Episode 121: Therac-25

Well There's Your Problem | Episode 121: Therac-25 .ass file - download the video off of YouTube and play with your media player of choice .ytt file - feed the subtitle file to your YouTube(!) in this convoluted manner Corrections (please!) to haitch dubya ei zed you you dubya at the google mail service JUSTIN: Okay. Hello, and welcome to Well There's Your Problem. It's a podcast about engineering disasters, with slides. I'm Justin Roczniak, I'm the person who's talking right now. My pronouns are “he” and “him”. All right, go. ALICE: I am Alice Caldwell-Kelly, I'm the person who's talking now. My pronouns are “she” and “her”. Yay, Liam! LIAM: Yay, Liam! Here's the thing. ALICE: Very formal “yay, Liam”. LIAM: My pronouns are “he”/”him”, I'm Liam Anderson, and one of my friends just spelled karaoke “K-E-R-O-K-E”. ALICE: Keroke? Isn't that like a potato dumpling? LIAM: Keroke! JUSTIN: Isn't that the guy who pioneered assisted suicide, right? LIAM: Yeah! ALICE: [laughs] Dr. Keroke, yeah! LIAM: I just got a text that said, “going to keroke late?” “Yeah, 10 to 1030.” “Damn, y'all young or something”, “y'all” spelled wrong. ALICE: I love your friends. LIAM: They're great, (?) ALICE: Speaking of doctors, we have a guest. Guest, please introduce yourself. TOM: Hello, my name is Tom Bowers, my pronouns are “he”/”him”, for the sake of argument. LIAM: Tom, why the fuck are you here? JUSTIN: [laughs] ALICE: Just like, we'll revisit this at a future time, yeah, maybe. TOM: I've done the visiting this(?), and we'll kinda come back around to it at some point, I'm sure. But... for now, “he”/”him” is fine. So. Yes. JUSTIN: So we seem to have... LIAM: Are we not gonna explain why Dr. Tom is here? JUSTIN: Oh, we do need to explain why Dr. Tom is here. LIAM: Dr. Tom, why are you here? TOM: Uh, I'm here because we are gonna talk about a medical disaster. LIAM: Yeah, you fuckers have been asking for this forever. JUSTIN: Yes. TOM: So, what we have on the screen here, this is a, well, this is in fact the Therac-6 linear accelerator. ALICE: This is very sort of Star Trek original series... It's The Device, they're gonna put you in The Device. You got too sick, and now they're gonna put you in a device, something which I would not enjoy. JUSTIN: I'm, I'm like halfway between “this is the medical device” and “this is what's inside the”, you know, the vaporization chambers in that episode where they're having a virtual war? LIAM: You guys remember the beginning of Quake 4? It's kinda like that. [laughter] TOM: Oh gosh. TOM: ...This has the exciting tagline, this is a “reliable, high output linear accelerator featuring "extra large fully flattened treatment fields”. ALICE: Looks like a big sewing machine, and it looks like... they just put you under it and then, I don't know, it uses a... big red flashlight and it just knocks all the illness out of you, is maybe the idea. TOM: I love that sort of even in sort of the, you know, what, early 80s, late 70s here, we've already agreed that sort of the laser, sort of laser eyes, laser sort of, that's good for marketing, that sells a product. Also coming in a color that just, you know, medical devices do not have this range of colors anymore. The 1970s, 1980s, sort of the end of an era for interestingly colored devices. ALICE: Oh my God. Contrast trim, in like an olive. It's wild. TOM: It's got some chromed bits, and now they all just come in beige. JUSTIN: Yeah, it's all beige or white. I think they'd look less intimidating if they had more design on them, y'know? LIAM: Oh, that sterile doctors office white? Yeah. ALICE: I think you could go the opposite direction, right? I think if you played it up a bit– If you put me in a jet black MRI, and you were like, yeah, this is the goth MRI, I would be a lot less scared. JUSTIN: It has like a pointed arch in front, as the entrance, you know. ALICE: [laughs] LIAM: It plays Depeche Mode the entire time you're in there. ALICE: Yeah, every time they put you near it, the radiologist is like, “prepare the sacrifice.” And just black metal starts playing? JUSTIN: [laughs] Yeah, alright, you know? This is sort of commensurate with what I'm about to be feeling. JUSTIN: Got like gargoyles on it and shit, you know? ALICE: [laughs] LIAM: Oh yees. [laughs] ALICE: I love the Warhammer 40k setting. JUSTIN: When it quenches, if it quenches the gas, it goes out a gargoyle on the outside of the building. LIAM: Yeah. TOM: Incredible. LIAM: You guys seen that scene from Hunchback of Notre Dame? Yeah! ALICE: [laughs] TOM: As someone who has at least some of a medical sort of radiation science degree, then, yeah, that's the kind of thing that a lot of... radiotherapists and radiation scientists would be into. ALICE: Yeah. Getting into this, because you really want some hooded robes, you know? Like driving music, all this stuff. TOM: I mean, it's sort of... Well, I do have a degree, but I have a third class degree in this, and research for this basically involved relearning my entire, sort of, the X-ray part of my course, which, had I done that at the time, I would have done significantly better. ALICE: I'm the only one here who is like, I will never go back to my degree for a subject, because we've made everyone else here except me revisit their degree, and I'm not going back to law school for this. I'm not sure how I would, but I'm not going back to any of my old books for this. But so, this thing is the precursor of what we're gonna talk about, which is... TOM: This is the one you can still find pictures of because it worked. ALICE: By popular demand, we're talking about this thing's less successful descendant, Therac-25. But first... JUSTIN: We have to do The God Damn News. ♪[news jingle]♪ JUSTIN: Come on. There we go. ALICE: So, we managed to have a helicopter crash in Australia, and two helicopters smacked directly into each other at, I guess, like a heliport? Two sightseeing helicopters, one of them climbing, the other one landing. And there's actually footage from on board, the one descending, which shows the thing that you don't want, which is a passenger grabbing the pilot's arm and pointing. So as to convey, there is a helicopter that you are about to fly into. JUSTIN: Oops. LIAM: Right. Oops. ALICE: So the helicopter taking off shears the front off of the one landing, including the entire windscreen. LIAM: Nope. ALICE: Like, the first sort of like, say, half a foot of the helicopter is just gone, just missing. That still manages to land safely, which, credit to that pilot who now does not have a windshield and has just had a very very close shave from some rotor blades or fuselage or whatever, managing to land that safely. The other one, as we see here, has lost the entire tail. That did not make a safe landing, it makes a very deadly landing and kills everyone on board. JUSTIN: That's gone poorly. TOM: The last I heard, there was one survivor from the crash, or has that now changed? ALICE: I believe that has now changed, I believe everyone on board... TOM: Oh no. Oh god. ALICE: has died, but, um. Yeah, so these are all, like, tourists, because this is a sightseeing helicopter, and, yeah, helicopters are obscenely dangerous machines, I'm terrified of them. LIAM: 50,000 moving parts, looking for a place to crash. ALICE: Uh-huh, uh-huh. As I put on the chyron here, the only thing more dangerous than a helicopter is two helicopters, in fact it's twice as dangerous. JUSTIN: Yes. TOM: Bloody hell. LIAM: You remember when we did that Kobe Bryant thread, and people got really mad at us? ALICE: Oh, yeah. Mm-hm. JUSTIN: [laughs] ALICE: Yeah. ALICE: I mean, I've been on helicopter tours before. I went on one over the Grand Canyon as a child, and the whole time I was like, well, if I die, at least it'll be an interesting fall. It'll be very scenic, y'know? Like, uh, twenty seconds of just, like, going into a big hole in the ground, like Joshua Graham, and I'm like, yeah, this is fine, to me. Um. I've done(?). JUSTIN: I can't imagine going on one of those tours just because it'd be so loud. You would get to the point where it's loud enough that, y'know, A. I can't hear anything, and then my vision starts going blurry... LIAM: You can't hear anything anyway. JUSTIN: This is true, but when there's enough noise, LIAM: [laughs] JUSTIN: when there's enough noise, my other senses start going. ALICE: It's actually not that, it's not that bad, like, once they put... the headset, the ear defenders and radio thing on you, it's really not so bad. It feels... I've been in a much louder situation–I've been on louder trains, I'm pretty certain. Although that may just be the poor quality of British rails. But, so, the Australians are now investigating how this happened, but as yet, we don't know, other than the fact that helicopters is bad when you crash them into each other. JUSTIN: Yeah, don't do that. ALICE: Don't do that. JUSTIN: Well There’s Your Problem top tip: don't get in a helicopter if you can avoid it. ALICE: Yeah, and if you do, try to like... I mean, this is a valuable example of, like, passenger-assisted see and avoid, right? Not successfully, but, y'know, if you as a passenger can see that the aircraft you're on is about to crash, it's probably a good thing to give the pilot a heads up. And so, y'know, very helpful. JUSTIN: So wait, was the helicopter that crashed and killed everyone the one where the pilot got the heads up, or was it...? ALICE: No, no, that was the other one, I believe. JUSTIN: Ah, okay. ALICE: So I believe they gave him the heads up right before the helicopter that killed everyone, like, crashed into it. JUSTIN: Ah, I see. Well. TOM: Yeah, that's... Yeah, I saw some of the stuff about the people involved, and that was pretty tragic. JUSTIN: In other... news. ♪[news jingle]♪ ALICE: There's Weather! Weather is occurring. Um. JUSTIN: Weather did happen. ALICE: This is, uh, like the second instance of Weather in the United States in the same week, I believe, because previously the East Coast got it, and now the West Coast had a bomb cyclone, which... JUSTIN: We didn't get shit here in Philly, we got like a little flurry for a couple hours, [pop] it was over. [laughs] ALICE: Meanwhile, Buffalo is like, trapped in a big block of ice like the end of The Shining. JUSTIN: Oh yeah, for the second time in the past couple months. ALICE: You know, it's almost as if there's some kind of climatic instability occurring here. It's like there's Weather, but more often. JUSTIN: Could be. ALICE: I don't know if anyone's noticed this or thought about this, maybe formed a sort of international working group about this. TOM: It's fascinating, isn't it, how the sort of... I mean, I don't know about anyone else, I'm somewhat getting quite accustomed to these once in a generation events... I mean, I don't know if it's just me, or if I'm getting old or whatever, but... every year, LIAM: ...just grinds you down, right? It's just... TOM: the once in a generation weather event seems to come around earlier and earlier. ALICE: [laughs] Yeah, it seems like I'm living through one of those one-in-a-hundred-year things ALICE: pretty often these days. LIAM: Every year. That's what (?) said, is that climate change is basically just watching the planet be destroyed through cellphone cameras until you're holding the camera. ALICE: Well— JUSTIN: I just want some snow to happen here, and what happens instead is it's consistently like 54 degrees in January. I mean, my God, this is lame. ALICE: But so, the worst part is, if you believe in sort of like, radical optimism, or like climate optimism or whatever, then that requires you to go, this is all locked in, it's gonna be like this forever now, this is normal. Therefore it is time for us to exert sort of maximum effort to make sure it is only this bad plus as worse as it gets, like, until it bottoms out, y'know? And I put down "There's Weather" because, like, all of this shit's already locked in, this is happening, whether we make it worse than this or not. And eventually we're gonna have to do like COVID, we're gonna have to, like, stop considering this as news, in order to cope with it psychologically, otherwise it would be all the news that there was. If you remember that New York Times front page that was like, here are the New Yorkers who have died of COVID, in like really small print. And then they just, the number of deaths from COVID in New York just like, quintupled after that, and they just kind of... well, okay? JUSTIN: Yeah, whatcha gonna do? ALICE: Well, it's Weather. My joke for this is that God sent a storm to kill Jeremy Renner, but instead he got Ken Block, which, RIP. JUSTIN: True. ALICE: Very sad. Snowmobiles, also, very dangerous machines. You wanna be careful with those too. JUSTIN: I was about to say, people seem to... get themselves killed on snowmobiles, like, constantly. TOM: I was looking up– JUSTIN: Mostly doing tricks with them. TOM: I was looking up to see which was... invented first, the snowmobile or the jet ski. But it's actually the snowmobile. So the jet ski was invented as a safer alternative to the snowmobile. ALICE: Wow. I mean, those things and ATVs as well, like, just generally, any sort of equipment that you put a powerful engine in, then you don't really require that much training to let people use. It's just fine. It's normal. And, y'know, this is just gonna continue, and it's up to you whether... you live in a climate change zone that gets this kind of fucked up weather, or Philadelphia where it's just like, oh, it's just hot in the winter. JUSTIN: It's just hot and... LIAM: Not (?) 100+ as the summers now, yeah. JUSTIN: ...miserable drizzle, just constantly. LIAM: Oh God, the cold rain(?) JUSTIN: Cloud cover for months on end, everything's gray. [laughs] LIAM: Like we live in Nova Scotia, yeah. JUSTIN: Yeah. [laughs] LIAM: I do want to say that, obviously, it's hard to, just on a personal level... the grind of this is so fucking exhausting, of watching... everything just go up in smoke and people not giving a shit about it. I know... I'm not even saying that to be depressing, I'm just saying that to be like, you're not the only person feeling that way. LIAM: And I suggest, personally, what you do is, and you will have to bleep this, but I'm doing it for comedic effect, go to the ExxonMobil CEO's house, or houses, or whichever, and Saudi Aramco... ALICE: With a katana. With a katana. LIAM: No, no, we're not doing katanas, that's inefficient. We are doing big halbards, baby. JUSTIN: Any one of those cobbled together guns, you know. LIAM: Halbards. ALICE: [laughs] LIAM: And what you do is, you usher his [bleep] into the basement, and you say, “stand right here, we're gonna [bleep], and then you just [bleep] like it's a 90s RPG. JUSTIN: This is gonna be a long bleep. LIAM: Yeah, good. TOM: [laughs] Oh no. LIAM: And then you [bleep] the [bleep] [bleep] on the front lawn, call the news and say, hey, [bleep] the CEO of ExxonMo[bleep], and... you might want to come quick, his entire family's [bleep]. And then you repeat that for Royal Shell and whoever the hell else, and then you take the entire Congress and [bleep]. Once again, you will need a halbard for this... ALICE: Uh, Liam, Liam... LIAM: I'm setting a record for longest bleep, Alice. How can I help you? I've had a very frustrating week. ALICE: ...I don't think so. I think... the path to progress is like moderate reforms within the law. I think if we, like, sort of petition the government peacefully for a redress of our grievances... JUSTIN: We could change the system from within. ALICE: That's right! LIAM: Yes. LIAM: That's why I've become a police officer. ALICE: [laughs] Yeah, that's right. JUSTIN: Yeah, Alice, JUSTIN: have you considered getting a petroleum engineering degree? ALICE: Ah! LIAM: Alice! [laughter] ALICE: It's strange, it's almost as if people think that a climate change– LIAM: Well, I have an oil field for you to blow up! ALICE: It's curious, people almost seem to believe that an oil company executive could not be a socialist revolutionary. Um. JUSTIN: [laughs] Um. In a third piece of news... ALICE: Oh, fuck, where'd I put the news button? ♪[news jingle]♪ ALICE: There we go. This guy is in hell. We put this guy in a living nightmare. TOM: [laughs] JUSTIN: Yes. LIAM: Yes. JUSTIN: This has been– LIAM: Of his own creation. ...This is, like, watching the Republic fall apart before our eyes (?) these dweebs sucks. It'd be cool if we were doing it, but like, watching this guy with that fucking shit-eating face, knowing he's about to just scoop more hot shit into his mouth, TOM: [laughs] LIAM: just a goddamn delight. TOM: Oh dear. ALICE: This is Kevin McCarthy, he is not the House Majority Leader. He wants to be. He wants to be the leader of the new Republican majority in the House of Representatives. However, they have to vote for him to do that, and he has been experiencing a sort of permanent nightmare where all of your colleagues get together in a room to vote about how cool they think you are, and, like, you can't get a majority of them. Like, uh, because... JUSTIN: Yes. JUSTIN: They think the other guys are cool. ALICE: Yeah, the House Freedom caucus thinks that Trump is cooler, and, uh, or thinks that, like, this guy's woke, and he's gonna do gender to them. And so they've been refusing to vote for him. Most noticeable was Matt Gaetz, or [different pronunciation] Gaetz, the sort of like, weird wrong aspect ratio looking ass, uh, yeah. LIAM: “Pedophile”. You can say it. ALICE: You can say that, yeah. JUSTIN: Are we able to say that? I'm not sure. LIAM: Oh, yeah. Yeah, he is a pedophile. Yeah, he is a pedophile. ALICE: [laughs] LIAM: Also, we– JUSTIN: We have facts to back that up. LIAM: Yes, we have... criminal investigations to back that up. JUSTIN: Oh, I see, okay. ALICE: I thought you were just doing more beeps, but... LIAM: Nope! Nope, dude's a straight up pedophile. ALICE: Wow. So, yeah, the– JUSTIN: You need a good pedophile to fight the bad pedophile. ALICE: [laughs] LIAM: Really good pedophiles, yeah. TOM: [laughs] LIAM: I don't really know where to go with that one. ALICE: You need a dead pedophile, a good pedophile... LIAM: Yeah, right there on the wiki page! LIAM: “Federal Investigations into Sex Trafficking”. We can call him a pedophile. ALICE: Ah, I see. Wow. Okay. ...in any case, Matt Gaetz and his caucus of freaks and weirdos are, like, refusing to vote for him, and all the Republicans can think to do is just run it again. Run it again. Run it again. We've had, what, twelve votes now that he's lost? JUSTIN: We're going into fourteen. Yeah. TOM: We’re up to fourteen. ALICE: Jesus TOM: Fourteenth time’s the charm. LIAM: He's married to– Matt Gaetz is married to the dweeb that invented Oculus, that guy's sister. ALICE: Palmer Luckey. I hate that I know that name. LIAM: I didn't want to say his name. ALICE: I hate that I know it. TOM: ...most recently in the news for... inventing an Oculus headset that will, um... LIAM: Kill its user. Yeah. ALICE: Kill you, yeah. TOM: Yeah, kill you... if you die in the game, you die in real life. ALICE: Yeah, that guy. JUSTIN: Well, what I think is interesting about this whole vote situation is that, um, y'know, once, like, they elect a Speaker, the entire purpose of the Republican majority will be to prevent any legislation from moving forward. So they don't actually have to elect a Speaker, because that has the same effect. LIAM: Right. ALICE: Yes. JUSTIN: But I think, y'know, maybe even more effective to further their agenda of blocking everything is just to continue this through another two years. ALICE: [laughs] JUSTIN: Because, y'know, Republicans in the House would not be able to do very much other than block legislation handed down by the Senate, regardless of... the outcome here. So I think, y'know, this is gonna probably go on for a while, I think. ALICE: It also adds a nice sort of fun thing, which is that Hakeem Jeffries can't be House Minority Leader yet. [laughs] Which is really funny, because he shouldn't be! He's terrible, and therefore... great. Perfect. Y'know? Anarchist–the anarcho-Republican party. JUSTIN: Yes. TOM: I'm sort of enjoying that we've had, y'know, over the summer, our own little contest of... just backstabbing and chaos with the Tory party leadership election. It's nice to go across the pond, sort of just, y'know, once again, America just does things bigger. ALICE: Oh yeah, absolutely. TOM: Yeah. It's lovely to see someone else's completely impenetrable, internecine process for selecting things. Do they have to, um... how do they vote? Is it sort of computerized, or do they have to stand up and walk through a lobby? You have to stand up while a guy calls your name and say “aye” or “nay”. This is why Krysten Sinema did the sort of, like, sassy thumbs down thing. TOM: [groans] JUSTIN: I think it depends on what the vote is. They do have buttons at their desk, I believe, but some votes are, I don't know what they call it, like a roll call vote or something, where, yeah, you have to physically go up. Um. ALICE: This is great. TOM: ...Democrats are drinking in the chamber or something. LIAM: They're drinking popcorn, which I... Apparently a couple of them were drinking, I would be drinking. I would be drinking, I would start dipping again... ALICE: [laughs] LIAM: and just spitting at a cup, being like, y'all ain't voted yet, huh? [laughs] ALICE: You wouldn't even have to use a cup, because I think Congress still has, like, congressionally mandated Civil War era spittoons. LIAM: If you were paying attention at our live show, you would know that they do. JUSTIN: I'm gonna say, uh... LIAM: Alice. ALICE: ...I was 70% aware of that live show. It was mostly chants, and to be honest– LIAM: Yeah, you fuckers love a chant. ALICE: Yeah. JUSTIN: Apologize for not getting recording then. Um, but anyway. LIAM: We did try, just so you know. JUSTIN: ...the net effect of this is, we're just not gonna get any post offices renamed for another two years, y'know? [laughs] ALICE: Yeah, we can just designate– [crosstalk] ALICE: ...3rd August in October to be, uh... y'know. LIAM: You saying “the 3rd August in October”, buddy? ALICE: Yeah. JUSTIN: 3rd August in October is... National Time Day. [laughter] JUSTIN: Calendar Literacy Day. [laughter] ALICE: Oh, I'm in such a frame of mind for this, yeah. TOM: This is the sort of thousandth and something day of March 2020, so I mean, at this point (?). ALICE: [laughs] Yeah. It's been 2016 for, oh, upwards of a million years. JUSTIN: ♪It's been... one week.