 95 00:06:41,501 --> 00:06:45,238 Dissemination implementation is can I take all of the information 96 00:06:45,238 --> 00:06:49,275 about how to deliver a therapy, and how deliver an intervention, 97 00:06:49,275 --> 00:06:52,945 and actually do it everywhere in the real world. 98 00:06:52,945 --> 00:06:56,649 You also then see comparative or cost effectiveness studies. 99 00:06:56,649 --> 00:07:02,522 So there's a large study done many years ago by the National Institute of Mental Health 100 00:07:02,522 --> 00:07:07,293 where they took several different therapies for people who had major depressive disorder. 101 00:07:07,293 --> 00:07:10,930 And they said, okay, we're putting them head to head. 102 00:07:10,930 --> 00:07:12,398 That's a comparative study. 103 00:07:12,398 --> 00:07:18,237 But your ideal study, the problem that comes up, is that we have these ideals, right. 104 00:07:18,237 --> 00:07:23,309 Whenever anybody looks at your study design, they are going to say, "I expect 105 00:07:23,309 --> 00:07:28,047 to have a treatment and a control arm." What about all those studies 106 00:07:28,047 --> 00:07:30,216 that don't have a control arms? 107 00:07:31,350 --> 00:07:32,118 They expect 108 00:07:32,118 --> 00:07:36,823 you to have parallel groups, that you're going to have randomized people, 109 00:07:36,823 --> 00:07:42,695 to two different arms of a study, and you're going to watch these folks simultaneously. 110 00:07:42,695 --> 00:07:44,664 Well, sometimes that's not feasible. 111 00:07:44,664 --> 00:07:47,400 They expect you to look for superiority. 112 00:07:47,400 --> 00:07:50,937 You know, drug A is better than drug B. 113 00:07:50,937 --> 00:07:57,210 Well, maybe, you know, drug A costs $30,000 a year and drug B costs 30 cents. 114 00:07:58,044 --> 00:08:01,914 Maybe it's more accessible to use that or maybe there're 115 00:08:01,914 --> 00:08:06,552 a lot fewer side effects with drug B than drug A. 116 00:08:06,552 --> 00:08:07,320 Prospective. 117 00:08:07,320 --> 00:08:11,224 They expect you to be following people into the future. 118 00:08:11,224 --> 00:08:17,797 If there are only 34 people in the world with your disease, you may not be able 119 00:08:17,797 --> 00:08:22,068 to follow them all prospectively in a randomized, parallel arm study. 120 00:08:22,068 --> 00:08:25,171 They expect to be double blinded and masked. 121 00:08:26,072 --> 00:08:29,208 Well, what if you're doing surgery versus non-surgical intervention. 122 00:08:29,208 --> 00:08:34,780 We used to actually blind those studies, but you may not be able to blind them. 123 00:08:34,780 --> 00:08:39,919 Although, sometimes you may say, well, kind of what am I trying to look at? 124 00:08:40,152 --> 00:08:44,590 Do I want to control for all the risks of opening somebody up 125 00:08:44,590 --> 00:08:48,794 and the extra infections they may get from opening and closing them? 126 00:08:48,794 --> 00:08:50,496 Or do I not? 127 00:08:50,496 --> 00:08:53,733 What is your exact question you want to answer? 128 00:08:53,733 --> 00:08:58,738 But if you're looking at a pill versus an IV in a pediatric population, 129 00:08:58,738 --> 00:09:03,042 you're probably not going to be able to give a fake IV. 130 00:09:03,743 --> 00:09:05,878 And they expect a randomized study. 131 00:09:05,878 --> 00:09:10,550 A lot of studies can be randomized if you get inventive, and you're 132 00:09:10,550 --> 00:09:14,820 working in the right population, but not all studies can be randomized. 133 00:09:15,121 --> 00:09:20,293 If I want to look at long-term antiretroviral therapy in HIV patients, it's 134 00:09:20,293 --> 00:09:25,097 going to be really hard for me to run a randomized trial. 135 00:09:25,097 --> 00:09:29,902 So we have these gold standards but sometimes we have to explain 136 00:09:29,902 --> 00:09:33,706 why we need to be a little bit bronze. 137 00:09:33,706 --> 00:09:38,711 Now the next two slides are a handful of studies from BMJ 138 00:09:38,711 --> 00:09:43,516 back in 2013 because it was actually easy to lift the information. 139 00:09:43,783 --> 00:09:47,453 You'll see actually I give a lot of examples from BMJ. 140 00:09:47,453 --> 00:09:49,789 That's because you can access it publicly. 141 00:09:49,789 --> 00:09:54,460 It's open for anybody in the world so what I show you is something 142 00:09:54,460 --> 00:09:57,463 I want you all to be able to access. 143 00:09:57,463 --> 00:10:00,466 If there are articles that are not publicly available, 144 00:10:00,466 --> 00:10:05,771 we will put them up as part of the course information along with my slides. 145 00:10:05,771 --> 00:10:10,776 So in BMJ, they had four articles in their research section this one week. 146 00:10:10,776 --> 00:10:13,779 Noninvasive versus evasive respiratory support, a systematic review meta-analysis. 147 00:10:14,413 --> 00:10:19,085 Another one was a multicenter randomized control trial that was blinded. 148 00:10:19,085 --> 00:10:21,621 At least the researchers were blinded. 149 00:10:21,621 --> 00:10:25,858 You had a population-based cohort study and a large-scale survey. 150 00:10:25,858 --> 00:10:28,394 A lot of different research there. 151 00:10:28,394 --> 00:10:33,032 Only one of those projects was actually randomized, double-blind, control trial. 152 00:10:33,032 --> 00:10:38,971 There's a lot of different types of research you can do that's meaningful. 153 00:10:38,971 --> 00:10:43,175 But as you're doing it, you still have to 154 00:10:43,175 --> 00:10:46,579 distinguish the observational studies from your randomized studies. 155 00:10:47,480 --> 00:10:52,318 A lot of times we start doing these analyses of observational studies 156 00:10:52,318 --> 00:10:54,754 thinking that they were a controlled, 157 00:10:54,754 --> 00:10:57,990 randomized trial, but that tacit assumption of randomness 158 00:10:57,990 --> 00:11:02,828 is what makes a lot of other assumptions work in statistics land. 159 00:11:02,828 --> 00:11:08,501 So, you really have to do a lot of extra work when you're analyzing 160 00:11:08,501 --> 00:11:10,136 observational trial data. 161 00:11:10,136 --> 00:11:15,374 The idea is in a nonrandomized study you can only show associations. 162 00:11:15,374 --> 00:11:19,045 You're never going to know all possible confounders. 163 00:11:19,045 --> 00:11:23,616 In a randomized studies, you can show association and causation. 164 00:11:23,616 --> 00:11:28,621 Now in a well-done nonadaptive randomization, so we'll get to that 165 00:11:28,621 --> 00:11:33,659 in a few weeks, the unknown confounder should not create problems. 166 00:11:33,659 --> 00:11:40,533 If you are doing an adaptive trial, unknown confounders can cause a lot of problems. 167 00:11:40,533 --> 00:11:43,269 But in nonadaptive studies, nonadaptive randomizations, 168 00:11:43,869 --> 00:11:47,973 the general idea is that unknown confounders should not create problems. 169 00:11:47,973 --> 00:11:52,378 But again, always remember that your questions are going to come first. 170 00:11:52,378 --> 00:11:57,149 So, as you're making all these changes, all these things that you're thinking about 171 00:11:57,149 --> 00:12:01,921 with your patients, and what's going to work, are you still answering the fundamental 172 00:12:01,921 --> 00:12:02,955 question of interest?