WEBVTT 00:00:00.000 --> 00:00:02.540 how can we really know if one treatment is actually 00:00:02.540 --> 00:00:05.000 better than another? Especially in orthopedics, 00:00:05.019 --> 00:00:07.740 it's not just guesswork, is it? No, absolutely 00:00:07.740 --> 00:00:11.000 not. It relies on, well, a really meticulous 00:00:11.000 --> 00:00:13.759 process of gathering solid evidence. And that's 00:00:13.759 --> 00:00:15.880 what we're diving into today, isn't it? The foundations, 00:00:16.059 --> 00:00:18.679 the methods behind clinical orthopedic research. 00:00:19.120 --> 00:00:21.679 Precisely. How we actually build that evidence 00:00:21.679 --> 00:00:24.019 base. We've got a stack of source material here, 00:00:24.399 --> 00:00:26.539 research papers, notes, and we're going to pull 00:00:26.539 --> 00:00:28.859 out the really crucial bits about how studies 00:00:28.859 --> 00:00:32.719 are designed, run, and importantly, how you can 00:00:32.719 --> 00:00:35.159 understand them. Which is key for anyone in the 00:00:35.159 --> 00:00:37.119 field, really, whether you're doing the research 00:00:37.119 --> 00:00:39.759 or using it to make clinical decisions. And I'm 00:00:39.759 --> 00:00:41.960 thrilled we have Professor Mo Imam with us to 00:00:41.960 --> 00:00:44.659 guide us through all this complexity. Welcome. 00:00:45.119 --> 00:00:47.579 Thank you. It's a pleasure. Thinking about how 00:00:47.579 --> 00:00:50.219 we know what we know, it's fundamental, isn't 00:00:50.219 --> 00:00:52.630 it? It gives you real insight into the evidence 00:00:52.630 --> 00:00:54.649 you're using every day. It really does. It's 00:00:54.649 --> 00:00:57.609 like knowing how the engine works. So over this 00:00:57.609 --> 00:01:00.649 deep dive, we'll cover the ethics, the absolute 00:01:00.649 --> 00:01:04.609 must do's. Non -negotiable. And then the practicalities 00:01:04.609 --> 00:01:07.909 of study design. Things like sample size, blinding, 00:01:08.409 --> 00:01:11.359 how do we measure. what actually matters to patients. 00:01:11.719 --> 00:01:13.620 Yes. Outcome measures are critical. And then 00:01:13.620 --> 00:01:16.120 understanding the data, the statistics, the different 00:01:16.120 --> 00:01:18.120 study types. Right. And we'll touch on the bigger 00:01:18.120 --> 00:01:20.180 picture, too. Things like multi -center trials, 00:01:20.680 --> 00:01:23.480 registries. Indeed. And even the economics of 00:01:23.480 --> 00:01:25.780 health care, which is becoming increasingly important. 00:01:25.879 --> 00:01:28.700 The whole ecosystem, really. Fantastic. A proc 00:01:28.700 --> 00:01:31.939 and deep dive. So let's start with a concept 00:01:31.939 --> 00:01:35.040 that's everywhere now. Evidence -based medicine. 00:01:35.540 --> 00:01:37.719 EBM. What does it actually mean? Where did it 00:01:37.719 --> 00:01:41.159 come from? OK. EBM. Well, at its heart, it's 00:01:41.159 --> 00:01:44.680 about using the best current evidence conscientiously, 00:01:45.260 --> 00:01:47.780 explicitly, you know, really thoughtfully when 00:01:47.780 --> 00:01:50.700 making decisions about individual patients. That 00:01:50.700 --> 00:01:52.920 definition from Gordon Geyette and the McMaster 00:01:52.920 --> 00:01:56.120 Group back in 91 really crystallized it. But 00:01:56.120 --> 00:01:58.079 it's not just about the research papers, is it? 00:01:58.120 --> 00:02:00.180 No, that's a common misconception. It's a three 00:02:00.180 --> 00:02:02.680 -legged stool, really. It's integrating that 00:02:02.680 --> 00:02:04.500 best external evidence with your own clinical 00:02:04.500 --> 00:02:07.319 expertise and crucially with the patient's values 00:02:07.319 --> 00:02:09.620 and preferences. All three are vital. That makes 00:02:09.620 --> 00:02:13.039 sense. A balance. Exactly. And while that definition 00:02:13.039 --> 00:02:15.580 is relatively recent, maybe late 20th century, 00:02:16.080 --> 00:02:18.340 the idea of looking systematically at outcomes. 00:02:19.360 --> 00:02:21.669 Well, that goes back much further. The source 00:02:21.669 --> 00:02:24.490 material mentions Dr. Ernest Codman. Ah yes, 00:02:24.669 --> 00:02:27.110 Codman. His story is quite something, isn't it? 00:02:27.150 --> 00:02:30.990 His end results idea. It really is. He was proposing, 00:02:31.150 --> 00:02:33.830 back in the early 1900s, that doctors should 00:02:33.830 --> 00:02:36.550 follow every patient to see the final outcome. 00:02:36.729 --> 00:02:39.909 Every single one. Every one. And if the result 00:02:39.909 --> 00:02:42.610 wasn't good, they needed to investigate why. 00:02:43.229 --> 00:02:45.189 Systematically tracking failures to learn from 00:02:45.189 --> 00:02:48.150 them. Sounds completely logical to us now. Why 00:02:48.150 --> 00:02:50.560 was that so radical then? Well, It flew in the 00:02:50.560 --> 00:02:52.879 face of the established practice, which was much 00:02:52.879 --> 00:02:55.099 more based on authority and tradition, perhaps 00:02:55.099 --> 00:02:58.360 anecdote, scrutinizing every failure publicly 00:02:58.360 --> 00:03:01.520 almost. It wasn't the done thing. So it caused 00:03:01.520 --> 00:03:04.020 friction. Significant friction. The source has 00:03:04.020 --> 00:03:05.659 mentioned it cost him his job at Massachusetts 00:03:05.659 --> 00:03:08.520 General Hospital. Wow. But his insistence on 00:03:08.520 --> 00:03:10.900 patient outcomes as the ultimate measure, well, 00:03:11.020 --> 00:03:13.319 that's now seen as a cornerstone, a really foundational 00:03:13.319 --> 00:03:15.379 step towards the empirical medicine we practice 00:03:15.379 --> 00:03:18.330 today. A difficult path for him, but crucial 00:03:18.330 --> 00:03:21.250 for progress. So we have Codman's early ideas, 00:03:21.629 --> 00:03:25.490 then the formal EBM definition. How did it become 00:03:25.490 --> 00:03:28.270 so widespread? It was a confluence of factors, 00:03:28.530 --> 00:03:30.270 really. You had the intellectual framework from 00:03:30.270 --> 00:03:33.710 Guyatt and others, but what really powered its... 00:03:33.770 --> 00:03:37.270 almost explosive integration was the simultaneous 00:03:37.270 --> 00:03:40.090 revolution in information technology. Ah, the 00:03:40.090 --> 00:03:42.990 rise of computers and databases. Exactly. Suddenly, 00:03:43.210 --> 00:03:45.030 you could search huge amounts of literature, 00:03:45.469 --> 00:03:48.009 manage large data sets, synthesize evidence in 00:03:48.009 --> 00:03:50.810 ways that were just impossible before. The technology 00:03:50.810 --> 00:03:53.930 made the EBM concept practically achievable on 00:03:53.930 --> 00:03:56.710 a grand scale. So the tech unlocked the potential 00:03:56.710 --> 00:03:59.520 of the idea. Precisely. It meant a scientific, 00:03:59.539 --> 00:04:02.300 data -driven approach could genuinely complement 00:04:02.300 --> 00:04:04.280 and sometimes challenge traditional ways of thinking 00:04:04.280 --> 00:04:06.659 in clinical practice. But before we even think 00:04:06.659 --> 00:04:08.719 about generating that data, there's the ethical 00:04:08.719 --> 00:04:10.879 foundation. It's mentioned as absolutely non 00:04:10.879 --> 00:04:13.740 -negotiable. Why is protecting research participants 00:04:13.740 --> 00:04:16.740 so critical, especially in orthopedics? It's 00:04:16.740 --> 00:04:19.540 paramount. Fundamentally, clinical research involves 00:04:19.540 --> 00:04:21.860 asking individuals to potentially accept some 00:04:21.860 --> 00:04:24.699 