♪ ALICE: [laugh] LIAM: You know that song came out like 23 years ago this week? TOM: Oh no. JUSTIN: Wow, wow. ALICE: Can we (?) to make the 3rd August in October National Well There’s Your Problem Day? I think he'd do it. LIAM: ...we could try. JUSTIN: Yeah, we could ask him. I mean... LIAM: Yeah, that's–oh, that song turns 25 this year, sorry. JUSTIN: Wow. TOM: Ugh. ALICE: God, we're old. JUSTIN: Anyway, it'll... probably be another five years before they elect a Speaker, and I for one am here for it, I think it's very funny. ALICE: Absolutely. JUSTIN: ...just wanna watch those numbers go up, y'know? ALICE: [laughs] JUSTIN: Anyway, that was The God Damn News. ♪[news jingle]♪ ALICE: Alright, what are we doing here? –pretty colours! TOM: Pretty colours, yes! This is the only bit of my day job I get to talk about here. So this is, uh... this is, uh... nature's... Oh, what? Okay. We'll, we'll... yeah. This is, um, this is nature's own engineering disaster. This is cancer. ALICE: Oh, this is the thing that I think I have every time I get a slight twinge anywhere. TOM: So this is cancer. On the, specifically, on the left, this is what's called a ductal carcinoma of the breast, and on the right is a Hodgkin's lymphoma. TOM: So that's, yeah, so... I'm not gonna go into the, uh, too much into the specifics of those cancers exactly, because... each of them is at least a sort of three hour presentation. So, basically cancer occurs when there are... ALICE: “But we have to ask ourselves, what is cancer?” TOM: What is cancer? So, cancer basically occurs... JUSTIN: It's a tropic. TOM: [laughs] LIAM: Goddammit, dude. [laughs] TOM: It arises when there are... failures in processes during cell replication. And this basically occurs in one of two ways: You either have mutations that are present from birth, and those are called germline mutations. So... LIAM: Like the BRCA gene, right? TOM: Like BRCA, like Lynch syndrome in bowel cancer, endometrial cancer. And then somatic mutations, which are ones that occur throughout life, you acquire them as you– the older you get, the more things can go wrong. So I've written, I've put, recapping a bit of basic biology here, but given that the standard for basic biology has been set so low... ALICE: Our viewers, our listeners, the standard beneath even that. I mean... LIAM: Yes. TOM: I mean, uh... JUSTIN: The less I know about how my body works, the happier I am. [laughs] LIAM: Oh, yeah. TOM: This is where it gets kind of terrifying, because this is when you find out that a lot of stuff is kind of... it's all under the control of, sort of, proteins. And there's no... ALICE: Oh, fuck, those guys, Jesus Christ. JUSTIN: Oh my God. TOM: It's not a really conscious process, and even less than that, it's literally just due to, like, these things have to be folded in a certain way. ALICE: You think that you're, like, a guy? Wrong. Incorrect. You are, like, 15 trillion little miniature guys, and they're all doing their own thing. JUSTIN: Your body is firmware. ALICE: Yeah. [laughs] JUSTIN: [laughs] ALICE: You are at best a sort of unruly democracy of, like, a lot of other shit going on. Yeah. LIAM: Gut flora, yes. LIAM: And everyone sort of has to march in the same general direction. If we think of the body as the Stanford marching band... LIAM: [laughs] ALICE: You think of the body as the House of Representatives, right. It has to elect a speaker. LIAM: Your dumb brain is the speaker. JUSTIN: You think of the body as one of several armies marching to the Crusades. ALICE: [laughs] ALICE: Yeah, this is why I'm (?) LIAM: Don't look, uh, wherever it was. Constantinople. ALICE: Constantinople, yeah. JUSTIN: Constantinople, yeah. [laughs] TOM: So, uh, next slide, please. ALICE: Hey, I recognize this guy too. TOM: Hey, it's this guy. LIAM: It's the dumb parts that make us up. TOM: Everyone's favourite character from Jurassic Park. ALICE: [laughs] TOM: So, this is a, uh, DNA double helix, and it's... so DNA is sorted in chromosomes, in humans there are 46 paired, and then sex chromosomes in varying combinations. ALICE: Weird, yeah. TOM: It's composed of four bases, and these, um, when they're assembled along the structure of the DNA, they're in triplets, and those triplets tell the molecules that are involved in replication, those are called codons. They basically give a little bit of code as to where to put a protein, or to start or stop reading the sequence. So when cells make proteins, the DNA for the relevant gene sort of unzips itself, and it's read by something called RNA polymerase. This transfers information into messenger RNA, so this is mRNA, this is messenger RNA, so the, uh, the mRNA vaccines for COVID are sort of a bit of a shortcut, so when they're injected they go into cells, and then the ribosomes in those cells reads the mRNA, and... it's got all the different building blocks, so think about it like Lego bricks, for instance, it tells you which Lego brick to stack on top of the other one. ALICE: Your body's dirtbag programmers. Your body's IT department. JUSTIN: It sounds a lot like a tape drive. TOM: I mean, yeah! Not dissimilar. It basically just gives it the set of coding instructions– LIAM: Which are okay. Bulk storage is okay. You don't need instant access to every file you have. Tape drives make sense, shut up. JUSTIN: [laughs] ALICE: It's okay for this–imagine solid state DNA. It'd be a hell of a lot more complicated. LIAM: Oh, I'd be running so fucking fast. JUSTIN: Probably be more reliable, though. TOM: That's probably a good anime plotline, let's face it. ALICE: Yeh? TOM: Eh, maybe. But, um, so, you encode it, and then it gets to the stop what's called a start code and a stop code, and it gets to the stop, and the protein sequence falls away, and because of... the ways in which the molecules in the proteins interact within the cell environment, it folds itself. So people talk about... ALICE: This is one of the things that, like, this is why I'm the opposite of you, Justin, I like knowing more stuff about my body on the basis that, this is incredible, we made a computer out of meat. ...It’s the reprogrammable, self-programmable meat computer, and it's made of meat. LIAM: Meat and electricity. ALICE: Yeah, and you plug enough electricity into this meat and it fucking paints the ceiling of the Sistine Chapel. Like, I do not like– how you get from– LIAM: Why I didn't study biology, buddy. ALICE: Because it rules! Because it rules, and you're afraid of your own greatness. JUSTIN: ...I think they should– LIAM: No, Alice, if anyone has never been afraid of its own greatness, it's me. I am the most arrogant piece of shit you can imagine. [laughs] ALICE: [laughs] JUSTIN: They should throw a solid state drive in there somewhere, so you can... you can flash the firmware every once in a while, and then reverse the aging process. LIAM: In your case, Rocz, you just throw it in your armpit because I've seen you try to install stuff. [laughs] ALICE: It's like, these guys from Silicon Valley are here to offer you 50 trillion dollars to try and do this. Because they're all afraid of their own aging. LIAM: Can I have some of that 50 trillion, please? ALICE: Yeah, sure. You just gotta tell them that there's a smoothie you can drink somewhere that's gonna make all of this work faultlessly forever. And entropy is not gonna apply to it. JUSTIN: I was thinking you could do it... you could do it with some kind of picture. LIAM: Yes, yes. But you gotta hide it in your house somewhere, is the thing. JUSTIN: Yeah, it's unfortunate. I recommend not looking at it. LIAM: I don't understand– –I've never read the book–why can't he just throw a sheet on the thing? JUSTIN: That's what I always thought, yeah. ALICE: He does! For like, years! LIAM: Oh, okay, well then, that’s–I’ve never read the book! JUSTIN: He should've put it in a drywall. ALICE and JUSTIN: [laugh] LIAM: Yeah, you gotta think smarter. JUSTIN: Yeah. Make it difficult to access! ALICE: So, so, how are we trying to salvage this, right? JUSTIN: [laughs] TOM: DNA... So, DNA makes the proteins, or DNA is the code for the proteins, basically. And so when cells divide, there's potential for errors to occur. So next slide, please. We'll talk... Next! ALICE: Proteins folding, like, in themselves, when that goes wrong in a different way, that's some shit I'm really scared of. That's prions. That's... TOM: That is prions, yes. That is... LIAM: Oh no. JUSTIN: I also wanted to say, the other thing Dorian Gray should have done, is just waited longer to look at the picture, and they'd be like, wow, I'm a cool skeleton. [laughter] ALICE: Other than misfolding prions, what you can do also that fucks up is you can just go, oh, just do these proteins. Fold them right, but in the wrong order, in the wrong place. You know, fold too many of them, too few of them. LIAM: I have a question. I never really understood what Folding@home was. I ran it for a while, is that related? TOM: So, TOM: yes, that is related. That is to do with the way that, as far as we're looking into the, the way that proteins folded in mutations in different kinds of cancer. So not talking about prions here, but talking about the ways that when a cancer gene encodes a protein, and we'll kind of come onto this, it's often a case whereby half of one protein stuck to another protein, and then it folds weird. And so the Folding@home thing was basically... running multiple sequences of these things, and in a simulated environment, seeing how they fold. With an idea of... being able to target specific therapies at different parts of that coded protein. LIAM: Interesting. Okay, thank you. TOM: That's all right. ALICE: And you can unfold these proteins again if you wanted, most notably by cooking, which is what makes your meat safe to eat. TOM: Yeah. LIAM: Is that a fucking joke? I am so out of my goddamn mind. TOM: No, no, no! That’s genuinely what happens! LIAM: [laughs] ALICE: ...Meat's got proteins in it! Like, that's what makes meat. LIAM: Yeah, okay. ALICE: And so if you wanna, like, one of the reasons why eating raw meat is very bad for you, right, is because it's got a lot of, like, proteins which are folded into the state that they're supposed to be in for being in muscle tissue. Which is, y'know. LIAM: And you have to break that down, right. ALICE: Yeah, you denature them by applying some combination of heat, (?), sauce, things of this nature, until you obtain a tasty snack that you can just digest. Just vampire the nutrients right out of–Biology's amazing! I, yeah. Sorry. [laughs] TOM: And also, this takes place in a very tight range of conditions as well. So one of the things I get really irritated at this time of year is the whole sort of thing about, you should detox, or you should eat this sort of thing because it's an alkaline diet or something. It's like, you cannot– LIAM: Juice cleanse, juice cleanse, juice cleanse. ALICE: [laughs] Yeah. TOM: Sort of saying, you can defeat cancer if you make your body into an alkaline environment... You literally cannot make your body into an alkaline environment, and if you do, you will die. Because there is, y'know, several million years of evolution has gone into maintaining a very specific and very tight range of pH within the body. So the sort of normal range is 7.35 to 7.45, if you go much outside that, it will kill you. And that's because the proteins don't fold properly, and the enzymes that are required to sort of break things down or go through other processes in the body just don't work. JUSTIN: You also have the risk, if you had an alkaline environment in your body and you touched a normal person, you'd turn into a sort of school science fair volcano. ALICE and TOM: [laugh] JUSTIN: [laughs] TOM: Just, sort of the full Mentos and Coke, y'know, it does happen sometimes, we don't know. ALICE: [laughs] TOM: That's never happened as far as I'm aware. ALICE: The medical industry is covering up spontaneous combustion by Diet Coke and Mentos. JUSTIN and LIAM: [laugh] TOM: So the cell goes through phases, and what it does is it has a rest phase which is called G0, and it's not undergoing any sort of division at that point, and then it moves into a growth one phase when there's a need for it, and that's usually, depending on where you are in the body, different cells replicate at different rates, some of them just don't replicate at all. So things like... mostly sort of neurons in your brain, the cells in your ear that have the... the cilia that control the reception hearing, and the cells in the retina in the back of the eye, they don't divide at all. Once you've got them, they've got them and they're fixed and that's it. ALICE: Those Silicon Valley guys are outside with another 50 trillion dollars with a project to try and get those to start growing again. TOM: Yeah, I mean, there was an interesting thing about retinoplasty recently that was absolutely horrific, but that's another topic. So if there's like a tissue injury, something like that, your skin cells, your soft tissue cells underneath will start to replicate, and those are the controlled chemicals called cyclins. So if you, and then we've sort of, lots of those released, if there's a, you know, you cut your hand, there'll be cyclins released locally, and then globally to start sort of cells there replicating both sort of tissue cells to heal up the injury, and also things like immune cells to go in there and sort of clean everything up. So then you get the synthesis phase, which is then progresses onto the next stage of replication, which is the growth two phase, which is where the cell just sort of bolts itself up to prepare for division. So it starts making copies of... so within the synthesis phase, it makes a copy of all of the DNA in the cell to make a sort of additional copy of each chromosome. And then in the growth two phase, it sort of, is making more organelles. So the organelles are like the little things like mitochondria– JUSTIN: Mitochondria, the powerhouse of the cell. ALICE: That’s right! TOM: Mitochondria, the powerhouse of the cell! TOM: It makes more of those. So that basically, you know, you have enough to sort of power the two cells. And at that point, the cell basically has to show it's working. And I've said sort of submitted code for review. Cause I think this was, that was a phrase that was in my head at the time. So it has to sort of print it, print out all of its code. And just– LIAM: ...does code reviews, of course. Yes. TOM: For the code review. ALICE: ...Elon Musk's cell that comes around. TOM: Yeah, well– ...not even a cell. This is under the control of another protein. This is called a protein known as PUMA, which is... coded from a gene called p53. So PUMA is the p53 unregulated mediator of apoptosis. And it checks all of the DNA at that point and make sure that basically it's done a good job of writing, of coding everything on those chromosomes, exactly the same. And if there are errors, then basically it's told to go back and code this again. Or if it's unfixably bad, the cell is basically then told, don't do anything. LIAM: You're going to cell jail. Yes. ALICE: Just die. TOM: You are going to–yeah, just die. You are going to, yeah. It's called apoptosis, programmed cell death. JUSTIN: So this is the commissar protein. ALICE: [laughs] LIAM: Yes. JUSTIN: [laughs] LIAM: ...commissar performance review. [laughs] ALICE: I think of it, right, as the Tumblr anon ask box of the body, right? In that LIAM: What the fuck? ALICE: It looks at what you've done, ALICE: and if it likes it, it's like, yeah, great. And if it doesn't, it tells you to kill yourself. LIAM and JUSTIN: [laugh] LIAM: How's it going over there on Tumblr, Alice? [laughter] TOM: So, yeah, this then ends, so if everything is fine, and it's definitely fine, and the cell does what's called the M phase, the mitosis phase, so it splits apart, and into the new, the chromosomes line themselves up across the middle of the cell, and everything pulls apart, and the nuclei separate, and then the cytoplasm, which is the sort of jelly bit of the cell, separates, and then you have two cells. And that's mitosis. ALICE: It's free real estate. Or free cells. Sorry? ALICE: It's free real estate. You have one cell, just tears itself apart, two cells. Easy. This can never go wrong. Never, no. Apart from when it does. I think we are to the next slide, please. ALICE: Oh fuck, the next slide, where it goes wrong. TOM: Yeah, the next slide. So this, for reference, this is a serous carcinoma. And these occur either from the ovary or the lining of the peritoneum, which is the lining of the abdominal cavity, or... from the fallopian tube, so those are the places it tends to arise from. And this actually is due to, when you have a, one of the things that can go wrong, is that you have a mutation in p53. So if you have a germline mutation in p53, which is what BRCA gives you, then this is one of the things that can go wrong. So– ALICE: Ah, my code-checking protein doesn't check code right. And it just... LIAM: Ah, Elon Musk, you're here again! ALICE: Yeah! TOM: [laughs] JUSTIN: My commissar showed too much mercy. [laughter] TOM: So eventually what happens is enough errors start to slip through, enough somatic errors slip through, that you get uncontrolled division of cells. So that's one of the—so you get failure in checking. The other thing is when the... chromosomes are pulling apart, then a bit of one chromosome can get attached to another, and this is called a translocation. Now in some cases, this is what's called a balanced translocation. So the bit of the one chromosome that's stuck to the other chromosome, and then the bit of the other one is stuck to the other one, it's what's called balanced translocation, so it still functions the same. But sometimes then it happens in a region where you get, you then get misfolding of proteins or the protein kind of cuts off early or something like that. And then you get a sort of misfolded protein. In some cases what will happen is that this will get recognised and the cell will be told to destroy itself again, and in some cases it won't. And then this is where the checking sort of thing. But p53 is the commissar protein, effectively. It is responsible for— so errors in p53 are responsible for about 50% of all cancers, based on the sort of current understanding we have of genomics. So. ALICE: What we need is a second commissar protein to check the work of the first commissar. LIAM: Oh no, a second commissar protein has hit the towers! TOM: [laughs] ALICE: Do you see why Stalinism takes off? We need NKVD blocking detachments of proteins. JUSTIN: I think the NKVD also had commissars, and uh... ALICE: Yeah! Commissars all the way down, y'know? JUSTIN: I was about to say. So you really need one sort of, uh... LIAM: Circle the 7th Army at Stalingrad. JUSTIN: ...biological Stalin. ALICE: Just so long as it's not at the level where I have individually approved cell division requests. Like I open my inbox and I'm like, I have like 16 million unopened things per day, where I'm like, yes, yes, no, yes, uh... LIAM: Alice, let me ask you a question. JUSTIN: The development of the body is a lot like the development of the Soviet Union. [laughter] LIAM: How many unopened emails do you have right now? ALICE: Currently? Oh God. Uh, let's see. Opening Gmail... 