risk, whether physical, psychological, or to 00:04:24.699 --> 00:04:27.040 their privacy, for the benefit of others, for 00:04:27.040 --> 00:04:29.769 future patients. The ethical framework is there 00:04:29.769 --> 00:04:32.870 to ensure the rights, dignity and well -being 00:04:32.870 --> 00:04:35.699 of those participants. always come first. Oh. 00:04:35.779 --> 00:04:38.120 Always. We're guests of our patients, as you 00:04:38.120 --> 00:04:40.860 said. Absolutely. And overseeing this are institutional 00:04:40.860 --> 00:04:44.160 review boards, IRBs, or ethics committees. Yes, 00:04:44.259 --> 00:04:46.899 IRBs or IECs, depending on where you are. Their 00:04:46.899 --> 00:04:49.240 job is essentially to be the advocates for the 00:04:49.240 --> 00:04:51.980 research participants. They scrutinize the research 00:04:51.980 --> 00:04:54.899 plan, the protocol. To ensure. To ensure risks 00:04:54.899 --> 00:04:57.779 are minimized and justified by potential benefits, 00:04:58.199 --> 00:05:00.420 that patient selection is fair and, truthfully, 00:05:00.879 --> 00:05:03.279 that the informed consent process is robust. 00:05:03.310 --> 00:05:06.240 and truly allows for an informed choice. Their 00:05:06.240 --> 00:05:08.399 mandate, particularly in the U .S., comes from 00:05:08.399 --> 00:05:10.240 principles like those in the Belmont report, 00:05:10.620 --> 00:05:12.579 formalized in regulations like the Common Rule. 00:05:12.959 --> 00:05:15.120 They're gatekeepers, ensuring research is conducted 00:05:15.120 --> 00:05:17.879 ethically. Does every single study need IRB approval 00:05:17.879 --> 00:05:20.639 then, or are there exceptions? Not absolutely 00:05:20.639 --> 00:05:23.060 everything. The sources point out that things 00:05:23.060 --> 00:05:26.100 like internal quality improvement projects, perhaps 00:05:26.100 --> 00:05:28.860 some types of case reports focusing on just one 00:05:28.860 --> 00:05:31.839 or two patients. These might not always meet 00:05:31.839 --> 00:05:34.060 the formal definition of research designed to 00:05:34.060 --> 00:05:36.160 create generalizable knowledge. So they might 00:05:36.160 --> 00:05:38.980 not need formal review. They might not. But, 00:05:39.100 --> 00:05:41.279 and this is a critical piece of advice, if there 00:05:41.279 --> 00:05:43.839 is any doubt at all, you absolutely must consult 00:05:43.839 --> 00:05:46.420 your IRB or ethics committee before you start 00:05:46.420 --> 00:05:48.899 anything. It protects everyone involved. Better 00:05:48.899 --> 00:05:51.980 to ask than to make assumptions. Sound advice. 00:05:52.500 --> 00:05:55.220 And a cornerstone of that protection is informed 00:05:55.220 --> 00:05:58.360 consent. It's called a fundamental pillar. What 00:05:58.360 --> 00:06:01.079 does informed really mean here? It's about genuine 00:06:01.079 --> 00:06:03.930 understanding and voluntary choice. It's not 00:06:03.930 --> 00:06:06.490 just giving someone a leaflet. You have to ensure 00:06:06.490 --> 00:06:09.629 the potential participant truly grasps the study's 00:06:09.629 --> 00:06:12.269 purpose, what procedures are involved, the potential 00:06:12.269 --> 00:06:14.910 risks and benefits, what alternatives exist, 00:06:16.029 --> 00:06:19.029 and critically, that taking part is entirely 00:06:19.029 --> 00:06:21.829 voluntary and they can withdraw at any point 00:06:21.829 --> 00:06:24.350 without any penalty, without affecting their 00:06:24.350 --> 00:06:26.509 normal care. They need to be able to explain 00:06:26.509 --> 00:06:29.490 it back to you in a way. It's an ongoing conversation, 00:06:29.730 --> 00:06:31.959 not just a signature on a form. That sounds like 00:06:31.959 --> 00:06:34.040 it could be tricky, especially with certain patient 00:06:34.040 --> 00:06:37.339 groups. It absolutely can be. The sources highlight 00:06:37.339 --> 00:06:40.019 the need for extra care with vulnerable populations 00:06:40.019 --> 00:06:42.680 of children, people with cognitive impairments, 00:06:43.100 --> 00:06:45.040 perhaps even situations where there's a power 00:06:45.040 --> 00:06:47.579 imbalance, like if the researcher is also the 00:06:47.579 --> 00:06:49.480 treating clinician. So how do you manage that? 00:06:49.779 --> 00:06:53.220 Well, for those lacking capacity, you need proxy 00:06:53.220 --> 00:06:55.920 consent from a legal representative. But even 00:06:55.920 --> 00:06:58.759 then, the ethical obligation is to respect the 00:06:58.759 --> 00:07:01.240 individual's potential views and best interests 00:07:01.240 --> 00:07:04.160 as much as possible. It requires real sensitivity 00:07:04.160 --> 00:07:06.720 and ensuring the process isn't coercive in any 00:07:06.720 --> 00:07:09.560 way. Respecting autonomy seems key. Now, once 00:07:09.560 --> 00:07:11.699 someone's in a study, what about their data? 00:07:12.279 --> 00:07:15.120 Protecting confidentiality? Hugely important. 00:07:15.339 --> 00:07:17.680 Data protection regulations like EPA in the U 00:07:17.680 --> 00:07:21.420 .S. or GDPR in Europe are very strict. In practice, 00:07:21.439 --> 00:07:23.939 it means using study codes instead of names on 00:07:23.939 --> 00:07:26.740 data forms. keeping records secure, ensuring 00:07:26.740 --> 00:07:29.240 data is accurate and legible. And the consequences 00:07:29.240 --> 00:07:31.959 if things go wrong? They can be severe, as the 00:07:31.959 --> 00:07:34.360 source material makes clear. Data breaches can 00:07:34.360 --> 00:07:37.000 lead to reputational damage, research being halted, 00:07:37.160 --> 00:07:39.500 significant fines, and potentially even legal 00:07:39.500 --> 00:07:42.779 action against researchers. Trust is incredibly 00:07:42.779 --> 00:07:45.759 hard to regain once lost. A heavy responsibility. 00:07:46.360 --> 00:07:48.699 Beyond patient data, there's the issue of conflicts 00:07:48.699 --> 00:07:51.699 of interest, COI. How is that defined in a research 00:07:51.699 --> 00:07:53.959 context? The Institute of Medicine definition 00:07:53.959 --> 00:07:57.180 is widely used. It's basically when circumstances 00:07:57.180 --> 00:07:59.759 create a risk that a researcher's professional 00:07:59.759 --> 00:08:01.740 judgment about a primary interest, like patient 00:08:01.740 --> 00:08:04.540 welfare or research integrity, might be unduly 00:08:04.540 --> 00:08:06.600 influenced by a secondary interest. Like financial 00:08:06.600 --> 00:08:09.329 gain. That's the most common one, yes. especially 00:08:09.329 --> 00:08:11.750 with industry funding. You often have the researcher, 00:08:12.009 --> 00:08:13.629 their institution, and the company providing 00:08:13.629 --> 00:08:16.290 funding. Their interests might not always align 00:08:16.290 --> 00:08:19.649 perfectly. An ethical issue arises if financial 00:08:19.649 --> 00:08:22.370 ties could potentially bias how the research 00:08:22.370 --> 00:08:25.410 is conducted, analyzed, or reported. And industry 00:08:25.410 --> 00:08:27.610 funding is pretty significant, isn't it? It is. 00:08:27.810 --> 00:08:31.009 The source notes figures like 71 % of R &D funding 00:08:31.009 --> 00:08:33.309 in the U .S. coming from industry. And there's 00:08:33.309 --> 00:08:35.809 documented evidence, quite a lot of it, actually 00:08:35.809 --> 00:08:39.210 showing that industry -sponsored trials are statistically 00:08:39.210 --> 00:08:42.110 more likely to report results favorable to the 00:08:42.110 --> 00:08:43.990 sponsor's product. That doesn't necessarily mean 00:08:43.990 --> 00:08:46.610 the research is flawed though. Not necessarily, 00:08:46.789 --> 00:08:49.429 no, but it highlights a potential for bias that 00:08:49.429 --> 00:08:51.679 absolutely must be managed transparently. How 00:08:51.679 --> 00:08:53.879 do researchers navigate that? Are there guidelines? 00:08:54.419 --> 00:08:57.299 Yes. Professional bodies like the AAOS provide 00:08:57.299 --> 00:08:59.799 guidance. For example, suggesting researchers 00:08:59.799 --> 00:09:01.519 shouldn't trade stock in the funding company 00:09:01.519 --> 00:09:04.480 during the project. Or, if a surgeon has designed 00:09:04.480 --> 00:09:06.980 an implant and receives royalties, research on 00:09:06.980 --> 00:09:08.779 that specific implant should ideally be done 00:09:08.779 --> 00:09:11.220 by someone completely independent. What about 00:09:11.220 --> 00:09:14.220 consulting fees? Legitimate consulting is fine, 00:09:14.500 --> 00:09:16.220 provided the compensation is fair market value 00:09:16.220 --> 00:09:19.240 for the work done and, crucially, It's fully 00:09:19.240 --> 00:09:22.700 disclosed. Transparency is key. Disclosure and 00:09:22.700 --> 00:09:26.000 management, not necessarily prohibition. Are 00:09:26.000 --> 00:09:28.679 there system -level responses to COI concerns? 00:09:28.820 --> 00:09:31.379 Yes. A major one is mandatory clinical trial 00:09:31.379 --> 00:09:34.480 registration. Public databases like clinicaltrials 00:09:34.480 --> 00:09:39.299 .gov, ISRCTN, or the WHO registries, they require 00:09:39.299 --> 00:09:42.139 researchers to outline their study design, their 00:09:42.139 --> 00:09:45.039 methods, their primary outcome measures before 00:09:45.039 --> 00:09:47.470 they start recruiting anyone. How does registering 00:09:47.470 --> 00:09:50.110 beforehand help? It locks the plan in publicly. 00:09:50.230 --> 00:09:52.149 It makes it much harder to engage in things like 00:09:52.149 --> 00:09:54.070 publication bias, where only positive results 00:09:54.070 --> 00:09:56.429 get published, or selective outcome reporting, 00:09:56.710 --> 00:09:58.509 where maybe you change your main outcome measure 00:09:58.509 --> 00:10:00.490 after you've seen the data because another one 00:10:00.490 --> 00:10:03.070 looks better. So it increases transparency. Hugely. 00:10:03.750 --> 00:10:05.809 It holds researchers accountable to their original 00:10:05.809 --> 00:10:09.399 plan. Many major journals following the ICMJE 00:10:09.399 --> 00:10:11.419 guidelines now won't even consider publishing 00:10:11.419 --> 00:10:13.620 a trial unless it was registered beforehand. 00:10:13.860 --> 00:10:18.139 A powerful incentive for good practice. Right, 00:10:18.259 --> 00:10:20.860 so ethical foundations are set. Let's move to 00:10:20.860 --> 00:10:23.220 designing the study itself. It all starts with 00:10:23.220 --> 00:10:24.860 a research question, doesn't it? Absolutely. 00:10:25.039 --> 00:10:27.240 A clear, focused question is the starting point. 00:10:27.480 --> 00:10:29.940 It might come from clinical observation, spotting 00:10:29.940 --> 00:10:32.419 a gap in the literature, wanting to follow up 00:10:32.419 --> 00:10:34.220 on earlier work. But how do you know if it's 00:10:34.220 --> 00:10:37.000 a good question? One worth pursuing. There's 00:10:37.000 --> 00:10:39.679 a useful framework called the finer criteria. 00:10:40.100 --> 00:10:41.960 Is the question feasible? Can you actually do 00:10:41.960 --> 00:10:44.659 it? Interesting? Will people care? Novel? Does 00:10:44.659 --> 00:10:47.019 it add something new? Ethical? Can it be done 00:10:47.019 --> 00:10:50.539 ethically? And relevant? Does it matter to practice 00:10:50.539 --> 00:10:54.879 or policy? Finer. That's neat. Feasible, interesting, 00:10:55.080 --> 00:10:57.559 novel, ethical, relevant. Exactly. It helps you 00:10:57.559 --> 00:10:59.820 refine your initial idea into something concrete 00:10:59.820 --> 00:11:02.279 and justifiable. Once you have that finer question, 00:11:02.299 --> 00:11:04.629 how do you structure it for a study plan? The 00:11:04.629 --> 00:11:07.450 PICOT format is incredibly useful here, especially 00:11:07.450 --> 00:11:09.710 for clinical questions. Population, who are the 00:11:09.710 --> 00:11:11.730 patients? Intervention, what's being tested? 00:11:12.350 --> 00:11:13.990 Comparator, what's it being compared against? 00:11:14.250 --> 00:11:17.110 Outcome, what are you measuring? And timeframe, 00:11:17.250 --> 00:11:20.429 over what period? PICOT, population, intervention, 00:11:20.830 --> 00:11:24.429 comparator, outcome, timeframe. Using PICOT clarifies 00:11:24.429 --> 00:11:26.929 exactly what the study is about. It makes communication 00:11:26.929 --> 00:11:29.450 much clearer. So PICOT defines the question. 00:11:29.690 --> 00:11:32.600 Then comes the study protocol. The master plan. 00:11:32.940 --> 00:11:35.279 Precisely. The protocol lays out everything in 00:11:35.279 --> 00:11:38.200 detail. The background, the objectives, the exact 00:11:38.200 --> 00:11:40.879 methods for recruitment, intervention, data collection, 00:11:41.360 --> 00:11:44.139 the statistical analysis plan, ethical considerations, 00:11:44.539 --> 00:11:47.600 timelines, everything. And the key point the 00:11:47.600 --> 00:11:50.600 source makes is this must be finalized before 00:11:50.600 --> 00:11:53.059 you recruit the first participant. Why is that 00:11:53.059 --> 00:11:55.879 timing so critical? It prevents people from changing 00:11:55.879 --> 00:11:58.250 the rules halfway through the game. essentially. 00:11:58.690 --> 00:12:00.970 If the methods or analysis plan aren't fixed 00:12:00.970 --> 00:12:03.850 beforehand, there's a risk, conscious or unconscious, 00:12:04.009 --> 00:12:06.429 of making changes based on early results, which 00:12:06.429 --> 00:12:09.350 can introduce bias. It ensures the study is conducted 00:12:09.350 --> 00:12:11.370 systematically as planned. And a crucial part 00:12:11.370 --> 00:12:13.629 of that plan is sample size calculation, getting 00:12:13.629 --> 00:12:15.950 the numbers right. Yes, vital for statistical 00:12:15.950 --> 00:12:18.669 power. Power is the study's ability to detect 00:12:18.669 --> 00:12:21.330 a true difference or effect if one actually exists. 00:12:21.370 --> 00:12:23.669 And if the sample size is too small, then the 00:12:23.669 --> 00:12:26.340 study is underpowered. You risk a type 2 error, 00:12:26.639 --> 00:12:28.740 a false negative. You might conclude there's 00:12:28.740 --> 00:12:31.639 no difference between treatments when, in reality, 00:12:31.919 --> 00:12:34.240 there is one. But your study just wasn't big 00:12:34.240 --> 00:12:36.700 enough to find it. Is that a common problem in 00:12:36.700 --> 00:12:39.679 orthopedics? Being underpowered? The source suggests 00:12:39.679 --> 00:12:42.259 it might be, yeah. Estimates suggest maybe up 00:12:42.259 --> 00:12:44.960 to half of orthopedic RCTs could be underpowered. 00:12:45.700 --> 00:12:48.399 Which makes interpreting those negative trials, 00:12:48.779 --> 00:12:51.340 the ones finding no difference, really tricky. 00:12:51.519 --> 00:12:54.120 Because you don't know if it's truly no difference 00:12:54.120 --> 00:12:57.740 or just missed. Exactly. And remember, the power 00:12:57.740 --> 00:13:00.700 calculation itself is based on estimate assumptions 00:13:00.700 --> 00:13:02.779 about how big the effect will be and how much 00:13:02.779 --> 00:13:05.279 variability there is. If those assumptions are 00:13:05.279 --> 00:13:07.580 off, your actual power might be lower than planned. 