193. LIAM: 44,244. ALICE: Jesus. I'll be honest, you are not checking your own cell division, Liam. You're just not doing it. You need a protein to do that LIAM: I'm an anarchist, therefore... JUSTIN: I got 61,000. ALICE: "In matters of boots, defer to the bootmaker. "In matters of cell division, refer to... PUMA." JUSTIN: Yes. TOM: I've got a very satisfying 123. If nothing else... Nice little sequence. That's fine. I don't think any of them are of particular relevance, so, anyway. ALICE: So this goes wrong, you get cancer. Cancer is... any number of different things, which is why saying “cure cancer” is weird, right? Like, “cure virus”. TOM: Yeah. So I mean, there are different factors that are involved. Some germline mutations are, to some extent, they're less of a... in terms of the somatic mutations, those are the kind of things where people say, oh, you know, making lifestyle changes and all that kind of stuff, you theoretically are reducing your exposure to certain carcinogens, which reduces the risk of picking up somatic mutations as you go along. ALICE: But we can absolutely give somebody cancer. We have that technology. LIAM: Oh, we're gonna find out how, even! TOM: Yes. So in terms of what we can do to treat cancer, the body itself does recognise malignant cancer cells, and will try to destroy them using the immune system, so this is using mostly the T lymphocytes, which are the things that are involved here, the cytotoxic T cells, as the cell acquires mutations, it expresses... ALICE: Next slide, Justin. TOM: Yeah, here's the next slide. It expresses, sort of, starts to stop expressing the self-proteins on its surface, which let the body know that this cell has gone wrong. And cells that have gone wrong, they will initially try to display those proteins so that they get killed off by the immune system. But as they acquire mutations, every time this mutant cell divides, it picks up more mutations as it goes along, and every time it does that, it... requires more and more resistance to the body's own defenses, and it starts outstripping the immune system's ability to... ALICE: Looking on itself, and is like, I'm no longer human. Um. And at that point, it's sort of hoisted its own flag, and it's getting along quite nicely, you know? It's growing itself, it's reproducing, it's doing its thing. But unfortunately, because doctors hate fun, and God forbid women be allowed to do anything, merely because this will kill a person, they're like, no, you're not allowed to do that, we have to stop you from doing this. TOM: So, we do have treatment, so that, broadly, the sort of things that we have to do, there are sort of non-chemotherapy medications. So a lot of cancers are things that are arising in endocrine organs, so organs that produce hormones. And they arise from cells that are under the control of hormones. And so we can do things that block or downregulate hormones, those things like, in breast cancer, a lot of things are oestrogen sensitive, or in prostate cancer, testosterone sensitive. So you can use sort of... ALICE: This is why, whenever you choose to alter your testosterone or oestrogen levels, they will give you a little handout that tells you, amongst other things, this will increase your risk of certain kinds of cancer. Um... TOM: Exactly. So you change... ALICE: ...Don’t give a fuck. Grown the tits. Don't care. But, y'know. JUSTIN: [laughs] ALICE: It was nice of you to warn me, I'm gonna disregard that warning, but, um, thank you. Y'know? TOM: But it will also, while it may increase certain types of cancer risk, it'll actually decrease others. ALICE: Yeah. Not a lot of trans women with prostate cancer. Some, but not many. TOM: No. And it is the same medication, and this is of course, on a slight tangent, why you get sort of certain individuals who will talk about, well, these are cancer drugs. These are hard, like, so many people die after using these drugs. Well, yes, because the majority of people... LIAM: They have cancer. JUSTIN: Have cancer, yeah. [laughs] TOM: But, to a certain extent, this is actually very, for sort of hormone– cancers under hormone control, very successful. So for people with lower grade prostate cancer, who, y'know, they'll have monitoring and stuff, and be on testosterone blockers, and they'll do well for several years before they start to run into problems on just suppressing testosterone. ALICE: Not only is your cancer very well controlled, but you're also achieving femboy aesthetics, which is, y'know, classic (?), as far as I understand. TOM: [laughs] JUSTIN: “I'm sorry, your only treatment is forced feminization.” ALICE: [laughs] LIAM: “Oh no, doctor, don't tell me that!” TOM: And then, the other thing that's sort of, I mean, I'm gonna change the order around I wrote these in, because I think the other thing that's pretty easy to understand is surgery. So, y'know, we cut the cancer out, and hope that we've sort of done that with enough of a margin of normal tissue around it, and before it's spread anywhere else. So I didn't say up top, but I just said my job, I am a histopathologist, I'm a trainee in histopathology, and most of my job is looking at lab specimens, and deciding on what sort of cancer someone has, and grading it, and staging it, and that helps then people at the team meetings decide what's the best form of treatment. And then also, looking at specimens, whether someone has had their cancer excised, and saying, yeah, there's sort of a margin of 1, 2, 3, 4, 5–I mean, the margins are tiny, we're talking about here, I mean, what is considered acceptable margins for some cancers is 1mm of normal tissue around it. But if it's a clear margin, it's a clear margin. ALICE: It's sort of like, you're applying a paradigmatic change to the cancer. The cancer's just replicating itself, having fun, and you're like, I am an organism that exists on a macro scale rather than a micro scale, therefore I can cut you off and throw you into the bin. And you can just exist there and die. TOM: I mean, into, well, into paraffin block slides, in my case, or the bin eventually. The bin in a controlled manner, according to the... ALICE: [laughs] ALICE: No, not just into an open mesh wastebasket... [laughs] JUSTIN: It's just bins all around the hospital full of masses of bleeding flesh. ALICE: [laughs] LIAM: Cut that shit out, baby! ALICE: In the operating theatre, the surgeon's setting up a little basketball hoop and doing a three-pointer with an excised tumour into the bin. LIAM: “LeBron!” [laughs] TOM: I mean, I am actually an ex-surgeon, so I switched career paths, but I can't really speak for that, what has gone on in operating theatres that I've seen. ALICE: [laughs] TOM: One of the things I have to sort of say, appearing on this, is you don't have to, any comments made are of course in line with GMC guidelines, and all disposed (?) tissue does take place in accordance with the Human Tissue Act. ALICE: I know, which sucks, by the way, because it means that if you get a genital reconstructive surgery, or whatever they're calling it this week, they won't let you keep your balls in a jar! They won't let you turn them into a, like, a little Newton's cradle or anything! TOM: No, you have to send them to me, so I look at them. [laughter] And I go... So, you know, I take a section of them and look under the microscope and go, well, you know, at least these were fine, and you did not have cancer at the time that they were removed. LIAM: Beautiful. ALICE: You're also going to have... like a desk toy of these, you know? TOM: The good news is, is congratulations on your affirmative surgery, and the better news is you didn't have cancer when you did! ALICE: [laughs] So we've got drugs, we've got surgery... TOM: Yeah, and then we've got another category of drugs. Now we get into the... This is the sort of section that is effectively... This is the “things we found out through war crimes”... section of cancer treatments. ALICE: Yes! LIAM: Oh boy. TOM: So, chemotherapy agents. Most chemotherapy agents are essentially... All drugs are poison, but chemotherapy is kind of poison poison. It's literal... it's really bad poison. And... a major class of these are derived from a chemical called mustine, which some people may have heard of. ALICE: Yes, yes, because FDR, in a sort of strange move that doesn't get talked about much, moved a shitload of chemical weapons to the European Theater of Operations, and particularly the Italian front, just in case, when they were pushed a bit too hard, the Nazis decided to do chemical weapons and the US had to retaliate, and this led to the, like, I may have this entirely backwards, but the bombing of ships in Bari harbour in Italy, which were filled with mustard gas, which is a vesicant, it blisters, it causes blisters. And this sort of dispersed over the town, killed a bunch of people, would be an episode in itself. However, one of those long cross-section studies figured out that, hey, in Bari, not a lot of people got a lot of cancers for a long time, because they had been exposed to mustard gas, or some component of it. And from there, it was sort of derived that, hey, maybe we could use something like this, maybe we can just gas people, and they won't get, you know, their cancer will go into remission or something. TOM: So... it works by... destroying fast replicating cells. That's what it acts on. It sort of... ALICE: That's why a blister agent affects the skin and the inner lining of the lungs and stuff. TOM: Yeah, skin, lining of the lungs, mucous membranes, anywhere where there is... So, skin has a rapid turnover, mucous membranes everywhere have a rapid turnover. And cancer cells have a rapid turnover, because they're not constrained by the body's own sort of checking mechanisms. They've gone beyond that. They just replicate. And it's why there are sort of such dramatic sort of side effects, those sort of hair loss, immunocompromised, infertility, you know, people have sort of vomiting, they can have sort of bleeding from the gut, because all of these cell processes are being disrupted. And it's almost kind of a bit of a race to see which is going to do, first of all cause significant harm to the patient in terms of, you know, the immunocompromised and things like that. It's why people are at such risk of sort of getting other things like, you know, pneumonias or whatever, versus some... depleting the cancer cells. And it's often used as a sort of an adjunct. So a lot of things like a lot of the breast cancer things I see whereby people have a rou— have some rounds of chemotherapy to shrink the tumor down. So you may have something that's sort of five, six centimeters, and it's sort of shrunk down to a centimeter or so, which means that you can do a much more limited excision of it. But it is horrible stuff. And yeah, so cyclophosphamide is a very common agent, and that is the kind of current treatment which is derived from mustine, which is the sort of mustard gas. And then the final one is radiotherapy. So radiotherapy, thank you to the sort of very early pioneers of radiation science, we know that– ALICE: All the guys who are, like, sleeping on big lumps of uranium, LIAM: Yes. ALICE: like, chewing on big lumps of uranium, LIAM: Thanks, Madame Curie. Yes. ALICE: all of these things, we know that radioactivity is a fucking great way of killing cells. Really good at it. JUSTIN: It works pretty good, yeah. LIAM: Proton therapy–You just stand here, and I'm gonna fire this particle accelerator at you, and you won't have cancer anymore. ALICE: We'll get there. TOM: Yeah, we will get there. Because, yeah. But, so, sort of 19th century, late 19th century even, there had been sort of use of therapeutic X-rays and other sort of radiation for cancer treatment. ALICE: ...get that an X-ray is ionizing radiation? It's just like a form of it that's weird. TOM: Yeah. But basically, it's very similar to sort of chemotherapy in that it affects most cells that are replicating quickly. And the quicker that they're replicating, the more they'll be disrupted by the radiation. So I think we're onto the next slide, where we're gonna talk about– This is also how cancer treatment causes huge headaches for structural engineers. [laughs] ALICE: [laughs] LIAM: Line everything in lead! Simply no worry about it. JUSTIN: You gotta design a room with several feet of concrete or lead. [laughs] TOM: Yeah. LIAM: Don't let your kids lick it. TOM: So this is a sort of, I'm gonna sort of summarize here, we're talking about sort of dose effects. And so basically measuring ionizing radiation that was sort of first used was the Röntgen, which measures air ionization. But it doesn't mention, it doesn't sort of really account for absorption of radiation. And it was superseded by the rad. You know, for sort of the Fallout fans, this is the one that they'll be familiar with. And this has sort of been around since the sort of early 50s, and it refers to the amount of energy it absorbs. And it's equivalent to 100 nanojoules per gram of tissue. Now this has been superseded in 1975 by the gray, which is the SI standard unit, which is equivalent to one joule per kilogram of matter, or 100 rads. But this is an SI unit, so the United States still uses rads, because– ALICE: Of course. TOM: Of course. LIAM: Fuck you, that’s why. ALICE: ...Metric system, yeah. TOM: So the other thing to bear in mind is also the radiation dose is not just an expression of the absorbed dose, it's how much (?) which is... the biological equivalent dose, which is mentioned in another unit called Sievert. With one Sievert equivalent to one joule per kilogram, one gray of human tissue, with different weighting factors being applied on the tissue being exposed. ALICE: I'm wearing my, like, you know, cool shirt–I'm wearing my cool sleeveless lead t-shirt, right? My equivalent dose is gonna be much lower under my lead t-shirt than it is gonna be on my arm, say. TOM: Yes. So you can– what it basically means is that you can sort of have a sort of big whole body radiation dose, but if it's sort of constrained to a– but if you're constraining that dose to a small area, then you can use higher doses. I mean, we'll talk about that more, I think, on the next slide. But the–sorry. JUSTIN: So we have, like, some coefficient here that is– corresponds to different types of tissue? TOM: Yes. JUSTIN: Is this what I'm s–okay, yeah. TOM: So, different tissue absorb at different rates. Which is why we'll get onto the design of some of the machines used to treat this in certain ways. I think we're onto the next slide. So this is sort of whole body radiation dose effects. ALICE: People will still be like, um, why does my, you know, dentist step out of the room when they do the X-rays? What the fuck? And post about that. TOM: Yeah. TOM: It's–and that's mainly because the radiation doses are stochastic. So the more doses you're exposed to over a length of time, the more they build up within the body. So you– ALICE: I had this explained to me, in sort of idiot form one time, very effectively, I should say, which is, if you go to a bar and you have a drink, you're gonna be fine. If the bartender has a drink with you, he is also gonna be fine. If the bartender does that with everyone who comes into the bar, they're all gonna be fine and he's gonna be dead. And I found that a very, like, [laughs] effective illustration of why... LIAM: Yeah. ALICE: your dentist steps out of the room. TOM: If that's the sort of illustration I'd had sort of during my medical imaging science course, then I might have done sort of better. But also, that would have required me to have not had undiagnosed ADHD and not got hopelessly addicted to EVE Online. ALICE: Killer. On both counts! JUSTIN: [laughs] JUSTIN: Oh, yep. That's a way to... That's a way to waste a lot of time. TOM: Yes. LIAM: I remember Everquest. TOM: Yeah, I mean, I could have been doing sort of boring work spreadsheets, but I was doing exciting combat spreadsheets instead. JUSTIN: I've always been afraid of touching that game, because I know I'd never be seen again. [laughs] LIAM: No, you wouldn't. No, you wouldn't. LIAM: I've seen you play goddamn, uh... what's it called? What do you like? JUSTIN: Factorio. LIAM: Yeah, there we go. JUSTIN: [laughs] TOM: Yeah, no, I don't wanna touch that, because it feels like it's a... I'm sort of coming up to my sort of... I'm sort of in the process of doing my final exams, for however many times it is when this goes out, but, um, yeah. I just have to sort of steer clear of that stuff for the time being. ALICE: Medicine in general just seems like one of those things where it's like, you're really good at homework, so we gave you some more homework. You like doing exams? We got 50 more of them, and they're all incredibly important. TOM: Yeah, and not to mention that, it's like, you have to pay for them yourself as well. ALICE: Of course. It wouldn't be a profession in the United Kingdom if it wasn't incredibly unsustainable to, like, do. TOM: No, it’s... ALICE: Despite its value. TOM: It's the thing that sort of people, sort of on a slight tangent, always seem to think, um, oh, your hospital pays for that, don't you see? But it's like, no, no, no, they don't. That is, you know, they expect me to pass them, but in no way do they, uh, wanna pay for me to do them. Anyway. LIAM: Awesome. ALICE: A lot like law, except you actually contribute something useful to society. [laughs] What you're paying for was to sit exams, what I was paying for was to, like, go to dinners. [laughs] TOM: I mean, there are... it's not to say that there aren't dinners. Uh, it's just a... ALICE: [laughs] TOM: You know. LIAM: I’ll be right back, keep going. TOM: Okay. So. I mean, in terms of exposure risk, then, there's a measure of both the effective dose, the time exposed for, and the number of doses, and we have pretty good data on what constitutes harmful doses of radiation. ALICE: You're able to survive a surprising amount sometimes, too. TOM: You are. And what tends to be distances to sources, and that's mainly due to, this is the second bit of war crimes, the development of nuclear weapons technology. ALICE: Oh, those! TOM: Ooh! So when the sort of prototype nuclear weapons... JUSTIN: If Liam were here right now, he'd be arguing. JUSTIN and ALICE: [laugh] ALICE: Yeah, but we lined up a bunch of conscripts, a bunch of countries did, including the UK, and just, like, went, okay, close your eyes, the nuclear weapon... we didn't even tell them the nuclear weapon was gonna go off, they were like, hey, we're gonna deploy you to the Caribbean for a bit, and you're gonna stand on this island, don't worry about it, and the next thing you know, [laughs] a nuclear weapon goes off, like, 25 miles away. [laughs] TOM: But in terms of the up-close doses, that was still stuff like the, um, the Demon Core experiments. ALICE: Ooh, yeah. TOM: Yeah. ALICE: How could you give your own scientist Slotin the Spicy Core? JUSTIN: What I'm getting is that the... the deadly joke from Monty Python, where they just tested it on a guy, was real. Just with radiation. [laughs] ALICE: [laughs] Pretty much! Yeah! ALICE: This has been a personal interest to me, but there's a whole load of shit, particularly in the early Cold War of the 50s and 60s, where, like, if you were in, say, the US Army in 1960, there's a decent chance your immediate commanding officer was a six foot tall sculpture made of uranium as part of a long-term study by the CIA to see how many different kinds of cancer they could give a person at once. JUSTIN: [laughs] TOM: So– JUSTIN: They did that in Iraq, too, but with burn pits. ALICE: Yeah. TOM: The, uh, the other sort of... the Demon Core is sort of a fascinating bit of thing, because it's basically two big hemispheres of plutonium, with, uh, sort of, on a shaft, and then shims inserted at sort of, every sort of few millimetre intervals. And they sort of remove these, one at a time, to see at which point this thing starts to become, starts to sort of react with each other. ALICE: Go critical, yeah. TOM: Go critical. And, there are two incidents that occurred, and one of which was when everything sort of fell out by accident. And then the second one of which was that someone had kind of invented their party trick, of they'd get everyone in– ALICE: There was (?) Slotin, yes! TOM: Yes! Get everyone in to demonstrate, you know, and they'd go, you know, we're moving the core, this is where they, you know, we've got the Geiger counter here, and this is where it starts to go clicky. And they'd just flick the shims out, one by one, with a screwdriver. Until he unfortunately flicked all of the shims out with the screwdriver, and the top half went on the bottom half. And... he sort of dived over it to sort of... ALICE: It was nice of him, considering. TOM: Yeah, nice of him considering. Managed to sort of prise these two things apart, and then sort of having calculated where everyone else was in the room relating to him, and there was a, as I understand it, there was a sort of army colonel or something standing behind him. They had an idea of how much radiation his entire body had absorbed, and then how much the guy behind him had absorbed, and then I think there were two other people in the room, and how far they were from the source when the thing went critical. And that sort of gave some useful information. And then the other useful information was how long it sort of took the, uh... It took him to... die after sort of... ALICE: A huge radiation sickness. TOM: Yes. This is in the... ALICE: One of the worst ways to go. TOM: So this is down the bottom of our chart here. So, yeah, this is sort of, y'know, this is the absolute sort of, this is a lethal dose of radiation over a very short period of time. ALICE: There's no pleasant radiation sickness, obviously, but you read accounts of some of these things, and like, this will come up when we do Chernobyl as well, but some of the time it's like, “I had what I thought was a cold,” right? And then, the far end of it is this thing, where it's like— okay, the Chernobyl miniseries was not accurate, like, in terms of you don't turn into a big piece of melty beef jerky, right, but it's... like, uniquely unpleasant. And... it also takes a long time, as well, like, as we see here, measured in days. And it's... TOM: Yeah. It's extremely grim. So that's what whole body dose is. So when radiotherapy is formed, we're only talking about a very small volume of tissue being targeted, so while higher intensity amounts of radiation are used, under normal circumstances only being applied to a very small amount of total body tissue, and the adverse effects are limited, it's not to sort of say it's completely eradicated, because it's not, and in cancers that treat with radiotherapy, patients will often have temporary effects in terms of immunosuppression, and can get secondary cancers as a result. So the most... common of which is something called angiosarcoma, which is a tumor which is sort of a blood vessel, effectively, it's a blood vessel forming tumor, and it's not nice, and it does not do well. ALICE: One thing that's sort of notable... JUSTIN: Invented by big radiation, so you'll do more radiation treatments after getting radiation treatments. ALICE: [laughs] ALICE: What's interesting to me is the extent to which there is sort of, like, very obvious feedback from radiation for a lot of the time. Like I mentioned people not knowing that they had radiation sickness before, we've talked about that in Goiânia and stuff, but like, a lot of the time, radiation workers and stuff will know very clearly that they have been exposed to something, because not only will there be a bright flash and you'll taste, like, metal, but a lot of accounts of people who have died of acute radiation sickness, radiation syndrome, whatever they're calling it now, have been like, they just know immediately that you have been heavily irradiated because you can feel it happening to you. TOM: [groans] Yes. ALICE: And the same thing is true on low doses. There is an X-ray feeling you can get, like, if you sometimes... Imagine, like, that scales up, you have to imagine, even if you're sort of precisely targeting the big Dr. No laser– It's not Dr. No–The big Goldfinger laser at you. TOM: So, uh, I think we're on the next slide. Right, so. Oh, the animation. Good. ALICE: Ah, LEGO! We have a cool LEGO diagram. TOM: We have a cool LEGO, we have a cool, uh, uh, is it, can we say "LEGO"? Connecting brick. Connect... ALICE: I think we can say so. A social democratic Danish brick building toy diagram... [laughs] of a radiotherapy suite. TOM: Yeah, and as Justin was alluding to earlier on, there's this sort of wall around, sort of very thick concrete or lead, or sometimes concrete and lead, and a big corridor with a big door at the end of it. And there are no windows between where the operator sits and where the therapy happens. ALICE: I like the extra line, like a light trap for film, where it's like, by all means, very applicable to radiation, but like, where it's like, at some point in the distant future, archaeologists will look at this and be like, "this was clearly... "a ritual chamber of some import," you know, it... has a mysterious arrangement of passages and very thick walls. TOM: Well, it's the, uh, it's the labyrinth, isn't it? It's... the labyrinth of Minos, which is sort of, just a very complicated palace design, but it's the sort of people who've come out of the Bronze Age Collapse, and it's, uh, “this is to contain a beast”. ALICE: Going very sort of Ancient Aliens and being like, the system of passageways and inlets and pyramids to let your pharaoh's car escape is actually an ancient radiotherapy suite. JUSTIN: I was about to say that, yeah. ALICE: Why do you think they built them so heavily, you know? JUSTIN: Really high energy radiation. We're talking like neutron star. ALICE: Ancient particle accelerator. TOM: So here, so what we've got, we've got two main therapy sources here. We've either got... spectrum radiation, so that's sort of X-rays and gamma rays, or particle... therapy. Both of them require the use of this thing which we sort of alluded to earlier, called a linear accelerator. So people sort of thought about, you know, know about your CERN, kind of thing, as a particle accelerator being that. But this is a particle accelerator. You have a gun that's an electron source, and then the reason for this sort of on and off rotating diagram is you have multiple magnets set at different electrical potentials which accelerate the electrons. And it's essentially sort of relativistic, about a millimetre wide, constrained in a narrow beam, with the electromagnets, and we measure the energy in which these are generated using something called electronvolts. So an electronvolt is the energy required to accelerate one electron through an electric potential difference of one volt, in a vacuum. ALICE: These are going through, like... you know in a video game where you have the... lines on the floor that light up and flash with a big arrow that speeds you up? TOM: Yeah, this is Wipeout for electrons. ALICE: Yeah, exactly. JUSTIN: [laughs] TOM: So they're just going faster, faster, faster, faster, faster through this beam, and then the majority of the accelerators for imaging work in sort of taking X-rays for, you know, you need to get a chest X-ray or something, that's sort of in the kilovolt range... kiloelectronvolt range, rather, and in the therapeutic range we're sort of talking about the megaelectronvolt. The beam itself can be therapeutic, so let's talk about particles, so you can move the beam around sort of using, these are called scanning, these are referred to as sort of, throughout the terminology we're using here, scanning magnets, bending magnets, or some people just use the term wigglers, to sort of move the beam. ALICE: Ha, yes! Thank you, finally– Bad and naughty electrons get put in the electron wiggler. JUSTIN: [laughs] TOM: Yeah, they get put in the electron wiggler, and that wiggles the beam around, and they sort of, (?), they sort of distribute the beam energy across an area, so the way this is set up is they go in a grid or whatever, so it sort of targets the tissue, and it does that, the two, the sort of, the... [sub note: Tom mixed up X-ray and electron beam re: deep stuff and surface; it’s the other way around. Check out the pinned tweet @fancywookiee] indications of these are that if you've got stuff that's on the skin surface, then you can sort of use X-ray therapy for that, because the skin is quite good at attenuating X-rays. So X-rays are kind of, you can't really treat stuff at depth with X-rays, whereas with an electron beam, you can actually target the depth at which you want this thing to hit, so the electrons will kind of miss stuff on the way in, but then hit a cancer that's, say, a couple of centimetres under the skin. So you've got electron beam mode, which is sort of firing the beam... through a sort of portal which can move it around, and then you have an X-ray mode, and to produce X-rays, the linear accelerator fires a beam of electrons into a tungsten sort of target, and the tungsten's used because it has a very high melting point, so high energy electrons hitting it at relativistic speeds don't melt it. ALICE: Next slide. TOM: Yeah, next slide. TOM: Oh, we, this hasn't come out so well on the black background, but... JUSTIN: Uh oh. TOM: Uh, yeah, we may have to mess this around in post, I dunno. I dunno if post happens. ALICE: I'm not sure if an image post happens. Certainly audio post happens. JUSTIN: Oh, an image post definitely happens, because Devon puts Devon in the... podcast sometimes. ALICE: [laughs] Yeah, of course. Okay, fine. Well in that case, we can supply a screenshot of the original. TOM: Yeah, sure. So I'd sort of done a thing here to sort of indicate what happens, where basically... you fire your electron beam at your tungsten target, and then X-ray photons are generated by one of two means. And, we're remembering that this is now, we've now gone from basic biology to basic atomic physics. So if– ALICE: I'm still stuck in military history, so I'm like, yeah, of course, the round hits the tungsten plate, causes spalling, which kills the crew. Um... JUSTIN: [laughs] TOM: So... the electron that's been fired by the linear accelerator smashes into one of the electrons in the lower orbit of the tungsten atom, and knocks it out. So the way that things work in physics is that things want to be going from the highest energy, things that are in high energy states want to be in low energy states. So as one of the electrons from the outer shell moves down to replace the electron that's been knocked out, it emits an X-ray photon, and it gets rid of the excess energy. Then you also get a phenomenon which is called bremsstrahlung, which is as the electrons from the linear accelerator are slowed by the mass of the tungsten nuclei, they lose energy. And again, that lost energy is emitted as an X-ray. And that's kind of about 80% of the generated X-rays. That's not really targeted, and that's kind of the background radiation. It's very, very inefficient. So about 1% of the input energy here is emitted as X-rays, and 99% is lost to heat. And so to account for that, you have to increase the power of the beam. The current of the beam is about 100 times higher in X-ray mode compared to that of the electron beam to generate the same output voltage, basically. So 25 megaelectronvolts of X-ray energy takes more input energy from the linear accelerator than outputting 25 megaelectronvolts of electrons. ALICE: Raising my hand for a question here. TOM: Sure. ALICE: You can target the depth of this when you're using it with, like, if you're using this in a particle setting, right, and you can target the depth. TOM: Sure. ALICE: Is the advantage of using this on skin level with an X-ray high enough to make it worth all of this extra bullshit that you have to do, including all the extra power? TOM: This is where I sort of start to fall down, because my clinical experience is not that of radiotherapy. What I know is, this is, from going into this, this is basically having the dual mode gives you a range of treatment options, and it's whether you're treating, like... What the electron beam, as I understand it, can treat is quite narrow, and quite small. ALICE: I see. People do get, like... TOM: Whereas X-ray– ALICE: Sorry, go ahead. TOM: Whereas with the X-ray, you can effectively sort of radiate a field. ALICE: Very inconveniently, people insist on getting a bunch of different weird cancers, in different places and times. TOM: Yeah, they happen in different places at different times, and... ALICE: ...rationalize this cancer thing, I think. It's very inefficient. JUSTIN: My question is, if you've got a machine that does both of these things, they're on separate circuits, right? TOM: No. JUSTIN: [laughs] LIAM: Grind never stops, baby. JUSTIN: Ah well. I can see some problems that could develop here. [laughs] TOM: So this is the... So, I think we may be onto the next slide. ALICE: It's time for us to go to Canada, and we must ask the world... TOM: Canada! LIAM: Oh no. JUSTIN: Canada! ALICE: What is Canada– LIAM: Eat shit, Riley! ALICE: [laughs] A miserable pile of secrets. TOM: So, after the war, there was a Crown corporation established in Canada. Which is a sort of Crown corporation–so Canada is obviously part of the Commonwealth, and so the Queen is still the head of state, and a Crown corporation is basically a sort of, as I understand it, profit-generating entity that's got some amount of... JUSTIN: Quasi-nationalized. TOM: Yeah. LIAM: Please see more for our—Newfoundland Rail episode. JUSTIN: Yeah, Canadian Nationals, a Crown corporation, stuff like— Or it was, it isn't anymore. It's sort of organized the same way Amtrak is. TOM: So this is... So AECL, the Atomic Energy of Canada Limited, was set up in... ALICE: The Amtrak of smashing atoms. TOM: The Amtrak of smashing atoms for specifically non-military nuclear energy devices. They developed us a early sort of civilian nuclear reactor, which was called the NRX. ALICE: It was a neo-reactionary nuclear reactor. TOM: Uh... [laughs] ALICE: Peter Thiel was actually generated in this reactor. LIAM: Oh no. TOM: So this first device, the NRX, was located in Chalk River, Ontario. This underwent a partial meltdown due to insufficient cooling. This resulted in contaminated water flooding the basement section of the reactor. Among the military personnel involved in the cleanup was one Lieutenant James Carter, of later peanut farming and US presidential fame. ALICE: ‘Cause he was a nuke guy, wasn't he? He was a nuclear engineer on submarines. TOM: This is where he kind of made his name, as part of the cleanup detail for the NRX. But yeah, this was sort of where he started to come to prominence. They then developed another design called the CANDU nuclear reactor. This is all... yeah. CANDU nuclear reactor. ALICE: Haha, very funny. ALICE: Canadian, like, depleted uranium, I imagine. Something like that. TOM and JUSTIN: Deuterium. ALICE: Ah, deuterium, okay. JUSTIN: It's fun, because you can't make nuclear weapons with it. So you can sell it to developing nations without as much of a proliferation risk. ALICE: I miss when they called it "heavy water". That was much more evocative. TOM: This is sort of their signature product, and they are, as far as 2011, they remain a Crown corporation, although in the same way that a lot of formerly national-type corporations have gone, all their operational activities are now privatised. They've brought in Canadian equivalent of Capita or whatever to actually run the day-to-day bits and pieces. ALICE: [groans] ALICE: AECL brought to you by... TOM: Yeah. So, and then... LIAM: Bleh! TOM: There's a lot of this... [sighs] Yes, they had this sort of asset, this is all reasonably good stuff. They did a lot of manufacturing of medical isotopes, so things like Cobalt-60 and Molybdenum-99, which is a contrast agent. In MRI imaging, you can't inject any sort of dye, because it's all to do with proton weights and stuff, but what you can do is give people Molybdenum-99, which has a heck of a lot of weight in it, and that is a contrast... ALICE: Gonna give you a very dense injection. TOM: Yeah, it gives you basically dense in nuclear... sorry, in um, in nuclear medicine, rather. So yeah, it's a kind of contrast agent, a conventional dye won't work, but if you give something that's heavy and radioactive, then it will show up. And then they developed linear acceleration technology for (?) radiotherapy, in conjunction with a French company, the Compagnie Générale de Radiologie, CGR. And they joint-marketed a device called the Therac-6, which the French– ALICE: ...You saw way in the beginning. This big sewing machine looking thing. TOM: The big sewing machine looking here, which the black and white images do not cover the sort of lovely colour scheme. JUSTIN: I'm getting a bit more of a stand mixer vibe from this, honestly. ALICE: Yeah, I can see that. TOM: It is your... It's the KitchenAid, isn't it? It's the KitchenAid. JUSTIN: Yeah, I mean, imagine you could put some flame decals on the side here, y'know? TOM: [laughs] ALICE: And they're gonna attach the dough hook and they're gonna fuckin' whisk the shit out of this guy's arm. TOM and JUSTIN: [laugh] LIAM: Oh dear. TOM: Oh no. Ah. So, the French named this, much... This was a... I don't know who was sort of here, but whoever was working at AECL was very much a sort of... pragmatist. ALICE: Sort of like a louche French guy, being like, [French accent] “(?) zis machine, zis is Neptune”, and then some Canadian guy from Ontario called Dave is like... “Okay. What’s that...” TOM: It's the Therapeutic Accelerator, and it outputs at 6 megaelectronvolts of X-rays: It's the Therac-6. ALICE: Beautiful. Poetry. Poetry all the time. TOM: Beautiful. And then, so they added in... So they took the French machine, le Neptune, and they added computer control elements. So they sort of added... They had a... Rather than the operator sort of have to set everything up manually, and then sort of go out of the room and just sort of press the on button, they sort of connected this to a PDP-11 mini computer. I mean, sort of relative term, because the “mini” is about... The base unit is about the size of a domestic fridge. ALICE: Oh, I mean, that's “mini” in that you don't need a LIAM: It’s pretty “mini”. ALICE: computer building, just a computer room. TOM: Yeah, you don't need a computer building. And then that's attached to to a terminal, which is in the operator suite. And then this was superseded by a second generation machine called the Therac-20. This was another CGR design. This was... Le Sagittaire. ALICE: Even nicer, you know? TOM: Yes. ALICE: “We're gonna put you in the Sagittaire”? Sounds delightful, sounds luxurious. It’s like, oh, you're gonna put me in the luxury suite of a hotel room in Tahiti. Fantastic. TOM: Yeah. LIAM: Just mini fridge that will stop your cancer, yes. ALICE: [laughs] Exactly. LIAM: Regular fridge, whichever. TOM: So this was, um... ALICE: “Macrofridge”. TOM: [laughs] “Macrofridge”. This is a bit more... marketed as–this could do both X-ray therapy and electron beam therapy. So this could do deep tissue... so the Therac-6 was an X-ray based machine, it could do field treatment, but not deep tissue treatment. Whereas this was like, okay, we can do the field treatment with the X-rays, but we've got to sort of... you can change the head around on it– ALICE: It slices, it dices, it's got things for removing stones from horseshoes. TOM: This is where you've got your dough hook, and then you've got your mixer attachment. ALICE: [laughs] Yes, yeah, exactly. TOM: So you do this by basically, you've got a head which we can sort of see here on the 6, the business end of it, where you can rotate