00:13:07.899 --> 00:13:10.919 So there's always some uncertainty. Do researchers 00:13:10.919 --> 00:13:13.539 build in a buffer? Yes, it's pretty standard 00:13:13.539 --> 00:13:16.120 practice to inflate the calculated sample size 00:13:16.120 --> 00:13:19.440 by, say, 10 or 20 percent just to account for 00:13:19.440 --> 00:13:22.039 people dropping out or maybe not recruiting quite 00:13:22.039 --> 00:13:24.460 as many as hoped. It's a pragmatic way to safeguard 00:13:24.460 --> 00:13:27.600 the power. Makes sense. Alongside sample size, 00:13:28.000 --> 00:13:29.779 selecting the right patients is obviously key. 00:13:30.159 --> 00:13:33.000 And then in trials, randomization comes in. Right. 00:13:33.519 --> 00:13:35.659 Appropriate patient selection defines who the 00:13:35.659 --> 00:13:39.409 results apply to. Randomization in RCTs is the 00:13:39.409 --> 00:13:41.730 best tool we have to minimize selection bias 00:13:41.730 --> 00:13:44.570 and confounding by assigning patients to groups 00:13:44.570 --> 00:13:47.090 purely by chance. You hope to. You hope to make 00:13:47.090 --> 00:13:49.169 the groups as similar as possible at the start 00:13:49.169 --> 00:13:51.870 in terms of both things you know about, like 00:13:51.870 --> 00:13:55.049 age, severity, and things you don't. So any difference 00:13:55.049 --> 00:13:57.230 you see at the end is more likely due to the 00:13:57.230 --> 00:14:00.049 treatment itself. not baseline differences between 00:14:00.049 --> 00:14:02.809 the groups. So randomization tackles bias head 00:14:02.809 --> 00:14:05.129 -on. It's crucial for tackling selection bias 00:14:05.129 --> 00:14:08.149 and confounding in RCTs, yes. Bias can creep 00:14:08.149 --> 00:14:10.350 in anywhere, though design, data collection, 00:14:10.549 --> 00:14:13.210 analysis. Selection bias is a big one, where 00:14:13.210 --> 00:14:14.929 the groups start off different in a systematic 00:14:14.929 --> 00:14:17.529 way. That's why historical controls are generally 00:14:17.529 --> 00:14:20.110 a bad idea comparing today's patients to yesterday's. 00:14:20.269 --> 00:14:22.529 Too many other things have changed. Okay. Now, 00:14:22.570 --> 00:14:24.990 the sources make a point about allocation concealment 00:14:24.990 --> 00:14:27.570 being distinct from randomization. What's that 00:14:27.570 --> 00:14:30.419 about? Ah, yes, a really important distinction. 00:14:31.720 --> 00:14:34.379 Randomization generates the random sequence who 00:14:34.379 --> 00:14:37.440 should go into which group. Allocation concealment 00:14:37.440 --> 00:14:39.940 is the practical step of shielding that sequence 00:14:39.940 --> 00:14:43.059 from the people enrolling participants. It ensures 00:14:43.059 --> 00:14:45.059 they don't know which group the next eligible 00:14:45.059 --> 00:14:47.500 patient will be assigned to before they decide 00:14:47.500 --> 00:14:50.299 to enroll them. Why does that matter if the sequence 00:14:50.299 --> 00:14:52.919 is already random? Because if the recruiter knows 00:14:52.919 --> 00:14:55.429 the next assignment, say the next one gets the 00:14:55.429 --> 00:14:58.269 new drug, they might consciously or subconsciously 00:14:58.269 --> 00:15:00.649 steer certain types of patients towards or away 00:15:00.649 --> 00:15:03.149 from that next slot. Maybe hold back a sicker 00:15:03.149 --> 00:15:05.789 patient if the good treatment is next, or rush 00:15:05.789 --> 00:15:08.470 to enroll someone if the placebo is next. It 00:15:08.470 --> 00:15:10.850 undermines the randomization. So knowing the 00:15:10.850 --> 00:15:14.259 next step can reintroduce bias. Exactly. And 00:15:14.259 --> 00:15:16.440 the source notes that older methods, like sealed 00:15:16.440 --> 00:15:19.120 envelopes, weren't foolproof. They could sometimes 00:15:19.120 --> 00:15:21.600 be predicted or tampered with, and studies using 00:15:21.600 --> 00:15:24.179 them poorly tend to overestimate treatment effects. 00:15:24.559 --> 00:15:26.840 So what's the gold standard for allocation concealment 00:15:26.840 --> 00:15:30.519 now? Centralized remote methods. Using a secure 00:15:30.519 --> 00:15:33.259 telephone service or an online system, managed 00:15:33.259 --> 00:15:36.019 independently from the trial site. The person 00:15:36.019 --> 00:15:38.100 enrolling the patient contacts the central service 00:15:38.100 --> 00:15:41.100 after confirming eligibility and only then are 00:15:41.100 --> 00:15:43.320 they told the group assignment. It completely 00:15:43.320 --> 00:15:45.840 removes the possibility of influencing the allocation. 00:15:46.440 --> 00:15:48.840 That makes total sense. Keeping it separate keeps 00:15:48.840 --> 00:15:51.259 it clean. Another bias -reducing technique is 00:15:51.259 --> 00:15:54.779 blinding or masking. Who gets blinded? Ideally, 00:15:55.059 --> 00:15:57.580 as many people involved as possible. Blinding 00:15:57.580 --> 00:15:59.419 prevents knowledge of the treatment allocation 00:15:59.419 --> 00:16:02.259 from influencing behavior or assessments. So 00:16:02.259 --> 00:16:04.779 you might blind the participants, subjects, the 00:16:04.779 --> 00:16:06.899 clinicians delivering the care, the people assessing 00:16:06.899 --> 00:16:10.039 outcomes, and the data analysts. But in orthopedics, 00:16:10.279 --> 00:16:13.720 blinding the surgeon seems difficult. Often impossible, 00:16:14.000 --> 00:16:16.379 yes. The surgeon usually knows exactly what procedure 00:16:16.379 --> 00:16:18.500 they performed. That's a major challenge in surgical 00:16:18.500 --> 00:16:21.240 trials. So what can you do? You focus on blinding 00:16:21.240 --> 00:16:24.519 everyone else. Patients can sometimes be blinded, 00:16:24.519 --> 00:16:26.860 perhaps with a sham procedure, though that raises 00:16:26.860 --> 00:16:29.480 its own ethical points. Outcome assessors can 00:16:29.480 --> 00:16:31.440 definitely be blinded. Have someone not involved 00:16:31.440 --> 00:16:33.740 in the surgery or care collect the follow -up 00:16:33.740 --> 00:16:37.139 data, especially PROMs. You can mask scars on 00:16:37.139 --> 00:16:39.799 x -rays, for example. And blinding the data analysts. 00:16:40.019 --> 00:16:41.840 You mentioned that earlier. Yes, that's usually 00:16:41.840 --> 00:16:43.860 the easiest part. Just give them the data labeled 00:16:43.860 --> 00:16:47.220 group A and group B. Yet the source... points 00:16:47.220 --> 00:16:49.440 out, it hasn't always been well reported in the 00:16:49.440 --> 00:16:52.200 past. It's a simple step to prevent bias in how 00:16:52.200 --> 00:16:55.100 the data is handled. All these techniques, randomization, 00:16:55.340 --> 00:16:57.980 blinding, sham procedures, link back to an ethical 00:16:57.980 --> 00:17:01.440 principle. Clinical equipoise. What is that exactly? 00:17:01.879 --> 00:17:04.359 Equipoise is the ethical justification for doing 00:17:04.359 --> 00:17:06.720 an RCT in the first place. It means there must 00:17:06.720 --> 00:17:09.319 be genuine uncertainty within the expert medical 00:17:09.319 --> 00:17:11.980 community about which treatment arm in the trial 00:17:11.980 --> 00:17:15.000 is actually better. It's not about the individual 00:17:15.000 --> 00:17:17.660 researcher's belief, but a state of collective 00:17:17.660 --> 00:17:20.059 uncertainty. So if everyone already knows one 00:17:20.059 --> 00:17:22.680 treatment is superior, then it's unethical to 00:17:22.680 --> 00:17:25.500 randomize. You shouldn't knowingly assign patients 00:17:25.500 --> 00:17:27.880 to what is believed to be an inferior treatment. 