that around. But the problem with it is that all of this is made out of steel, and it's got a huge lump of tungsten in it, and then on the other side of it, it's got a counterweight for the tungsten, which is sort of iron or concrete or something extremely heavy to balance it, and this thing's all on rails and it sort of rotates around and clocks into place. And again, this was controlled by a PDP-11 minicomputer. So we'll move on, I think we're onto the next slide. And we're onto... ALICE: This is a beautiful little computer. I like this a lot. LIAM: An hour and 26 minutes in, it's the subject of today's episode. TOM: The subject of today’s episode. JUSTIN: [laughs] TOM: And I have to say, you've got this beautiful little VT100 terminal emulator. Well, not even terminal emulator, sorry, I say “terminal emulator” because when I graduated from medical school, we were still using a VT220 terminal emulator for getting results of stuff from the lab. ALICE: Really? TOM: Yes. So that was back in... A quick digression into IT in the NHS, when I graduated from medical school in 2007, they were still looking up results on a VT-220 terminal emulator, which is running in DOSBox. ALICE: Amazing. LIAM: Beautiful. TOM: Yep. We are now in 2023, and most NHS systems are just about running Windows 7. LIAM: Wow. That's terrible. TOM: Yeah. ALICE: The United Kingdom of Great Britain and Northern Ireland. TOM: There are... the most advanced I've got was when I did one of my exams, and I have a laptop for that, and that was actually on Windows 11, and I didn't know what I was doing, because frankly... ALICE: “Where's the fucking start button? Where–what?” TOM: Yes. TOM: So then you've got the actual computer itself, which is the thing with the tape drives, and beautiful pink trim. ALICE: Oh yeah. Someone colour-coded that. TOM: Yeah, this is, full... ALICE: Someone went, like, the Pantone cap, and was like... TOM: ...full, um, kind of, you know, it's the early 80s, we are gonna have our, sort of, we're gonna have our synth pop on the radio, and we are gonna have our synth pop-themed mini-computer, and then, this is the only... So, you know, for an example of, like, you know, people have been calling out for this, an example of how bad this is gonna go: this is the only extant picture that I could find of a Therac-25. If you put “Therac-25” into Google Image Search, you will get lots and lots of pictures of other linear accelerators, and people go, oh, this is the Therac-25–No. I... very much tried to do my due diligence on this, and find a picture of one, and this thing has been scrubbed from existence. ALICE: Wow. TOM: You can find the Therac-6, you can find, I think, a couple of pictures of the Therac-20. The Therac-25, this is the only picture, which is an artist's impression from a promotional brochure, which was held as part of a library into the Atomic Energy Archives of Canada, I think that's the source of this. And it's the only extant picture of the thing that exists. ALICE: Just like, grainy drawing. It looks more hostile than Therac-6 does, it looks a lot less, sort of, kitchen art... TOM: Yeah. The... JUSTIN: You got a sort of ominous cantilever, y'know? LIAM: Yeah. ALICE: Yeah. TOM: Yeah. So this was... ALICE: Puts me in mind of, like, ancient Egyptians again, I'm not sure why. TOM: They, so at this point, this was developed independently by AECL, so they'd ended their partnership with the CGR, the French company. ALICE: It wasn't named, like, Le Seductrice, or whatever. TOM: No, no, this is... ALICE: Pure Canadian. TOM: No, this is the Therac-25. It does therapy at 25 megaelectronvolts. Does what it says on the tin. It was, so this was using their, sort of, they patented in the 1970s what was called a dual-pass linear accelerator. So basically, it fires the electron beam through that system of alternating electromagnets twice, which is why they can get the... therapy megaelectronvoltage up to, sort of, 25 MeVs. And it can do 25 MeV X-ray, and it only does 25 MeV X-ray, because that's the energy it takes to, that's, you know, using pretty much the maximum sort of beam output to generate that at 25 MeV. Or electron beam therapy for 5 to 25 MeV. And the reason you vary the the megaelectronvoltage is to, sort of, you know, account for depth, attenuation, different types of tissue. And the other significant change is that this was built from the ground up as a, because this is now the, sort of, early to mid-1980s, this is a fully software controlled system. LIAM: Oh no. JUSTIN: Wow. Oh boy. ALICE: It’s going on the computer! JUSTIN: On the computer, yeah. ALICE: No more (?), knobs, and buttons, it’s all gonna be controlled through this new little terminal. TOM: Yep. TOM: It’s the future. We do things with computers now. So, it’s fully software-dependent, which is responsible for both the machine operations in terms of setting the... the voltage of the treatments, and then also, responsible for the safety systems. Because, in the previous case, that used hardware interlocks, which– the system software had software checking on it, but it was backed up by hardware interlocks, so the thing physically wouldn’t fire unless it checked, y’know, there is a switch that says, “the head’s in the correct position”, “The beam’s set correctly”, it won’t fire. I’m gonna shout someone out now, specifically, because there is one person who’s... who's basically done the bulk of work on contemporaneous documentation of the Therac-25, and that is Professor Nancy G. Leveson. She is currently still working as the Professor of Aeronautics and Astronautics at MIT, and a specialist in software systems and safety. ALICE: I assume because the software systems and safety on this went so well. TOM: I mean... one of the primary sources for this is an extensive article, a paper that she basically wrote as a complete breakdown of what went on with Therac-25, is why we know so much about it, because a lot of the documentation that otherwise might have happened, it's just buried in things like lawsuits that were settled out of court. There was a net, we'll get onto what happened with the FDA, but a lot of this was documented by her quite extensively. The appendix on Therac-25 is published in Safeware: System Safety and Computers, and quotes an AECL assurance manager stating that “the Therac-6 package was used by the AECL software people when they started the Therac-25 software. The Therac-20 and Therac-25 programs were done independently from a common base.” One thing to remember here is the use of– ALICE: No worries, then. TOM: Yeah–the use of the word “people”, and we'll come back to that. Just park that. That is... Chekhov's people. ALICE: [laughs] TOM: Park that under the chain and come back to that. Next slide, please. Again, this is one way we probably need to do it in post. ALICE: We'll just– I'll send Dev the link to the thing, they'll fix it. TOM: So we've got a– LIAM: Thanks, Dev. TOM: Thank you, Dev. Thank you so much. So there are three main components to this, and they rotate back and forth around the central axis. You've got a stainless steel mirror and a light assembly, and this is–it's a light. And that shows where the thing's gonna show. So you get your patient into position, they lie down on the table– ALICE: Cool laser beam looking thing. And be like, we're just gonna aim this at the cancer. TOM: It's not even lasers, it's just a– this is the 1980s, this is a 100 watt bulb... ALICE: We're gonna aim this weak flashlight at the cancer. TOM: This completely– LIAM: (?) Goddammit. TOM: Yeah, so that basically shines the light in the treatment area. They sort of fiddle around with it and get it so, okay, we get it in the middle of the field, get it so that the thing's gonna be set up. You've got your X-ray target, which is a conical piece of tungsten that's called a beam flattener. And Leveson describes that as being an inverted ice cream cone. So yeah, that's what's in the earlier picture. It's sort of a conical bit of tungsten. So where the beam hits it sort of filters out through the conical bit of tungsten. And then what you get is a flat X-ray beam that kind of comes out with an even spread. And then you can do stuff like put a sort of lead sort of trays on the patient to sort of further direct that. So... if you don't want to sort of irradiate a huge area, you can say, well, okay, we're doing a sort of skin thing, but we want to just kind of have the sort of eight centimeters where this thing is, and then the rest of it is like a tungsten plate– a lead plate that goes over the top. ALICE: Upturned baking tray with a hole punched out in the shape of your melanoma or whatever. TOM: Yes, exactly. And then the electron mode, which is sort of surrounded by the scanning magnets, the wigglers as we've referred. That was a delightful term that I sort of heard when I was sort of doing the research for this on another video. They called these things wigglers. ALICE: I'm so pleased that that's what they call them. TOM: Yeah. So they then have what's called an ion chamber. So, that measures the ionization levels in the air as the therapy beam or the X-rays exit. So that records the dosage that's delivered to the patient and then puts that dosage back on the terminal. And then as we talked about, we've got a counterweight, and on the counterweight are three micro switches and they control the position of the therapy head. And so the position of the head can be set by the operator terminal or by use of a hand control. And this is one of the things on the Therac-20 apparently that sort of operators complain about, this thing is, like, y'know, this is very heavy, and we have to manually click this into place every time. So we've automated this process. LIAM: Oh good. TOM: Yes. ALICE: You have to look down through this and like, aim it first, aim with the mirror, then switch it to the thing. TOM: Then switch it to the thing, yeah... LIAM: That sounds like a lot of work! ALICE: I have to use this submarine periscope thing. TOM: Yeah. So... JUSTIN: Big wheel you gotta crank, y'know? TOM: Yeah. LIAM: Hamster wheel you actually have to run on. ALICE: Radiation medicine Sisyphus, y'know? LIAM: [laughs] TOM: And then there's a little piston here, which is like a locking device that kinda helps to lock the head in place once the thing's been selected. Next slide please. So then, the software component. So... this is the simulation of the operating layout that the operator would see. LIAM: That looks very Kevorkian, I like that. TOM: Yeah. JUSTIN: It's got very... whatchamacallit, it's got Matrix energy here. LIAM: Yeah, bad ones. Bad Matrix energy. ALICE: I particularly enjoy “treatment mode fix”, as opposed to “treatment mode break”. [laughter] ALICE: “Treatment mode make worse, as a joke”. JUSTIN: Yeah. TOM: So, the reason you'll see sort of dates on this that are sort of saying, 20-something, this is running in an emulator, so people have been able to reverse engineer and get hold of the code for this thing and then run it in a terminal emulator to see how it does, and then sort of to replicate some of these things that might have gone wrong with it. LIAM: Software nerds are amazing. TOM: So, the things that are on the top line are the beam type, where operators entered E for Electron. So this has got to attach a little keyboard. E for Electron. ALICE: E, E, E. TOM: Or X for X-ray. LIAM: E. ALICE: [laughs] JUSTIN: If you put in other letters, did it give you weird kinds of beams no one ever heard about? TOM: Uh, as far as– LIAM: Oh yeah, my favorite, F-rays. ALICE: Getting shot with a Q-ray, yeah. JUSTIN: Oh God. [laughs] TOM: Yeah. LIAM: “Everyone around me is a pedophile!” [laughter] ALICE: Yeah, you get shot with a Q-ray, you start believing some really strange things about JFK Jr. TOM: So, the next sort of–so, that's actually the one thing just noted on this screen, this is talking about kiloelectronvolts, we've already said this is a megaelectronvolt thing, and if you tell it to X-ray mode, this thing will just default to putting in 25 megaelectronvolts as the energy. It doesn't have any other options for X-rays. The second set of fields is the prescribed dose, so they go in and they say, okay, I want this amount of radiation for this many rads, this is all rads, this is all using rads because, again, this is... ten years after the implementation of grays, but this is in the US. ALICE: 200 rads, 2 grays. TOM: Yeah, yeah. So that's how many rads of radiation. And then you enter that, the third set of fields is where the gantry is, so that's the position of the head, and then the bottom field, the date, time, the operator ID, you know, this is Kevin doing the thing today, and tells the operator the beam's ready to use. ALICE: Alright, so you look at this, everything's actually– TOM: Everything's good, you go da-da-da-da-da, enter in all the information, you know. ALICE: Phewsh. Great, perfect. Cancer over. LIAM: And kill a guy. TOM: So, because we've sort of discussed the design of the setup of this before, because you need to protect your staff from the ionising radiation, the terminal itself is located outside the therapy room, you've then got radiation shielding in the walls and a sealed door between the treatment area and the control area, and there's usually a two-way intercom system between the treatment area and the operating area so the staff can say, you know, okay, we're gonna do this thing now, there's gonna be a bit of noise, you know, if you've had an MRI or something like that, they kinda go, yeah, it's gonna be noisy, and then it sort of sounds like the wheels are coming off every single tricycle in the world at once. [laughter] ALICE: And then you have a panic attack, and they go, stop doing that, and you go, I can't, and they go, well, try harder. LIAM: [laughs] ALICE: You continue on this way... until you do the thing where they sedate you, y'know? JUSTIN: I had an MRI once and it was fine. You know why? Because I knew I could easily get out of that hole if I had to. LIAM: That's nice. Congratulations, fuck-ass. JUSTIN: Yeah, exactly. ALICE: Whereas with this thing, you gotta, like, JUSTIN: It’s not a cage, it’s a machine. ALICE: get up, sprint down a corridor and try and like open a sealed door. Which is... LIAM: No, well, that's no match for Rocz, of course. JUSTIN: I'm simply built different. LIAM: You are simply built different. And we are... TOM: There's... TOM: probably another episode's worth of stuff in talking about MRI machines and what's called “missile effect”. JUSTIN: [laughs] LIAM: What a great name. [laughs] ALICE: Just directly sort of laser target my personal fears, why don't you. TOM: Yes. TOM: This is the, y'know, why the MRI machines have the big lockable door, and everything in there is plastic. Everything in there is plastic. If there is a crashcall... the separate sort of plastic things that they have to kind of grab, because... missile effect. ALICE: Yeah, the time they killed a kid in an MRI because he, like, went into cardiac arrest and there was someone brought in a metal oxygen tank that then got fired across the room, killing the patient. TOM: Yeah. It's awful. JUSTIN: It's gone poorly. LIAM: Medicine's great. TOM: Yeah, it is absolutely, y'know, that is the kind of thing that happens when I'm doing radiation safety, and imaging safety as part of my course, that was kind of things like, this is why this is metal– why nothing is metal here. Absolutely nothing is metal. If you have metal, leave the metal outside. Just do not bring metal into the unit, don't bring metal onto the floor, don't have metal. Metal not required in the MRI environment. ALICE: Me talking to Spotify. [laughter] TOM: So, during this sort of development system, test operators complained that entering the data again on the terminal was too slow, and AECL implemented a function whereby you could sort of copy the treatment site data by sort of process of a (?). So yep, yep, yep, okay, fine. You know, it just did copying all over. And if the software detected an error, it would respond with one of two error states. So either treatment suspend, in which case it shuts the machine down, and the whole thing would have to be restarted all over again. ALICE: The annoying error. TOM: The annoying error. LIAM: Yeah, that sounds convenient. ALICE: That's sort of, by default, the thing that... incentivises you to use or work around the other error form, so that you don't have to do this. TOM: And then a treatment pause, which is where– again, as we've established, dealing with a lot of literalists at AECL. The treatment pause, which could be overridden by the operator pressing the P key. And up to five treatment pauses were allowed before the machine would require a system restart. Error messages were generated alongside a number code. So that's good, you get your number code, and you can look up in the documentation. ALICE: Error code 69. Uh... TOM: Error code 69 not implemented, the current one to 64. JUSTIN: Ah, well. TOM: Yeah. LIAM: Lame. TOM: Yeah. JUSTIN: Unfortunate. TOM: The supplied operator's manual did not contain a descriptor of what the error messages were. LIAM: Good. ALICE: [laughs] LIAM: Thank you. ALICE: You don't need to know. Error code 13? What is it? Don't worry about it. JUSTIN: Don't worry about it. TOM: Don’t worry about it. The maintenance manual did contain descriptions, but not whether they were of clinical risk to patients. ALICE: Of course not. You don't need the maintenance people to know that. TOM: Yeah. So, before... ALICE: Just sort of like, some real, like, 60s medical chauvinism of being like, “well, he's not a doctor, is he? So, “whatever, to be honest.” TOM: Before commercial distribution they did a safety analysis of the device, but this was conducted on a hardware simulator under treatment conditions and they ran a sort of fault tree. But this only accounted for hardware failures, and not software. And the assumptions that AECL made that sort of Leveson summarised were that one, is that sort of programming errors have been, this is what they said, they said that “programming errors “have been reduced by extensive testing on hardware simulator “and under field conditions on teletherapy units. “Any residual software errors are not included in the analysis. LIAM: Oh, good. “Program software does not degrade “due to wear, fatigue or reproduction process.” JUSTIN: Oh, okay. ALICE: Sure. JUSTIN: Wasn't that like the first half of this presentation? How programs decay... JUSTIN and ALICE: [laugh] TOM: So, I originally sort of came on because I was like, hey, there are some interesting medical failures that I'm aware of, of engineering. And then Alice sort of mentioned to me, oh hey, we've been looking to do the Therac-25 for a while. And I sort of went away and I looked at it, and I went, there's a throughline here. Because cancer is in itself a failure of essentially meat coding. And we can look here, we can draw an important parallel between that and a sort of a cancer treatment which is hampered by an error in sort of silicon software coding. So... ALICE: it's all computers all the way down. TOM: It’s all computers all the way down. Yeah. You cannot escape the computer. JUSTIN: We are all going on the computer. We have all always been going on the computer. ALICE: Yes, yes! LIAM: [laughs] Looking through the box! TOM: The third point is that computer execution errors are caused by faulty hardware components and soft, random errors induced by alpha particles and electromagnetic noise. ALICE: That's such a, like, “my computer is the size of a fridge” kind of conception of how computer errors happen. It's like, this is a machine, this is a machine on a human comprehensible scale, I don't need to think about circuit boards yet, really. There is, like, something has got into this cabinet and made a connection wrong. TOM: So this is the sort of, this random error thing, this is an an alpha particle error, this is a technique in speedrunning. ALICE: Oh yeah, of course, that happened, didn't it? TOM: Yes, this is the single-event upset. ALICE: A bit flip. TOM: Yes, a bit flip. So that's what they said, basically, if an error occurs in our software, it's because a solar ray just happened to hit that bit of the... LIAM: It just did that. TOM: And it just did that. JUSTIN: It just did that. TOM: Yeah, so the famous one is the 2013 Mario 64 speedrun, in which DOTA_Teabag... ALICE: Great name. TOM: was doing speedrunning things, and had Mario suddenly walk to the upper floor of a level, which sort of shaved seconds off their time, and allowed them to sort of do the best Mario's could. ALICE: Like an absurd cosmic error in your favour, that this one bit has been flipped by an outside force in such a way as to benefit you. JUSTIN: I have a new theory about JFK. LIAM: Oh boy. [laughter] ALICE: Yeah, he got bit flipped. LIAM: He got bit flipped. JUSTIN: Then his head just Did That. ALICE: The bit here is the IsAlive modifier in the human head. JUSTIN: [laughs] TOM: So a total of 11 Therac-25 devices were installed in hospitals throughout Canada and the USA. And so, we get to do a date! We finally get to... Justin! ALICE: Oh no. ♪[“Local Forecast - Elevator” by Kevin MacLeod]♪ JUSTIN: Hi, it's Justin. So this is a commercial for the podcast that you're already listening to. People are annoyed by these, so let me get to the point. We have this thing called Patreon, right? The deal is, you give us two bucks a month, and we give you an extra episode once a month. Sometimes it's a little inconsistent, but you know, it's two bucks, you get what you pay for. It also gets you our full back catalog of bonus episodes, so you can learn about exciting topics like guns, pickup trucks, or pickup trucks with guns on them. The money we raise through Patreon goes to making sure that the only ad you hear on this podcast is this one. Anyway, that's something to consider if you have two bucks to spare each month. Join at patreon.com/wtyppod. Do it if you want. Or don't. It's your decision, and we respect that. Back to the show. JUSTIN: Oh boy. June 3rd, 1985. TOM: So, 61-year-old manicurist Katie Yarbrough attended the Kennestone Regional Oncology Center in Georgia, in Marietta, Georgia, for her 12th dose of radiotherapy treatment. She had had a previous surgical excision of a breast tumor, and had follow-up treatment with radiotherapy for local lymph nodes. Now, from... where there is medical information, I've kind of jotted down medical notes. This is pretty standard for breast cancer treatment today, even. So you can do... a small enough tumor, you can take a lump of breast tissue out. We call this a wide local excision, but a "lumpectomy" is still a term that's often used, and it's removed with a cuff of normal tissue around it. And then they send often a lymph node sample along with it, and.... for the closest group of lymph nodes, and... in the armpit usually, where we kind of look at one of those nodes and we look at it under the microscope and determine whether there's cancer in it. So if there is spread to local lymph nodes, or if they suspect there might be, then these can still be treated either with radiotherapy or excision of the local lymph nodes. And this is still pretty standard relative to modern practice. 