00:17:28.720 --> 00:17:31.180 Equipoise insurer's randomization only happens 00:17:31.180 --> 00:17:33.259 when there's a real question about the best approach, 00:17:33.819 --> 00:17:36.000 making both options ethically acceptable within 00:17:36.000 --> 00:17:39.079 the trial context. It's particularly relevant 00:17:39.079 --> 00:17:41.839 when considering sham surgery. You need genuine 00:17:41.839 --> 00:17:44.259 uncertainty about the active procedures benefit 00:17:44.259 --> 00:17:46.960 over the sham. Randomize only when you truly 00:17:46.960 --> 00:17:49.539 don't know. Okay, so we've designed the study 00:17:49.539 --> 00:17:51.539 carefully. Now we need to measure the results. 00:17:51.940 --> 00:17:53.900 The science of outcome measurement. Absolutely 00:17:53.900 --> 00:17:56.519 critical. Outcome measurement is how we quantify 00:17:56.519 --> 00:17:58.799 the effect of the treatment. Is the patient better? 00:17:59.019 --> 00:18:01.799 How much better? It's essential for showing value, 00:18:01.880 --> 00:18:04.420 comparing care, seeing if evidence -based practice 00:18:04.420 --> 00:18:07.359 actually works. The WHO defines it as the change 00:18:07.359 --> 00:18:09.619 in health due to an intervention. And how do 00:18:09.619 --> 00:18:11.180 we measure these outcomes? What are the main 00:18:11.180 --> 00:18:14.380 types? Broadly, two categories. Objective or 00:18:14.380 --> 00:18:17.059 observer -reported outcome measures OROMs. These 00:18:17.059 --> 00:18:19.099 are things measured by a clinician or a device. 00:18:19.619 --> 00:18:21.579 Range of motion, strength, maybe using an instrument 00:18:21.579 --> 00:18:24.180 like the KT1000 for the knee laxity, functional 00:18:24.180 --> 00:18:27.000 tests like hop tests. And the second type? Subjective. 00:18:27.150 --> 00:18:30.369 or patient reported outcome measures, PROMs. 00:18:30.490 --> 00:18:32.089 These come directly from the patient, usually 00:18:32.089 --> 00:18:34.789 via questionnaires, about their pain, function, 00:18:34.950 --> 00:18:37.609 quality of life, symptoms, their perspective. 00:18:37.769 --> 00:18:40.109 What are the limitations of the objective OROMs? 00:18:40.210 --> 00:18:42.329 Well, instrumented measures can be precise, but 00:18:42.329 --> 00:18:45.490 also expensive or complex. Functional tests try 00:18:45.490 --> 00:18:48.390 to mimic real life, but can be affected by patient 00:18:48.390 --> 00:18:51.089 effort or technique, sometimes making them less 00:18:51.089 --> 00:18:53.890 reliable. And crucially, an objective measure 00:18:53.890 --> 00:18:55.869 might not capture what the patient actually cares 00:18:55.869 --> 00:18:59.259 about. The KT -1000 example is good. It measures 00:18:59.259 --> 00:19:01.359 forward, backward, and knee laxity well after 00:19:01.359 --> 00:19:04.359 ACL surgery, but not rotational stability, which 00:19:04.359 --> 00:19:06.240 might be the patient's main problem. Which leads 00:19:06.240 --> 00:19:10.079 us to PROMs, patient -reported outcomes. They 00:19:10.079 --> 00:19:12.200 seem increasingly important. Hugely important 00:19:12.200 --> 00:19:14.319 because they capture the patient's experience. 00:19:14.539 --> 00:19:17.059 How do they feel? How well can they function 00:19:17.059 --> 00:19:19.240 in daily life? It's about patient -centered care. 00:19:19.680 --> 00:19:21.359 What does treatment success look like from their 00:19:21.359 --> 00:19:23.470 point of view? Are all PROMs the same? No, there 00:19:23.470 --> 00:19:25.690 are different types. Generic ones like the SF 00:19:25.690 --> 00:19:29.269 -36 or SF -12 measure overall health. They're 00:19:29.269 --> 00:19:31.130 good for comparing across different diseases 00:19:31.130 --> 00:19:33.890 or picking up broad effects, but they might not 00:19:33.890 --> 00:19:36.529 be very sensitive to changes in a specific joint 00:19:36.529 --> 00:19:39.500 problem. So you also have specific ones? Yes, 00:19:39.779 --> 00:19:42.579 condition -specific or region -specific PROMs. 00:19:42.720 --> 00:19:45.200 Things like the COS for knees, the WOOS for shoulders, 00:19:45.619 --> 00:19:47.920 the Harris hip score. These are designed to be 00:19:47.920 --> 00:19:50.799 really sensitive to changes related to that specific 00:19:50.799 --> 00:19:52.799 condition. They're generally more responsive. 00:19:53.059 --> 00:19:55.140 The choice depends on what you're studying. Choosing 00:19:55.140 --> 00:19:57.160 the right measure is fruitful then. What makes 00:19:57.160 --> 00:19:59.700 a PROM or any outcome measure a good measure? 00:19:59.920 --> 00:20:03.319 What properties does it need? Three key psychometric 00:20:03.319 --> 00:20:06.009 properties. First is reliability. Is the measure 00:20:06.009 --> 00:20:08.289 consistent? If you measure the same thing twice, 00:20:08.430 --> 00:20:09.950 do you get the same result? Assuming nothing 00:20:09.950 --> 00:20:13.069 has changed. You need reliability to trust that 00:20:13.069 --> 00:20:15.269 any change you see is real, not just measurement 00:20:15.269 --> 00:20:18.089 noise. Different kinds of reliability. Yes, test 00:20:18.089 --> 00:20:21.210 or retest reliability. Interrater between different 00:20:21.210 --> 00:20:23.809 assessors. Interrater, same assessor. Internal 00:20:23.809 --> 00:20:26.049 consistency. Do items on a scale hang together? 00:20:26.240 --> 00:20:29.420 We quantify it with stats like ICC or Cronbach's 00:20:29.420 --> 00:20:32.259 Alpha. For trials, you often want reliability 00:20:32.259 --> 00:20:35.440 above .9. Reliability is consistency. What's 00:20:35.440 --> 00:20:39.019 next? Validity. Does the measure actually measure 00:20:39.019 --> 00:20:41.299 what it claims to measure? Does your knee score 00:20:41.299 --> 00:20:44.240 truly reflect knee function? There's criterion 00:20:44.240 --> 00:20:47.529 validity. compared to a gold standard, construct 00:20:47.529 --> 00:20:50.569 validity does it behave as theory predicts, discriminant 00:20:50.569 --> 00:20:52.950 validity does it not correlate with unrelated 00:20:52.950 --> 00:20:56.049 things. Reliability, validity, and the third. 00:20:56.369 --> 00:20:58.910 Responsiveness. Can the measure detect change 00:20:58.910 --> 00:21:01.470 when it actually occurs, especially clinically 00:21:01.470 --> 00:21:03.950 meaningful change? A measure could be reliable 00:21:03.950 --> 00:21:06.910 and valid, but if it's too blunt to pick up improvements, 00:21:07.410 --> 00:21:09.779 it's not much use in a trial. Right. Reliability, 00:21:09.839 --> 00:21:11.880 ability, responsiveness. So you've got your scores 00:21:11.880 --> 00:21:13.819 changing. How do you know if the change matters? 