1985 to now, things haven't changed that much. So she was just– ALICE: She's had breast cancer, has been successfully treated, and she's now in for this follow-up in the armpit. TOM: Yes. LIAM: Right. TOM: And she was prescribed a 10 megaelectronvolt electron therapy dose. So she goes in, (?) positioned, the operator exits the room, presses the on button, and then she describes feeling a tremendous force of heat, a red hot sensation, and the technician– ALICE: Is this a– ALICE: That's not, it's not supposed to do that, right? TOM: No. JUSTIN: Right. TOM: ...This is her 12th dose of, 12th fraction of radiotherapy treatment. So she's had 11 previous ones, and they've all been kind of fine. You know, she's gone in there– JUSTIN: “This patient was one day from retirement.” ALICE and JUSTIN: [laugh] TOM: Oh no. So she's gone in there, she's had the... 11 doses, they've all been fine. And then suddenly, you know, that really, really burned. The technician said, this is impossible. She said, you know, she said, you burned me. And they said, that's impossible, it can't happen. ...the treatment area did sort of feel warm to the touch. And then they went, we're sorry about that. And she was sent home. And... ‘cause she's had her treatment, she's returned, that's, you know, sorry, you know, fine. Over the next few days, however, the skin over the upper chest and the shoulder became painful. Hot, red– ALICE: Acute radiation exposure, it takes a minute. It takes a minute, you know? TOM: She was admitted to the hospital in Atlanta. But then, so she was admitted to Marietta, which– my geography of Atlanta and of Georgia is not amazing. She was admitted to the hospital in Atlanta, so the sort of secondary or tertiary center, continued to be sent to Kennestone for further doses of radiotherapy while she was inpatient with her radiation injury. The pain sort of continued to increase, her shoulder joint became immobile. And the treating doctors thought this might be due to treatment effect from the radiotherapy, so they just thought, okay, so she's had several doses of this, and we know that radiation has sort of stochastic effects, but then it's gone beyond that now to where the skin is breaking down. And not only has she got this redness on the front of her shoulder, it's also on the back. ALICE: Like you've been shot with a big beam of radiation. TOM: Like you've been shot with a relativistic beam of electrons. ALICE: Einstein gun. Sort of Einstein gunshot wound. TOM: So she was then seen by the medical physicist. So this is a big shout out, whenever you have radiation emitting things in a hospital setting, not only do you have technicians and radiooncologists and people who prescribe and deliver these sorts of doses, you also have to have a medical physicist. So a person who's qualified at radiation and knowing how the radiation is made, to a sort of... this is the person who knows all the theory of the radiation stuff. ALICE: Knows all the German words about why... TOM: Yes. JUSTIN: It does seem like almost universally, radiation is always like the last thing anyone thinks about, about any particular injury. TOM: Yeah. So you have your, and that's not to say they don't have a knowledge... I'd say having done some sort of bachelor level degree stuff, you have to have an idea of the radiation safety, but these are the people who absolutely know their stuff about medical radiation physics and in their place in the medical setting. So this guy, Tim Still, is the Kennestone physicist. And he's exactly the kind of obsessive nerd you need on staff to recognise radiation problems, and then make meticulous notes about it. So it was obvious that this was a radiation injury. He reviewed her, determined she'd received one to two doses of radiation in the 15 to 20,000 rad range to the effect... LIAM: Jesus wept. ALICE: If we go back, by the way, to the thing, more than 1,000 rads is in the “acute illness, early death”, lethal dose 100 after 10 days category. The highest one that we put up. TOM: Yes. The highest one. So, the key thing to say again, this is not a whole body radiation dose, this is just to a very small area. You know, effectively, the width of an electron beam. So about a millimetre, so you imagine 15, 20,000 rads being applied over about a millimetre wide area of tissue, basically. So, Tim Still contacted AECL to ask if the Therac-25 could operate in electron mode without the beam scanning. And was told– ALICE: And of course, they're like, nah. TOM: This is impossible. LIAM: Nah, don't worry about that. TOM: Uh, this was not reported to the FDA, because they could not report it to the FDA. Because the FDA protocol at the time was that reports regarding medical devices could only be made by the medical device producers and importers, and not by end users or patients. LIAM: Beautiful. That is smart as shit. ALICE: So if the Canadians don't want to do shit about it, nothing you can do. TOM: Yeah. So unless the manufacturer wants to do something about a device, and their position is, well, this is far safer than the previous device because of computer. JUSTIN: Yes. TOM: This uses computers, this is safer. There is no way in which this could cause an incident. And fair play to Tim Still, because... as we'll discover later, this is kind of on the money here. This is, all the... medical physicists involved here were absolutely on the money on what was going wrong. And, uh... JUSTIN: Instantaneously diagnosed the problem. Not that anything's gonna be done about it, but... [laughs] LIAM: But. TOM: So. A second date. JUSTIN: Oh boy. LIAM: Oh, so you know it's really good. JUSTIN: July 26th, 1985. TOM: [sighs] TOM: So, this, in this case, is a 40 year old. This is, so we're now, we've gone north, we are now in the... Ontario Cancer Federation Radiotherapy Facility in Hamilton, Ontario, for a 24th treatment for cervical carcinoma. Again, cervical carcinoma is something that is kind of treated with radiation in the sort of early or locally advanced stages. One of these things whereby if you've got sort of the risk of recurrence after surgery, or if someone's got, you know, an inoperable sort of tumour, but that's one that's bleeding, then you can just do palliative treatment with radiotherapy, sort of basically sclerose the tumour and stop it from bleeding and give someone—you know, while you're not gonna cure their cancer, you know, they're not sort of getting lots of sort of vaginal bleeding or something like that that's uncomfortable and it's trying to give people some, you know, some element of dignity, even though they're in the sort of final phase of the cancer. And this is either done by radio, sort of external radiotherapy like this, or you can implant radioactive pellets that sort of emit radiation locally, that's called brachytherapy. So on this occasion, the patient was receiving therapy, they were positioned for treatment, the technician set the device up for therapy. It was initiated, and the machine shut down after five seconds, reporting a H-tilt error. And the terminal screen... JUSTIN: ...Sort of a pinball machine thing... TOM: Yeah. Sort of tilt, you know... LIAM: You made it mad. TOM: Yeah, you made it– ALICE: Do we have an idea of, like, how long a typical exposure is supposed to be, just to contextualise that? TOM: Ugghhh. ALICE: Just a ballpark, are you there for like seconds, minutes? TOM: It's kind of seconds, really. Some of the more scanning electron beam may take a minute or two, but this is pretty quick. An X-ray exposure, people have had diagnostic X-rays, that's a couple of quick seconds, that's the field irradiated. ALICE: So there's a joke I remember about chiropracty, which is like, a medical doctor is like, well, we'll do some tests, and we'll try some things, and if it doesn't work, then we'll know some more things. Whereas a chiropractor is sort of like, sprinting towards you, setting up for a football kick, like, say goodbye to back pain! LIAM: [laughs] Sort of like this, but with cancer, is like, sprinting towards the beam, just like... TOM: So, the terminal screen health... JUSTIN: You know what the real joke is, is chiropracty. LIAM: Chiropracty. ALICE: Oh yeah. [laughter] ALICE: Learned how to do it from a ghost. TOM: Yeah, as someone that's formally done some training in orthopaedics, I have seen a non-zero number of chiropractic injuries. Anyway, I don't want to get sued by a ghost, so I'll not say any more. [laughter] TOM: So the terminal screen displayed no dose, and indicated that there was a treatment pause. So the operator presses the P key. Again, another no dose error. And most operators at this point... this was a finicky piece of kit. They put the patient in the thing, they have to go all the way out of the room, and then they press the button to go, and then it goes... No. [laughter] And they sort of go back into the room, and they check it, and they come back out, and they press P again. And it's like... No. Okay. So, this was not an infrequent occurrence. And on this occasion, the operator pressed the P key a further three times. This is triggering the magic five times they've had to press the P key. LIAM: Ah, yeah, Sticky keys, is now enabled. ALICE: [laughs] TOM: So, the Sticky keys is now enabled, and the machine enters treatment suspend. So it shuts down, and then you have to call the technician, and the technician goes, “what have you done?” They have to basically come and boot everything back up again, and this is a... ALICE: Turn it off at the wall, turn it back on again. TOM: Turn it off at the wall, turn it back on again, wait for the tape drive to spin up, et cetera. ALICE: [laughs] TOM: And then, shortly after treatment, the patient complained of burning and tingling pain in the treatment area. They returned for further treatment three days later. By this point, they were then reporting severe pain, swelling and redness, and then were hospitalised the following day. ALICE: Radiation injuries. You feel kinda weird, then you feel really weird. LIAM: Then you feel real bad. TOM: So, the machine was taken out of service, and the AECL was again informed of a potential radiation injury. They sent an engineer to investigate the machine, and as the manufacturer was now involved in the proceedings, the Canadian Radiation Protection Board was informed of the incident, and the FDA was sort of starting to be made aware of this. The Canadian Radiation Protection Board issued a notice that users have to manually confirm the treatment head position, but it didn't state anywhere there'd been any patient harm. So the patient who received an excess of dosage in this scenario, unfortunately succumbed to her cancer a few months later. ALICE: It was a terminal case coming in. TOM: Yeah, this is the sort of case I was talking about where someone's having palliative radiotherapy for a cancer that's, you know, it's not curable, but we're trying to make your last days reasonably comfortable. Which wasn't the case here, because– ALICE: And we've actually made them much more uncomfortable. TOM and LIAM: Yes. TOM: The degree of radiation injury– So this person had an autopsy, and it was reported that had she not died as a result of the cancer, the degree of radiation injury sustained would have necessitated a hip replacement procedure to restore mobility of the joint. ALICE: Just like, melted part of your pelvis or something. TOM: Yeah. LIAM: “Sorry about that.” ALICE: Yeah, you got the effect of, like, when you open a fridge door with your hips, but it's the door to the containment room at Chernobyl. TOM: Yeah. JUSTIN: Yes. TOM: So the estimated dose here was in the region of 13,000 to 17,000 rads. JUSTIN: Not very good. TOM: No. ALICE: No. TOM: So, uh, ACL carried out an investigation of the incident, and failed to reproduce the circumstances. They did, however, identify mechanical issues relating to the plunger locking mechanism and the microswitch array. And what they, sort of, AECL's investigation, they were purely looking at hardware. Because again... this is software control. ALICE: Software can't go wrong. Yeah. JUSTIN: Software can't go wrong, yeah. ALICE: It only goes wrong, we've tested it, it only goes wrong when, like, uh, fucking protons or whatever get into the case. LIAM: You idiots fuck it up. TOM: Yeah. TOM: And they kind of, what they sort of assessed was that, sort of, they got a three-bit microswitch array, so each of your microswitches can be in the sort of zero or one position. So what they modified the software to do was tolerate a one-bit error. So that, effectively you've got three bits there, so if two of them are in the correct position, it's okay. It's an introduced... ALICE: It's now immune to the speedrunning sort of error. TOM: Yeah. But it's got an additional authority to sort of track the fact that it's in motion. They issued a voluntary recall notice at this point, and this was a Class II recall from the FDA, which describes, just making sure I get everything, so I've dotted the i’s and crossed the t’s here, “a situation in which the use or exposure “to a violative product may cause temporary “or medically reversible adverse health consequences, or “where the probability of serious adverse health consequences is remote.” ALICE: So, “it would be nice to fix it,” as opposed to your Class I violation, LIAM: “But we’re not gonna!” ALICE: which is, “fix it,” sort of thing. JUSTIN: I assume this is not a situation where they physically recall the machines, because that sounds very difficult. TOM: No. They sent, basically they sort of sent engineers out to go, this is the 80s, you've gotta go and redo the code by hand. But there's only eleven sites where this is installed. So you've gotta do it eleven times. You've gotta send some guys off for a week to recode things, basically. And they audited the changes internally, and end users were told they could resume use once necessary modifications had been made. And they touted this as a safety improvement of five orders of magnitude over the previous system. ALICE: Sounds good. TOM: Yes! Number big. JUSTIN: How do you measure that? JUSTIN: [laughs] TOM: So, there is, towards the end of the Leveson document, and please allow me to scroll through... Here we go. "The evidence for the belief that radiation burn "could not have been caused by the machine "included a probabilistic risk assessment, "showing the safety had increased by five orders of magnitude "as a result of the microswitch fix." ALICE: Meanwhile, your medical physicist is there, just like, “and yes, but how did they get the radiation burns?” TOM: Yes. Yes. The belief — LIAM: “Melting Into glue,” right. ALICE: Just like, presumably being like a very, very localized cosmic gamma ray burst. Just comes in through a window or something, while they're in the waiting room. TOM: So, the belief that safety had been–this is from Leveson. “Perhaps it was based on the probability of failure of the microswitch, “which is typically 10 to the power of minus five, "AND-ed with the other interlocks.” JUSTIN: Okay. LIAM: Alright. ALICE: Sure. TOM: So, one microswitch only fails sort of, you know, 10 to the minus five times, and then they've gone, well, we've got three of these. ALICE: I mean, it makes sense. TOM: Yeah, so five–that's it. ALICE: Faultless. TOM: Yeah, we've made it sort of so that there's a bit of error, so it's five times safer. JUSTIN: Alright, case closed! TOM: Case closed! [laughter] [crosstalk] ALICE: Next episode is Chernobyl. TOM: Next slide, please. ALICE: Gonna look at a nice, older gentleman. TOM: A nice, older gentleman. This is Gordon Gord Symonds, who passed away in May of last year, and in 19- ALICE: RIP. TOM: RIP. This is the thing with this device, this thing was so... People know about it because it's got notoriety, it's kind of gone online, people have talked about it various times, but actually trying to find sort of contemporaneous pictures or anything of anyone involved, so this, this, as far as I can tell, is the same guy. I sort of found it from a local newspaper article from May of this year. ALICE: (?) love to be named “Gord”, I will say that. LIAM: We do, we do love to be named “Gord”. TOM: Another sort of medical, another sort of physicist, and this guy is the head of advanced X-ray systems at the Canadian RPB, and he was tasked with investigating the July 26th... Hamilton. He concluded, in addition to the faulty microswitch design, there were four other areas that needed to be addressed in order for the Therac-25 to pass the Canadian certification for radiation emitting devices. Most significant of these was that if there was to be a dose error detected, the machine would move immediately to treatment suspend rather than treatment pause. ALICE: Oh, so you only have to press the P key and you can hit them with the wrong dose. TOM: Yeah, so what this guy was saying is that it shouldn't let you do that. Basically, if there's an error, if anything goes wrong, it should suspend treatment. Which... this seems sensible. ALICE: Yeah, but then you'd have to get up and walk around and turn the thing off... TOM: Yes, and call the technician, and then, yeah. ALICE: You have to make a phone call? Just because I was about to give someone radiation poisoning? TOM: So, TOM: AECL did respond to the, um, to this edict from the Canadian RPB, and their response was to alter the number of times the treatment could be overridden. From five times to three times. LIAM: Beautiful. ALICE: [laughs] JUSTIN: Ah, well. Alright. ALICE: You can only get three lethal doses. JUSTIN: Compromise is how we get things done. I mean... ALICE: That's right. JUSTIN: [laughs] TOM: So, they had made changes to the microswitches in all of the devices. However, the other recommendations, including alterations to the sort of testing of the beam and the error reporting which Symonds had told AECL would need to be in place to actually receive the Canadian certification, were still pending when another incident occurred in December of 1985. LIAM: Oh no. TOM: Uh, worth noting that, just before we move on, and Justin gets to do another date, that the team at Hamilton were not happy with AECL's response, and they requested that AECL install a potentiometer system and a mechanical interlock on the treatment head. AECL refused. Again, medical physicists went, no, we're doing it anyway, and installed it themselves. JUSTIN: [laughs] ALICE: Void the warranty of your... TOM: Yeah, void the warranty and make– LIAM: Wearing an “I void warranties” t-shirt under my scrubs as I (?) this extremely expensive machine with a hammer. ALICE: [laughs] ALICE: Just like, yeah, I actually need to jailbreak my linear accelerator. LIAM: Yeah, me and John Deere farmers, you know how it is. TOM: Yes! ALICE: [laughs] That’s right! LIAM: “Right to repair,” I scream! ALICE: [laughs] JUSTIN: Stop me if I'm getting ahead of us, but, am I correct in thinking the previous version of this device had mechanical interlocks? TOM: The previous version of the device had mechanical interlocks. LIAM: You don't need ‘em, Rocz, they had (?). JUSTIN: Ah yeah. ALICE: ‘Cause this one’s software-controlled. TOM: So the previous one had hardware interlocks. One thing that people did note, which will be relevant later, is that the previous version did have a tendency to sort of blow fuses and things. It was, again, the reason that some of the operators were sort of used to the software on this being finicky is that they'd worked with the previous versions of the device, and this was like, okay, it would shut down, it would suspend, they’d have to hit the P key loads of times, and sometimes a fuse would blow, and the machine would be out of commission, someone would have to come replace the fuses, et cetera. TOM: Anyway, these were user-end errors, of course, so if anything happened like that, that didn't go anywhere. Certainly not to the FDA, because it wasn't a manufacturer report. So... the third incident, next slide. JUSTIN: December, 1985. TOM: Yeah, we don't— this is very, very sketchy, the information here, but what we do know is that following the... this was at Yakima Valley Memorial Hospital in Washington. ALICE: Yakima. TOM: Yakima. ALICE: Fantastic. LIAM: (?) come from there. That's what I know. TOM: The patient had, they were sort of on their first cycle of therapy, and they had redness of the skin around the hip area, with a sort of unusual pattern of injury resembling stripes. The staff who examined the patient were concerned this pattern matched the... I've mentioned this before, these are all blocking trays, so we're talking about a sort of lead oven tray with a circle cut out of it. They sort of were concerned that the pattern matched the blocking trays that they had. However, because they took the trays and after they've received so much radiation dose they become a radiation hazard, so they have to throw the trays away. They couldn't match it. ALICE: So it's like, the radiation dose is so strong that it has burned the outline of the tray. It's just like, scattering off. Sure. TOM: Yeah. But the... treating doctors also said, well, it could've been caused by an electrical thermal pad or blanket–The patient, someone who's got cancer, and they've got, it's achy, and they put a heat blanket on it. They put an electric blanket on it to warm up. And they sort of said– ALICE: “Yeah, it's possible that we fucked up “but on the other hand it's also possible the patient's an idiot.” TOM: Yeah, or the patient's electric blanket fucked up or something like that. You know, we don't know. LIAM: Never mind the guy has the medical equivalent of drawing a dick into your sunburned friend's skin. TOM: Yeah. ALICE: [laughs] JUSTIN: [laughs] TOM: So the staff contacted the AECL and AECL were told there was in no way in which the device or operator could have resulted in injury. ALICE: (?) We fixed it. JUSTIN: [laughs] ALICE: We fixed it, you know? TOM: And again, TOM: this device had undergone the required changes to the microswitch array. And had a five order of magnitude improvement in safety. ALICE: So it's fine. Clearly it's nothing. TOM: So this device had been upgraded in September of 1985 and had been operating for two, three months at this point without incident. The doctors described the (?) cause as unknown, but the sort of, you know, what we know of the long term of this is about two years later, as the sort of discovery process was sort of going on here, that the patient had gone on to suffer extensive skin and soft tissue injury. Non-healing ulceration requiring skin grafts, basically, to repair the skin damage. Yes. Okay. Uh, and... incident number four. ALICE: Oh, this is the real bad one. Having read that. TOM: This is the real bad one. So you know, like– ALICE: This is the radiation sort of nightmare story. TOM: Yeah, so, you know, if everyone hadn't sort of checked up at this point, this is, you know, content warning, content warning, content warning, medical radiation injury. JUSTIN: March 21st, 1986. TOM: So, uh, Voyne Ray Cox, a 40 year old gentleman attends the East Texas Cancer Center in Tyler, Texas, for post-surgical radiotherapy for a tumour on the back. It doesn't mention what the tumour is. This is, you know, we use it for melanoma, we use it for sort of larger squamous cell carcinoma, so that sort of... melanoma obviously is, you know, melanocytes, squamous cells are the sort of cells of the skin, but we tend to treat those with surgery, so this is, you know, might have been something like a melanoma, we don't know. The patient was prescribed a dose of a 22 megaelectronvolt electron beam therapy to the left upper back, and the, um, the patient at that point was to deliver a total fraction of 180 rads, with repeat doses over the next several weeks, totalling 6,000 rads. So we've already talked about 6,000 rads, you know, in one go, to your whole body. ALICE: Very bad. TOM: Very bad, very bad. ALICE: Whereas if you do it spread out on the sort of like, you know, eight radiotherapy fractions and your ninth one is free, sort of like, uh, heart-stamp thing, it's not as bad. TOM: Yeah, exactly. So... he was, this was his ninth cycle, again, like the first, this is someone who knows what to expect. They go into the room, they get sort of positioned, they sort of lie down or sit up, or sort of get into the place where they need to be to... have their therapy, the beam's sort of all set up and then off they go. So again, the operator here placed behind a heavy shielded door down a corridor. There are no windows between the treatment room and the operator suite, and the means of monitoring is a video camera, so sort of, you know, 1980s scanline quality internal video, and an intercom system. However... LIAM: [coughs] Sorry. TOM: However, on this occasion, the intercom system had broken. ALICE: Oh, great. Sort of... ...Final Destination to which they were throwing. You know. TOM: So the patient was positioned, the operator exited, closed the shielded door between the treatment suite and the control room, entered the prescription data on the terminal, on reviewing, noted she'd inadvertently typed “X” for X-ray mode. So she promptly skipped back up to the modality field, you know, where we had our little screen earlier, and changed it to “E” for electron beam. Everything else was correct. Hit the key to proceed. LIAM: Should've changed it to “G” for good. ALICE: Yeah. JUSTIN: No, no, that's the one that changes your gender. LIAM: Oh. TOM: [laughs] LIAM: They don't allow that in the UK. Fucking pussies. ALICE: [laughs] The G-rays, yeah. JUSTIN: Got hit by the gender beam. LIAM: Mmm, forced feminization. ALICE: [laughs] TOM: So on this occasion, the machine entered a treatment pause, and returned the words... Remember... the operator manual has no description of the errors. It returned the words, “malfunction 54”. ALICE: Great. Okay. Could be anything. TOM: Could be anything. The dosage report on the machine however suggested substantial underdose of radiation compared to what was prescribed. LIAM: Oh good! [laughs] JUSTIN: This is why it's always a good idea to have an OBD-II reader with you when you're doing this. ALICE: [laughs] LIAM: Yeah, never know when you're gonna need it. LIAM: You need to flash– ALICE: I brought this Geiger counter from home, and I'm holding it in between me and the fucking thing. [laughs] JUSTIN: It just flashes the check engine light. LIAM: Flash the Therac's ECU. Tuning the Therac. ALICE: [laughs] LIAM: I want the Pops and Bangs mod for my Therac-25. ALICE: [laughs] TOM: So, this machine, as we've said, this is a finicky machine. This does this all the time, and it pauses and you press the P key to override the pause and proceed with the dose as prescribed. The machine... malfunction 54. ALICE: That's weird. Gotta go press P again. TOM: Well, you would press P, ...was going to press P again, but then was alarmed by the sound of the audibly upset, disturbed patient banging on the operator room door. ALICE: Oh fuck. JUSTIN: Not good, not good. ALICE: Oh shit. You... JUSTIN: That's not what you want. TOM: Next slide, I think. Yes. So, this, again, we'll just go back to our setup here to remind ourselves of the situation here. So we’ve got... operator terminal, sealed away, intercom that's not working, patient has come off of the treatment table, and as they're coming off the treatment table, the operator presses P again. ALICE: You got hit with a fucking, like, glancing blow by this thing as well? TOM: Yes. It's sort of implied. ALICE: This is why it wouldn't be me, right, because you put me in a situation like that, I'm like, they find me a melted puddle of goo because she's done it five times, and I've been like, well, I don’t want to be a bother, I'm not gonna sit up. She did say to sit still, so, whatever. This guy clearly has just, like, flown out of the thing. TOM: Yeah. So this is, so he knows that this is not normal. But basically he sort of felt a, what he reported as a severe pain like someone had poured hot coffee on his back when the operator engaged the device for the first time, and was rising from the table to alert her when the second dose was delivered, describing severe pain in his arm, and a sensation like his hand was leaving his body. ALICE: What the fu–Jesus Christ. So he just, like, stood up, he had his arm in the way of it, and LIAM: Don’t like that. ALICE: we just accidentally sent this man's hand's atoms to fucking Jupiter. JUSTIN: [laughs] TOM: Yeah, this is horrendous. This is–Jesus Christ. ALICE: It's like, it's like, fucking orbiting the rings of Saturn, and he's just like, uh, okay. Still attached to it, y'know? TOM: So, the patient was seen by the medical team on site, who sort of noted reddening in the treatment area. As he wasn't unwell in himself, was discharged home. Because the– ALICE: For the minute, y'know? TOM: Yeah. Because they said, well, y'know, this is, this is a safe, established device, we've used this for, y'know, years. This is, y'know, there must have been another freak sort of electrical shock. JUSTIN: Like, from like a Tesla coil that's been embedded in the machine? ALICE: [laughs] Like, where is this electricity coming from? [laughs] TOM: Again, I think part of the thing is here, is that, y'know, documentation is limited, but I think sort of from a point of view of someone who has been trained at medical school, never been sort of trained, y'know, although sort of having some additional experience in radiation bits and pieces, you don't expect to see a radiation injury. What was far more, y'know, what you think is, well, this is an electrical device, electric shock's far more common, and what they've reported sounds like an electric shock. But y'know, their cardiac function is okay, they've not got an arrhythmia, and y'know, the treatment's sort of severe electric shocks, so if there's no arrhythmia, then, y'know, the patient's otherwise well. Then they sort of, then there can be discharge time (?). JUSTIN: “We know they were just in a radiation machine, but “it probably wasn't a radiation injury, right guys?” [laughs] TOM: Well no, because this is a radiation machine that's very safe, the previous versions of it were very safe, the manufacturer has just upgraded this machine six months earlier, and it is now five-fold safer than it was before the upgrade. JUSTIN: Not just five-fold, five orders of magnitude! TOM: Sorry, sorry, five orders of magnitude, safer! JUSTIN: [laughs] TOM: Than before the upgrade. So it's the kind of thing whereby, y'know, in medicine, we often talk about, if you see sort of hoof prints, expect horses, not zebras, kind of thing. I suspect that, y'know, it's very difficult to speculate on a sort of, at this point, 40 year old medical case, for which I don't have any notes. But that's... ALICE: Also, very little that they could have done. Had they immediately been able to go, ah, this guy's suffered a catastrophic radiation injury. I mean, the thing you do at that point is you fucking roll your chair over to the drawer and hand him one of those, uh, “so, you're gonna die” pamphlets. Y’know. JUSTIN: Yeah. Give the guy a cigarette. [laughs] ALICE: [laughs] Yeah. JUSTIN: “Manufacturer claims, “this device will have one accident in the lifetime of the observable universe.” JUSTIN and ALICE: [laugh] ALICE: And this was it, so we can all relax, y'know? TOM: So, tragically, over the next several months, the patient lost use of the left arm, and several bouts of nausea, vomiting, developed radiation-induced myelitis, and myelitis is nerve inflammation in the spinal cord, and his left arm became paralysed entirely, as did his vocal cords. I mean, this is just horrific. In addition, he was severely immunocompromised, and then had opportunistic herpesvirus infections of skin in the affected areas, lost bladder and bowel control, and died five months afterwards, with what was later estimated to be a dose of between 16,500 and 25,000 rads having been delivered. ALICE: Presumably not knowing why, as well. That also sort of rankles with me, is the idea that not only did you kill the guy horribly, but you killed the guy in a sort of horrible, confusing way. TOM: Yes. It's awful. I mean, as I say, this is, yeah, but combined with the sort of horror of the... intercom being sort of out, this is the real, sort of, this is the nightmare scenario for any sort of medical treatment gone wrong, really. ALICE: If you're a radiologist in that field, it's all like, hearing the guy sort of bang on the inside of the door? That's got to be the worst thing that you can imagine happening. TOM: I say, from what it sounds like, it's not recorded who the operator was, and I think for the purposes of, you know, that their anonymity has been maintained, and... given the role they have to play later, which is fairly crucial, then I think it's quite, you know, they were doing their job as they were routinely expected to do their job. They were working with the machine within the parameters they'd been told to operate the machine, and... both from a patient and a clinician perspective, this is absolutely the... worst case scenario that could ever happen in medicine. Because, you know, there is... varying degrees to which there is a sort of Hippocratic Oath as such, but first do no harm, to a greater extent, remains the foundational principle of medicine. In the case of the Therac-25 device here, this was used to successfully treat other patients later that day. JUSTIN: Ah. TOM: But taken out of service the following day for testing, they flew an AECL engineer down from Canada to try and reproduce the “malfunction 54” error. ALICE: Did we ever find out what the malfunction 54 error is? TOM: Malfunction 54 is very helpful. It tells the operator that the patient has either received too little, or too much radiation. JUSTIN: Ah. ALICE: [laughs] ALICE: Great, thanks. JUSTIN: Okay. LIAM: That really narrows it down. Thank God. JUSTIN: I was about to say, that... eliminates one number. ALICE: Reminds me that... a lot of NASA software had to develop different states for off-scale high versus off-scale low. Because if you just say, this is not a reportable—this is not a number I can return for this query, it leads you to this sort of ambiguity. TOM: Yeah. So AECL basically, still at this point, flatly denied it was possible for the device to overdose a patient. The electrical systems were checked, we're still functioning on... this is an electrical system error, but the machine was determined to be properly grounded, and could not have delivered an electrical shock to the patient. And the machine was returned to service a few weeks later, on the 7th of April, 1986. ALICE: I mean, it's still busily, you know, treating cancers. TOM: Yeah, it's still treating cancers. ALICE: Presumably successfully, so. LIAM: Yeah. JUSTIN: The guy didn't die of cancer. Um. ALICE and LIAM: [laugh] LIAM: Yeah, he just got a Chernobyl dropped on his face, but it's fine. TOM: Yes. However. JUSTIN: April 11th, 1986. TOM: So, four days after the device has been returned to service, another patient, Verdon Kidd, attended for cancer treatment on the face. We have very limited other clinical information. The same operator that had treated the previous patient set up the machine for therapy. ALICE: Great. What's the worst part of your body you can imagine getting this malfunctioning thing fired at? LIAM: Taint. Taint. ALICE: No, it's face. It's face. LIAM: Face or taint, whichever. TOM: So again, this is one of the more extensively documented incidents, for better in terms of this, it helped to uncover what was going on, but worse in terms of, again, a heavy content warning for a medically sustained radiation injury. The same operator as previously, as well, having set this patient up for treatment, the intercom had been fixed. ALICE: [laughs] Yeah, great, okay. There was one saving grace of the system, the intercom had been fixed. And as it had been previously, she entered "X" for X-ray before proceeding to treat, noticed her inputs, went back, corrected this to "E" for electron beam mode, commenced the treatment. Over the now repaired intercom, she heard a loud noise from the machine, with the terminal once again displaying “malfunction 54”. Also audible, the sounds of a distressed patient moaning for help. The operator entered the room, helped, see, this was on the face, so you get the person in position, and then you have to, for a lot of therapeutic and diagnostic procedures, sometimes you have to keep people in position, so sometimes you have to lightly tape people in position. ALICE: Use some light bondage, yeah. TOM: Yeah. For medical reasons. LIAM: Yeah Alice, for medical reasons. ALICE and JUSTIN: [laugh] ALICE: I'm sorry if that appeared prurient of me, yeah. TOM: He complained feeling fire on the side of his face, and the medical physicist attended– LIAM: Ooh, fireplay. TOM: So, having had this ongoing investigation for the last several weeks, the medical physicist was very interested in seeing what had gone on here, and got a statement from the patient. And the statement was that he saw an intense flash of light. JUSTIN: Oh, that’s bad. LIAM: No, you don't wanna see that! JUSTIN: Don't wanna see that! ALICE: No... [laughs] LIAM: Never wanna see that! Look at the pretty blue–what was it? Look at the pretty blue trolley, not the pretty blue light. JUSTIN: Yes. TOM: Yeah. Describing the sound like that of frying eggs. LIAM: Ohhhhh. TOM: Leveson, Nancy Leveson, has further documentation on this, and she reports in her appendix that he repeatedly was asking “what happened to me” and was extremely distressed. ALICE: Yeah, I bet, Jesus. TOM: Yeah. LIAM: I’d be distressed too. TOM: Unfortunately, what happened to him was radiation encephalopathy. LIAM: Oh, Jesus. TOM: So he became increasingly delirious and confused, entering a coma; because they were onto, something is wrong with this machine, they had an autopsy performed, demonstrating acute, high-dose radiation injury to the temporal lobe. And this gentleman, Verdon Kidd, has the dubious honour of being the first