00:21:14.500 --> 00:21:16.359 Statistical significance isn't everything, you 00:21:16.359 --> 00:21:18.680 said. Exactly. It's where MCID and MDC come in. 00:21:18.940 --> 00:21:22.799 Hugely important concepts. MCID. MCID is the 00:21:22.799 --> 00:21:24.940 minimally clinically important difference. It's 00:21:24.940 --> 00:21:27.519 the smallest change in score that patients actually 00:21:27.519 --> 00:21:30.779 perceive as beneficial or important. The smallest 00:21:30.779 --> 00:21:32.640 change that makes a real difference to them. 00:21:33.250 --> 00:21:35.890 MDC is the minimum detectable change, the smallest 00:21:35.890 --> 00:21:38.509 change that's statistically likely to be real, 00:21:38.789 --> 00:21:40.869 not just measurement error. So MCID is about 00:21:40.869 --> 00:21:43.450 patient perception, MDC is about measurement 00:21:43.450 --> 00:21:46.430 precision. You've got it. And ideally, for a 00:21:46.430 --> 00:21:48.630 change to be truly meaningful, it needs to be 00:21:48.630 --> 00:21:51.190 statistically significant and exceed the MCID. 00:21:51.519 --> 00:21:53.920 The source gives a great example. A trial finds 00:21:53.920 --> 00:21:56.460 a four -point difference on the IKDC score, P 00:21:56.460 --> 00:21:59.619 .05, statistically significant. But if the MCID 00:21:59.619 --> 00:22:02.660 for the IKDC is known to be, say, 11 .5 points, 00:22:03.000 --> 00:22:04.740 that four -point difference is statistically 00:22:04.740 --> 00:22:06.759 real but probably doesn't feel important to the 00:22:06.759 --> 00:22:08.619 patient. That really clarifies the difference 00:22:08.619 --> 00:22:11.039 between statistical noise and meaningful improvement. 00:22:11.240 --> 00:22:13.539 Is there another patient -focused concept emerging? 00:22:13.960 --> 00:22:16.500 Yes. The patient -acceptable symptom state, or 00:22:16.500 --> 00:22:19.140 PASS. It's usually a single question. Is your 00:22:19.140 --> 00:22:21.250 current state acceptable to you? It's a threshold. 00:22:21.849 --> 00:22:24.329 Are you feeling well or satisfactory, regardless 00:22:24.329 --> 00:22:27.049 of how much you've changed? It complements change 00:22:27.049 --> 00:22:29.750 scores by asking about the absolute state achieved. 00:22:30.329 --> 00:22:32.369 Interesting. A different way of looking at success. 00:22:32.670 --> 00:22:34.750 Okay, so we've collected the outcome data. Now 00:22:34.750 --> 00:22:38.349 we need to analyze it. Statistics. What are the 00:22:38.349 --> 00:22:41.230 basics we need to grasp? First, know your data 00:22:41.230 --> 00:22:45.049 type. Is it continuous, like age or height? Or 00:22:45.049 --> 00:22:47.250 categorical, like yes, no, or mild, moderate, 00:22:47.450 --> 00:22:49.690 severe? That dictates the stats you can use. 00:22:49.880 --> 00:22:52.599 Then you use descriptive statistics to summarize 00:22:52.599 --> 00:22:55.220 your sample frequencies, mean, median, mode for 00:22:55.220 --> 00:22:57.420 the central point, standard deviation, or range 00:22:57.420 --> 00:22:59.119 for the spread. It gives you a picture of who 00:22:59.119 --> 00:23:00.980 was in the study. And then for comparing groups 00:23:00.980 --> 00:23:03.519 or looking for relationships. You choose inferential 00:23:03.519 --> 00:23:05.640 statistical tests based on your question and 00:23:05.640 --> 00:23:07.900 data type, comparing two group means, maybe a 00:23:07.900 --> 00:23:11.259 t -test, three or more groups, ANOVA, looking 00:23:11.259 --> 00:23:13.779 for an association, correlation, predicting an 00:23:13.779 --> 00:23:16.160 outcome, regression. The test must match the 00:23:16.160 --> 00:23:18.700 question. When we get the results, the p -values, 00:23:18.880 --> 00:23:21.440 the confidence intervals, how do we interpret 00:23:21.440 --> 00:23:24.680 them beyond just p .05? Right, the p -value just 00:23:24.680 --> 00:23:26.680 tells you the probability of seeing your result 00:23:26.680 --> 00:23:30.180 if there was truly no effect. Less than .05 usually 00:23:30.180 --> 00:23:33.460 means statistically significant, unlikely due 00:23:33.460 --> 00:23:36.039 to chance alone. But the confidence interval, 00:23:36.220 --> 00:23:39.539 the CI, is often more informative. Why the CI? 00:23:39.740 --> 00:23:41.799 It gives you a range of plausible values for 00:23:41.799 --> 00:23:44.700 the true effect in the population. A 95 % CI 00:23:44.700 --> 00:23:47.519 means we're 95 % confident the true value lies 00:23:47.519 --> 00:23:49.980 within that range. It tells you about the precision. 00:23:50.250 --> 00:23:53.309 Narrow CI means more precise estimate. Wide CI 00:23:53.309 --> 00:23:55.690 means less precise. And crucially, you can compare 00:23:55.690 --> 00:23:58.670 the CI to the MCID. Does the entire range of 00:23:58.670 --> 00:24:01.769 the CI fall below the MCID? Or does it comfortably 00:24:01.769 --> 00:24:04.150 cross it? That gives you a much better sense 00:24:04.150 --> 00:24:06.470 of clinical significance than the p -value alone. 00:24:06.690 --> 00:24:09.250 The CI gives magnitude and precision. Beyond 00:24:09.250 --> 00:24:12.170 the numbers, presenting data visually. Tuft's 00:24:12.170 --> 00:24:14.430 principles. Yes, Edward Tuft's ideas about graphical 00:24:14.430 --> 00:24:16.809 excellence and visual integrity are key. Show 00:24:16.809 --> 00:24:19.200 the data clearly. Avoid clutter, chart junk. 00:24:19.559 --> 00:24:21.319 Make sure the graphics accurately represent the 00:24:21.319 --> 00:24:23.940 numbers without distortion. Maximize the data 00:24:23.940 --> 00:24:26.859 ink ratio, essentially. And the Cleveland and 00:24:26.859 --> 00:24:29.480 McGill hierarchy. What's that telling us? It's 00:24:29.480 --> 00:24:32.220 about how accurately we perceive different visual 00:24:32.220 --> 00:24:34.440 elements. We're best at judging position along 00:24:34.440 --> 00:24:38.220 a common scale, like bar charts. Less good with 00:24:38.220 --> 00:24:41.099 angles or areas, like pie charts. So it guides 00:24:41.099 --> 00:24:43.640 choices. Bar charts are often better than pie 00:24:43.640 --> 00:24:45.839 charts for comparing proportions because we judge 00:24:45.839 --> 00:24:48.819 length more accurately than angle or area. Practical 00:24:48.819 --> 00:24:51.539 tips for figures and papers, then? Make them 00:24:51.539 --> 00:24:54.380 stand alone. The legend should explain everything 00:24:54.380 --> 00:24:56.119 needed to understand the figure without going 00:24:56.119 --> 00:24:59.140 back to the text. Define abbreviations. Keep 00:24:59.140 --> 00:25:01.720 it clean. Focus on the data. Let the substance 00:25:01.720 --> 00:25:04.079 shine through, not fancy graphics. Brilliant. 00:25:04.200 --> 00:25:06.220 Okay, we've looked deep into individual studies. 00:25:06.720 --> 00:25:08.599 But knowledge builds up over time. How do we 00:25:08.599 --> 00:25:11.299 handle multiple studies? Different designs? The 00:25:11.299 --> 00:25:14.000 hierarchy of evidence. Right. The hierarchy concept 00:25:14.000 --> 00:25:16.700 ranks study designs based on how well they protect 00:25:16.700 --> 00:25:19.059 against bias, especially when looking at cause 00:25:19.059 --> 00:25:22.079 and effect. RCTs are generally at the top for 00:25:22.079 --> 00:25:24.299 intervention questions because randomization 00:25:24.299 --> 00:25:28.119 minimizes bias. Below that, you have observational 00:25:28.119 --> 00:25:31.359 studies. Expert consensus often shapes these 00:25:31.359 --> 00:25:33.759 hierarchies and reporting guidelines like consort 00:25:33.759 --> 00:25:36.859 for RCTs. RCTs are the gold standard for interventions 00:25:36.859 --> 00:25:40.490 then? Generally, yes. for asking, does treatment 00:25:40.490 --> 00:25:43.289 X work better than treatment Y? But they have 00:25:43.289 --> 00:25:45.750 drawbacks, as we discussed. They do. Not always 00:25:45.750 --> 00:25:48.650 possible or ethical. Strict criteria can limit 00:25:48.650 --> 00:25:50.950 generalizability. How will the results apply 00:25:50.950 --> 00:25:53.430 to typical patients? And the underpowering issue 00:25:53.430 --> 00:25:55.990 means we have to be careful interpreting no -difference 00:25:55.990 --> 00:25:58.849 findings. So observational studies have a key 00:25:58.849 --> 00:26:01.710 role, too. Cohorts, case controls. Absolutely 00:26:01.710 --> 00:26:04.180 essential. When RCTs aren't feasible, they provide 00:26:04.180 --> 00:26:06.839 vital information, especially on prognosis, risk 00:26:06.839 --> 00:26:09.640 factors, real -world effectiveness. They often 00:26:09.640 --> 00:26:12.180 generate the hypotheses that RCTs then go on 00:26:12.180 --> 00:26:14.279 to test. They complement each other. You mentioned 00:26:14.279 --> 00:26:16.240 case control studies are research in reverse. 00:26:16.519 --> 00:26:19.000 Yes. That's a good way to think about it. You 00:26:19.000 --> 00:26:21.099 start with people who have the outcome, cases, 00:26:21.279 --> 00:26:23.180 and people who don't, controls. And then you 00:26:23.180 --> 00:26:25.200 look backwards in time to see if exposure to 00:26:25.200 --> 00:26:27.440 potential risk factors differed between the groups. 00:26:27.680 --> 00:26:31.119 and case reports or series. Lowest level of evidence, 00:26:31.180 --> 00:26:35.119 but still useful. Level V, yes. Not great for 00:26:35.119 --> 00:26:37.799 proving cause and effect, but invaluable for 00:26:37.799 --> 00:26:40.400 reporting something new. A rare side effect, 00:26:40.920 --> 00:26:43.920 a novel surgical technique, an unusual presentation. 00:26:44.519 --> 00:26:47.440 They alert the community, spark ideas, generate 00:26:47.440 --> 00:26:49.920 hypotheses for better studies later on. Often 00:26:49.920 --> 00:26:51.940 the very first signal of something important. 00:26:52.009 --> 00:26:54.509 So a whole ecosystem of study designs, how do 00:26:54.509 --> 00:26:56.470 we bring all this disparate evidence together? 00:26:56.829 --> 00:26:59.190 Systematic reviews. Exactly. That's the next 00:26:59.190 --> 00:27:01.529 level up, synthesizing primary studies. What's 00:27:01.529 --> 00:27:03.769 the difference between a quick narrative review 00:27:03.769 --> 00:27:06.650 and a proper systematic review? A narrative review 00:27:06.650 --> 00:27:09.150 is more like an expert summary. They pick key 00:27:09.150 --> 00:27:11.730 papers, give their perspective. It's useful, 00:27:12.190 --> 00:27:15.269 but can be subjective. A systematic review is 00:27:15.269 --> 00:27:18.269 a research project in itself. It uses explicit 00:27:18.269 --> 00:27:20.930 predefined methods to find all relevant studies 00:27:20.930 --> 00:27:23.589 on a specific question, assess their quality 00:27:23.589 --> 00:27:26.349 using tools like Cochrane's risk of bias or minors, 00:27:26.589 --> 00:27:29.289 and then synthesize the findings transparently. 00:27:30.009 --> 00:27:32.910 It aims to be objective and reproducible. And 00:27:32.910 --> 00:27:35.630 a meta -analysis fits in where? A meta -analysis 00:27:35.630 --> 00:27:37.789 is the statistical part you can sometimes do 00:27:37.789 --> 00:27:40.359 within a systematic review. If the included studies 00:27:40.359 --> 00:27:42.839 are similar enough in patients, interventions, 00:27:43.160 --> 00:27:45.920 outcomes, you can pool their numerical results 00:27:45.920 --> 00:27:48.440 together to get a single overall estimate of 00:27:48.440 --> 00:27:50.849 the effect. This pooled result is usually more 00:27:50.849 --> 00:27:53.150 precise than any individual study. Often shown 00:27:53.150 --> 00:27:55.410 in those forest plots. Yes, the forest plot is 00:27:55.410 --> 00:27:57.529 the classic way to visualize a meta -analysis. 00:27:57.910 --> 00:28:00.130 Each study gets a line showing its result and 00:28:00.130 --> 00:28:02.150 confidence interval, and a diamond at the bottom 00:28:02.150 --> 00:28:04.289 shows the pooled result. It lets you see the 00:28:04.289 --> 00:28:06.430 individual studies and the overall picture at 00:28:06.430 --> 00:28:08.470 a glance. But what if the studies being pooled 00:28:08.470 --> 00:28:11.490 are really different? Heterogeneity. Yes, that's 00:28:11.490 --> 00:28:14.559 a major issue. Heterogeneity means the studies 00:28:14.559 --> 00:28:16.779 aren't all measuring the same underlying effect. 00:28:16.819 --> 00:28:19.539 It could be due to differences in patients, clinical 00:28:19.539 --> 00:28:23.240 study methods, methodological, or just more variation 00:28:23.240 --> 00:28:25.779 in results than you'd expect by chance. Statistical. 00:28:26.160 --> 00:28:28.539 We measure statistical heterogeneity with things 00:28:28.539 --> 00:28:30.900 like the I squared statistic. And if I squared 00:28:30.900 --> 00:28:33.140 is high? High heterogeneity means you have to 00:28:33.140 --> 00:28:36.240 be very cautious about pooling the results. The 00:28:36.240 --> 00:28:38.200 average might not mean much if the studies are 00:28:38.200 --> 00:28:40.890 too dissimilar. You might explore why they differ 00:28:40.890 --> 00:28:43.789 using subgroup analyses, but the source rightly 00:28:43.789 --> 00:28:45.970 warns against just chucking out studies to make 00:28:45.970 --> 00:28:48.130 heterogeneity go down that can introduce its 00:28:48.130 --> 00:28:51.470 own bias. It requires careful judgment. Synthesis 00:28:51.470 --> 00:28:54.509 is complex, too. Okay, shifting gears slightly 00:28:54.509 --> 00:28:56.930 to practicalities. Multi -center trials needed 00:28:56.930 --> 00:28:59.349 for big numbers, but sound like a nightmare to 00:28:59.349 --> 00:29:02.109 run. They are incredibly challenging, yes. The 00:29:02.109 --> 00:29:04.950 benefits are clear bigger samples, more generalizable 00:29:04.950 --> 00:29:08.380 results, but the logistics. getting ethics approval 00:29:08.380 --> 00:29:10.759 from every single site. The source mentions the 00:29:10.759 --> 00:29:13.140 Jupiter trial experience, how delays can mount 00:29:13.140 --> 00:29:15.720 up. Central IRBs might help streamline that. 00:29:15.799 --> 00:29:18.200 What else makes them tough? Standardizing everything. 00:29:18.400 --> 00:29:20.740 making sure every site does the procedure the 00:29:20.740 --> 00:29:23.259 same way, measures outcomes identically, fills 00:29:23.259 --> 00:29:26.079 in the forms correctly. Communication is absolutely 00:29:26.079 --> 00:29:28.500 key, keeping everyone connected and motivated 00:29:28.500 --> 00:29:30.779 across different hospitals or even countries. 00:29:31.480 --> 00:29:34.359 Lessons from trials like PIVL highlight the need 00:29:34.359 --> 00:29:37.019 for constant contact, maybe face -to -face meetings. 00:29:37.700 --> 00:29:40.220 And agreeing on authorship early on is crucial 00:29:40.220 --> 00:29:42.680 when you have so many contributors. A massive 00:29:42.680 --> 00:29:44.920 coordination effort. What about registries? How 00:29:44.920 --> 00:29:47.720 do they fit in? Registries collect data on routine 00:29:47.720 --> 00:29:50.460 practice. They track outcomes for large numbers 00:29:50.460 --> 00:29:52.460 of patients undergoing certain procedures like 00:29:52.460 --> 00:29:54.960 joint replacements or ACL reconstructions in 00:29:54.960 --> 00:29:57.240 the real world. They're brilliant for long -term 00:29:57.240 --> 00:29:59.680 follow -up, spotting rare problems, and seeing 00:29:59.680 --> 00:30:01.720 how things work outside the controlled setting 00:30:01.720 --> 00:30:04.680 of an RCT. The Norwegian knee ligament registry 00:30:04.680 --> 00:30:06.759 is a great example. Do registries have weaknesses? 00:30:07.700 --> 00:30:10.079 Data quality can vary. Completeness might depend 00:30:10.079 --> 00:30:12.599 on whether reporting is mandatory. Local resources. 00:30:12.940 --> 00:30:15.599 And comparing outcomes between centers using 00:30:15.599 --> 00:30:18.319 registry data needs careful statistical adjustment 00:30:18.319 --> 00:30:19.940 because the patients might be very different. 00:30:20.319 --> 00:30:23.200 Case mix. Finally, economic studies. Why the 00:30:23.200 --> 00:30:25.900 focus on cost effectiveness now? Because healthcare 00:30:25.900 --> 00:30:28.599 resources aren't infinite. Costs are rising. 00:30:29.000 --> 00:30:31.380 And there's a push toward value -based care, 00:30:31.680 --> 00:30:33.519 getting the best outcomes for the money spent. 00:30:33.900 --> 00:30:36.420 So we need to evaluate not just if a treatment 00:30:36.420 --> 00:30:39.200 works, but if it's worth the cost compared to 00:30:39.200 --> 00:30:42.869 alternatives. The value equation. outcome divided 00:30:42.869 --> 00:30:44.829 by cost. How do you approach that? You need to 00:30:44.829 --> 00:30:46.910 measure both the costs, surgery, hospital stay, 00:30:47.230 --> 00:30:50.230 rehab, complications, and the outcomes. Clinical 00:30:50.230 --> 00:30:52.569 scores, quality of life. But the crucial thing 00:30:52.569 --> 00:30:55.130 the source highlights is defining the perspective. 00:30:55.470 --> 00:30:57.769 Are you looking at costs to the patient, the 00:30:57.769 --> 00:31:01.490 hospital, the NHS or insurer? Society as a whole. 00:31:01.670 --> 00:31:04.069 Why does the perspective matter so much? Because 00:31:04.069 --> 00:31:06.210 it changes what costs and benefits you include. 00:31:06.849 --> 00:31:08.930 A societal perspective includes lost productivity, 00:31:09.150 --> 00:31:11.490 which a hospital perspective wouldn't, the best 00:31:11.490 --> 00:31:13.349 or most cost -effective option can look very 00:31:13.349 --> 00:31:15.450 different depending on whose costs and benefits 00:31:15.450 --> 00:31:18.069 you count. The example in the source about different 00:31:18.069 --> 00:31:19.809 bone -cutting tools makes this really clear. 00:31:19.890 --> 00:31:22.349 The cheapest tool isn't necessarily the fastest 00:31:22.349 --> 00:31:25.329 or most accurate. You need clear criteria. A 00:31:25.329 --> 00:31:27.589 really important nuance we've covered so much. 00:31:28.130 --> 00:31:31.910 From EBM origins, ethics, study design, outcomes, 00:31:32.349 --> 00:31:35.569 stats, synthesis, practicalities, how do researchers 00:31:35.569 --> 00:31:38.990 share all this work? Dissemination is key. usually 00:31:38.990 --> 00:31:41.269 through peer -reviewed publications, scientific 00:31:41.269 --> 00:31:43.789 papers, and presentations at conferences like 00:31:43.789 --> 00:31:46.029 posters or talks. Tips for writing a good paper. 00:31:46.349 --> 00:31:49.349 Follow the standard IMRAD structure. Intro, methods, 00:31:49.549 --> 00:31:52.690 results, discussion. Avoid common pitfalls like 00:31:52.690 --> 00:31:55.069 trying to include absolutely everything, repeating 00:31:55.069 --> 00:31:58.349 yourself, or overstating your conclusions. Don't 00:31:58.349 --> 00:32:00.269 claim causation if your design doesn't support 00:32:00.269 --> 00:32:03.579 it. Outline first, work with co -authors, manage 00:32:03.579 --> 00:32:06.240 references properly, create clear figures and 00:32:06.240 --> 00:32:09.279 legends, and discuss authorship early. Abstracts 00:32:09.279 --> 00:32:11.420 and posters. Abstracts are the concise, pitch 00:32:11.420 --> 00:32:14.420 -grab attention, convey the key message. Posters 00:32:14.420 --> 00:32:16.700 need to be visually clear, tell the story graphically, 00:32:17.059 --> 00:32:18.819 but have enough detail for someone to understand 00:32:18.819 --> 00:32:20.720 the basics, even if you're not standing right 00:32:20.720 --> 00:32:23.099 there. Good formatting, clear fonts, and always 00:32:23.099 --> 00:32:25.299 disclose conflicts of interest. And presentations. 00:32:25.480 --> 00:32:28.119 Tell a story. The discussion is vital. Interpret 00:32:28.119 --> 00:32:30.559 your findings, place them in context, acknowledge 00:32:30.559 --> 00:32:34.140 limitations, be ready for the Q &A. Lastly, scientific 00:32:34.140 --> 00:32:36.339 meetings. More than just listening to talks. 00:32:36.579 --> 00:32:39.180 Oh, much more. They're vital for learning, seeing 00:32:39.180 --> 00:32:41.859 unpublished work, refining your thinking, prepare 00:32:41.859 --> 00:32:45.359 beforehand, check the program, abstracts. But 00:32:45.359 --> 00:32:48.779 the networking, that's equally crucial. Networking. 00:32:49.019 --> 00:32:51.400 Yes. connecting with peers, senior figures, potential 00:32:51.400 --> 00:32:54.220 mentors. Especially for junior people, the source 00:32:54.220 --> 00:32:57.099 emphasizes this. You build collaborations, get 00:32:57.099 --> 00:33:00.240 advice, find opportunities. That face -to -face 00:33:00.240 --> 00:33:02.519 interaction is still incredibly valuable, perhaps 00:33:02.519 --> 00:33:05.240 even more so in our digital age. Watching senior 00:33:05.240 --> 00:33:07.259 colleagues interact is also a great way to learn 00:33:07.259 --> 00:33:09.339 professional norms. This has been incredibly 00:33:09.339 --> 00:33:12.440 comprehensive, a real journey. It has, from how 00:33:12.440 --> 00:33:15.279 do we know, through ethics. design like randomization 00:33:15.279 --> 00:33:17.740 and blinding, measuring patient outcomes with 00:33:17.740 --> 00:33:20.839 PROMs and MCID, interpreting data, understanding 00:33:20.839 --> 00:33:22.839 different study types and synthesis, ready for 00:33:22.839 --> 00:33:25.299 the practicalities of multi -center trials, registries, 00:33:25.400 --> 00:33:27.779 and sharing the work. And for you listening as 00:33:27.779 --> 00:33:30.420 professionals in this area, grasping these principles 00:33:30.420 --> 00:33:33.440 isn't just academic. It's absolutely vital for 00:33:33.440 --> 00:33:35.559 critically evaluating the research that informs 00:33:35.559 --> 00:33:38.000 your practice every single day. It helps you 00:33:38.000 --> 00:33:40.740 make better, more informed decisions. Knowing 00:33:40.740 --> 00:33:43.160 how the evidence is generated allows you to be 00:33:43.160 --> 00:33:45.640 a much more critical and insightful consumer. 00:33:46.359 --> 00:33:48.339 You can ask the right questions about a study's 00:33:48.339 --> 00:33:50.880 methods and understand the true clinical weight 00:33:50.880 --> 00:33:53.319 of its findings. If this deep dive has been helpful, 00:33:53.519 --> 00:33:55.259 we'd really appreciate it if you could rate the 00:33:55.259 --> 00:33:57.099 show and perhaps share it with colleagues, maybe 00:33:57.099 --> 00:33:59.619 on LinkedIn or X, and to leave you with something 00:33:59.619 --> 00:34:02.900 to think about. We've seen the power of RCTs 00:34:02.900 --> 00:34:05.500 for controlling bias and the richness of registry 00:34:05.500 --> 00:34:08.980 data for real -world insights. How might linking 00:34:08.980 --> 00:34:11.579 these two worlds, combining large -scale registry 00:34:11.579 --> 00:34:14.559 data with embedded randomized comparisons, perhaps 00:34:14.559 --> 00:34:16.699 change how we understand treatment effectiveness 00:34:16.699 --> 00:34:19.420 across the diverse reality of orthopedic practice 00:34:19.420 --> 00:34:20.019 in the future?