person documented to have died due to therapeutic radiation therapy. ALICE: Wow. I mean, horribly, too. JUSTIN: Yeah, pretty bad. TOM: So, the, the, the... So, as we mentioned, this is another medical physicist. Next slide, please. There is a (?). LIAM: A (?), as people get wildly drunk in my home. I'm not sure if you can hear that at home, viewers, but god damn. JUSTIN: You might be able to. I can hear it. LIAM: Can’t do anything about it. TOM: So, this time, this is the Tyler East Medical Center, the Tyler East Cancer Center's Fritz Hager, who, again, I've tried to find pictures of people, and I searched for Fritz Hager in Tyler, Texas, and returned a picture of this young gentleman, who is American Idol's top five contestant, Fritz Hager III. So... ALICE: Grandson? Question mark? TOM: Grandson, question, question, question mark. Unless there is an abundance of Fritz Hagers in Tyler, Texas, and specifically Tyler, Texas... JUSTIN: Lots of Germans. In Texas. ALICE: That's true. That is true. LIAM: And some Czechs. TOM: Potentially some Malaysians, As far as I can establish... ALICE: Well, I'm gonna be picturing the guy, the radiation... TOM: Yeah, I mean, given the amount of documentation that this guy did, into what went on with the... this is the guy... largely responsible for documenting what one of the major flaws of the system was. And he and the operator who'd been involved in both cases, pulled.. you know, it's not specifically sort of said in the documentation, but you get the impression that all-nighters were pulled, trying to replicate the error. JUSTIN: Right, yeah. ALICE: I have a small, perhaps nonsensical thing, which is I hope it was the operator's unwitting fault, because that seems less cruel than it being random chance, both times. TOM: The key thing here is that this is a very experienced operator. So they have been using... the Therac-25, and presumably possibly previous versions of the machine, or other similar machines. So they were very, very quick... you know, you imagine someone who's using the system day in, day out, to treat 20 patients per day, 100 patients in a week. You know, your hours on the machine build up to the fact that you can, you know, you rapidly edit stuff, and you go, oh, okay, fine, how many times a day do you sort of accidentally typo something, oh, this is an electron beam, this is an X-ray, this is an electron beam. Quickly correct the error, off we go. So they recreated the error, basically, by... achieving this, by rapidly editing the X-ray modality to electron beam in the Dose Setup screen. And because they were so quick at entering it and making the corrections, they could, both the operator and the physicist, could reproduce the error at will. The AECL... ALICE: So, it's a timing problem. TOM: It's a timing problem. So the AECL engineer the following day was unable to reproduce the error. ALICE: He's like, hunting and pecking on the keyboard. TOM: Yeah, so as Leveson says, the ETCC physicist explained the procedure had to be performed quite rapidly. Eventually, someone at AECL found a fast enough typist to reproduce the error on a test machine, where engineers reported a centre of field dose of around 25,000 rads. LIAM: Jesus. TOM: 250 grays. The frying sound was determined to be... so, a while ago now, we mentioned that each of the treatment heads has an ion chamber which records the dose at the exit point. And these were completely saturated. So it's... ALICE: Fuck. TOM: So the frying sound, at least for some respite, the frying sound was not patient-related. ALICE: No, just the burning dosimeter. TOM: Yeah, burning dosi–the completely saturated ion chambers, every single ion in them saturated with a lot of radiation. In litigation arising– ALICE: The ion detector just returns, “yes”. LIAM: “Whoops!” [laughs] ALICE: “Yes, ion present”. LIAM: This just says, “liable”. TOM: [sighs] ALICE: [laughs] TOM: So, in one of the cases, and I think it was... incident number four, there was already a litigation arising from that. And AECL representatives admitted they had been aware of the cursor up problem in other centres over a year earlier, after previous incidents, but thought the software issues had been fixed. ALICE: How many previous incidents aren't there (?) here? TOM: Yes. That's the thing. This is a device that's gonna amuse every–we, y'know, there are six incidents that we're aware of. We are– ALICE: I mean, I mean, it's hard to hide radiation injuries, but on the other hand, with people who are very sick already, that's... TOM: Yeah. TOM: That's one of the things that I think confounds some of the issues, that you know, people who are very, very sick already with cancer, but some of these people, I think, and the issues where this is kind of picked up, are where people kind of had cancers that were, y'know, maybe less severe, or, y'know, but then had horrific consequences from their therapy. So we mentioned that the FDA had sort of been involved at this point. At this point, the FDA really gets involved. So these two incidents, in rapid succession, reported by the Texas Health Department to the FDA anyway, despite the fact that there's not a reporting mechanism, they sort of just kept calling the FDA until they sort of created one. ALICE: Someone answered them. TOM: Yes. So, AECL's initial response, that they made three days after the second incident, was that all current device users were to refrain from using the up arrow to edit treatment functions. ALICE: [laughs] JUSTIN: [laughs] LIAM: [laughs] Okay. LIAM: Alright, I like this. I like this BSDS(?) solution. TOM: No, even better. So that no one accidentally presses the up arrow, the up arrow was required to be removed from the terminal. ALICE: Ah. Yes! Now we're getting serious about access control. If you don't want people to use the button, take the button off the machine and lock it in a cupboard somewhere. TOM: Yeah. Take the button off the keyboard, tape the key microswitch over so it's in the open position, and that's only able to re-enter the prescription information from scratch in the case of an error. LIAM: Oh, she's beautiful. [laughs] TOM: So far, in terms of the corrective actions issued (?), we've sort of messed around with the microswitches, and we've got a screwdriver out to lever the key off the keyboard. ALICE: This is, what, a how many million dollar machine here? TOM: Do you know what, the one thing I didn't look up was how much this thing cost. I probably should have done that. But, y'know, a lots. LIAM: Lots. TOM: A lots of money. ALICE: A state-of-the-art piece of medical equipment. TOM: State-of-the-art, computer-controlled... JUSTIN: Hold on, hold on, I got it, but, uh, Price is Right rules, what do you got? ALICE: Two million dollars. LIAM: Are we playing with inflation or without? JUSTIN: Without. LIAM: Uh, half a million. TOM: I'm gonna go with 750,000. Yeah. JUSTIN: Doctor Tom's got it, it's a million. LIAM: Fuck! [laughter] ALICE: I went too high, I don't know how much these things cost. Quite literally, a million dollars. LIAM: Damn it, dude. JUSTIN: [laughs] TOM: So, the FDA were unsatisfied with this response. ALICE: Everyone's a critic. TOM: ...declares the Therac-25 defective on May 2nd, 1986, requiring that the AECL produce a corrective action plan. And then shortly after this, we enter, because this is May, and then in medicine, then June through August, before the new... term of medical school tends to kick out, and medical training tends to kick out, and before the new input comes in October, this is conference season! ALICE: Ah! TOM: Yes! ALICE: Well, lots to talk about, y'know. TOM: Yes. So, this has been installed at 11 sites, and a users group convenes at the American Association of Physicists in Medicine. So... ALICE: To ask some questions, like, “why does this kill my patients?” JUSTIN: Yes! TOM: So, this is where our previous Tim Still from Hamilton, and Fritz Hager and other operators, could all talk to each other about the ways in which the Therac-25 seems to be going wrong, and the safety modifications they've been forced to make, such as retrofitting their own safety hardware interlock, and bully the attending AECL representative. JUSTIN: [laughs] ALICE: Oh, that guy's gonna have an interesting time at that conference. Being put the question, why has my colleague had one of the worst experiences you can have as a medical professional, and inadvertently killed a patient horribly, due to your company's product for which we have paid one million dollars. LIAM: How would you like to be beat to death with a shovel? TOM: I mean, TOM: Alice is gonna be somewhat familiar with this, but there is a way of documenting things in medical legal note-keeping. ALICE: The question was asked in a sort of robust tone. “I put it to him, that he had assisted in the murder of my patient.” TOM: So, this is meeting notes that were recorded and then recounted in Leveson's appendix, which was... “There was a general complaint by all users “about the lack of information propagation. “The users were not happy about receiving incomplete information. JUSTIN: [laughs] TOM: “The AECL representative countered by stating that AECL “does not wish to spread rumours, “and that the AECL has no policy to, quote, ‘keep things quiet’, unquote. LIAM: You sound like you don't, bud. TOM: “The consensus among the users is that improvement is necessary.” ALICE: Yeah. I bet. I kinda get the sense of that consensus. JUSTIN: "Millimeters from an outright brawl here" is what I'm getting. ALICE: Yeah. LIAM: With the chair! Get him with the chair! JUSTIN: [laughs] TOM: So, not only sort of content with bullying the AECL representative in person, they put together a zine! ALICE: Oh, well, it was the 80s, you know? Just anyone could do it. TOM: Yeah. TOM: So, there were two issues of the Therac-25 user group zine, the first of which contained commentary by Tim Still, the physicist at Kennestone where the first incident occurred, outlining eight significant issues he felt were present in the Therac-25 software systems. Second issue expanded on just one of those, a function relating to the head position microswitches further, and the problems that potentially arose. Again, I would love to find this. Again, this is something that has almost certainly lost to time because there are... This was at eleven centres. ALICE: Mimeographed, yeah. TOM: Yeah, sort of eleven centres of which half the people involved are sort of probably dead. And this is... There may be a copy in someone's attic somewhere, but if we ever see it, that would be amazing. But not in the scope of this recording. Next slide, please. ALICE: What actually went wrong with software? TOM: So, what... So. TOM: The Therac-25 software was developed using the Therac-6 and the Therac-20 operating software as a base, entirely proprietary system, created in assembly language of the PDP-11, and as far as can be established, remember we parked the phrase “people”? ALICE: “People”, yes. TOM: As far as can be established was written by a single anonymous programmer. ALICE: “Person”! This is something for the software person. TOM: The software person. JUSTIN: Yes. In assembly? TOM: In assembly. JUSTIN: So it’s Chris Sawyer. LIAM: Yes, Chris Sawyer, I was gonna make that joke, goddammit. JUSTIN and ALICE: [laugh] ALICE: Programming was different back in those days, you could do the Margaret Hamilton thing of, you know, Chris Sawyer programmed all 11,000 lines of code necessary to horribly kill a patient by hand. LIAM: [laughs] TOM: So, when the initial X is entered, the software starts setting up the electron gun in readiness to deliver the, we've mentioned this before, when you put in an X, it defaults to 25 megaelectronvolts of X-ray photons. ALICE: And then it has to rotate the head of the machine into the dough hook position for X-ray. TOM: Yeah, and also set up the magnets inside to bend the beam around to hit the target. During the magnet setup process, the software clears all of the internal data several times. So if you make changes to the prescribed treatment modality within that approximately eight-second window, it displays the corrected prescription information on the operator screen. However, the software has configured the magnets within the linear accelerator to produce a particle accelerator beam at the intensity intended for X-ray generation, not electron beam therapy. ALICE: Oh no. LIAM: I don't like that. TOM: So, they dug out the software for the Therac-6 and the Therac-20, and it couldn't produce the same error. But because of the hardware interlocks, then they didn't have any incidents, because the beam didn't fire or a fuse would blow. ALICE: This is... not something that comes up until you make it purely software, and then it's firing an electron beam at the intensity required for an X-ray beam which is gonna draw off 99% of that power anyway... TOM: Yes. But it's– ALICE: Gotcha. TOM: Yes. It's not doing that. It's firing it through the electron window. JUSTIN: I'm surprised it didn't just straight up explode! LIAM: No. ALICE: It's a testament to particle accelerators, right, as a science, that like, hey, you can make one of these, it'll do it, no problems, the machine works fine. TOM: Yeah. LIAM: I regret to inform you that I have to drop off at this point, it being 8 o'clock. TOM: Sorry, Liam. ALICE: Okay. LIAM: No worries. Yeah, I'll just keep the window open. ALICE: Yeah, okay. Tom, are you good on time, or... do you wanna record this in two parts, what do we feel like? Just to feel flagged(?), I don't wanna keep you... for 3 hours. TOM: I mean, I've not got many more pages left of my spiel here. We're kind of in the home stretch. JUSTIN: We've got seven more slides, and then we've got Safety Third, which we can skip. ALICE: Okay. LIAM: Have a good night, everybody. Thank you. ALICE: Alright, goodnight, Liam. TOM: Thank you, Liam. So, the investigators noted that basically the speed at which the experienced operator can make the changes to the prescription information, um, could– ALICE: I mean, that's eight seconds. TOM: Yeah. ALICE: To hit an up key and an E key is not that much. TOM: No, but it would sort of, you know, they've gotta skip through all of the previous lines up to the thing. And this is someone who's very, very experienced, and can just be like, [clicking noises] click. Back down to the bottom line, go. So, they– basically, the speed at which they can operate, um, meant that you could make the changes within that eight second window, but also because they're so quick at operating this, it's kind of, you know, the faster you type, the more errors you may introduce. So, and because the system is so finicky, they're used to operating with their fingers going, “it's paused again.” ALICE: This feels like one of those cultural things where it's like, I remember reading something about, um, it was a fighter pilot's memoir from the Second World War, where his friend's like a, not... he can't quite get to grips with the idea of, like, dogfighting. His brain doesn't work that way. What he's good at is, like, flying very straight and level, despite distractions. And what they eventually tell him is, you know, go and be a bomber pilot. Go and do that instead. At which he's very successful. Um... And it's like, it feels like this is one of those applications where what you want is sort of like extreme consistency and like thoughtfulness rather than like speed of action. Um... But if you've got, you know, 20 patients to see that day, then it doesn't, you know, different sets of incentives, you know? TOM: Yes. Yeah. Uh, exactly. This is, I mean, very relevant to sort of the current state of healthcare in the UK, but that's probably for another time. Um... ALICE: I imagine much harder to be sort of that kind of very deliberative personality. TOM: Yeah. You've got to be, you've, you've got to be on it the whole time and the amount of time you've got to be on it is rapidly increasing. But you know, that's, that's, that is the way it is. Um, so, uh, I've got a hard mode version of what exactly is going on here. This is, this is ripped. This is sort of taken straight from Leveson. So next slide, please. ALICE: What is wrong with computer brackets, (Alice will not understand). TOM: Yes. I mean, I would say this is not my wheelhouse and this is me sort of looking through this, reading this several times and then sort of trying to sort of approximate it from my own sort of knowledge. But, so you have the monitor task, which is called “treat”, and this has eight subroutines, “datent” being the data entry task. And it communicates with the keyboard handler task. That's the bit that tells you what keys you put in And that shares a variable in the data entry, uh, uh, called data entry complete. When the prescription data is entered, the flags change to denote this. And when it's set, the treat subroutine moved to variable three, which is “set up test”. Exit it, exits that, and then restarts the “treat” subroutine. In the original version of the software, to determine whether the data entry was flagged as complete, depending on the position of the cursor on the command line at the bottom of the terminal screen, but not when it had been moved subsequently. ALICE: So it just, it's not intended for you to edit this again. If it's like, uh, what the programmer has in mind is you, you put in the prescription details and you get it right. If you get it wrong, you ditch the whole thing, start process from like beginning. TOM: Yes. And then it's not entered again. If it's been flagged as complete. It would usually force the technician to enter the prescription data except when X-ray photon mode is selected. So this is why the eight seconds is quite quick. They've got to go through and put every bit of prescription information in. Uh, if it's electron beam. If it's actually–it just defaults to 25 megaelectronvolts and skips back down to the bottom. So this is the portion of the software that sets another task in motion called “hand”, which sets the position of the treatment head. But because the data entry has not been exited and won't be entered again during the cycle, there's no way in the software to detect the incompatibility between the treatment information entered and the treatment head setting. So then the final bit of this software is called a subroutine called “magnet”. And that sets the, that's where the eight second delay is relevant. So this has an artificial delay introduced to account for the time it sets all of those electron magnets up to deliver the beam. And that's set up using a routine called a “Ptime”. It's cycled through several times and checked with the keyboard handler to see if there are any edits made. Of course, way it executes, because it clears that flag, then it never detects that there's been an edit made and defaults back to the data entry. So giving you the eight second window, ALICE: It’s like watching a gun being assembled, it's horrible. TOM: Leaves an eight second window. It's possible to edit the modality and dose, have it be displayed correctly on the terminal screen, not be recognized by the software, and then resulting in the two overdose incidents that were at East Texas. [sighs] ALICE: That's such a, like, easy thing to miss. JUSTIN: No, it's software, it's perfect. ALICE: [laughs] TOM: So, the radiation scientists, they are continually bullying AECL throughout all of this. The FDA are similarly scathing with internal memoran– Leveson managed to get hold of some internal memoranda for her report and indicated that “they possibly emitted awareness of previous radiation safety incidents “involving the Therac-25, having settled out of court “the injuries sustained by Katie Yarbrough,” the first victim in Marietta, Georgia. “Insufficient detail and clarity of their quality assurance process, “with no response provided to the Canadiaration”–”Radiation Protection Board “after they requested further details of software revision and requirements.” So this is... Gord. So they didn't reply. So they... JUSTIN: Did you just say... ALICE: ...(?) Gord's emails. JUSTIN: Did you just say “Canadiation”, yeah. ALICE: “Canadiation”, yeah. JUSTIN: [laughs] ALICE: That's what they should call it. That's like... There's always a background level of Canadians around, right? JUSTIN: Yes. ALICE: But then above that, there's like an area, a point at which you might stochastically be exposed to a Canadian, in a way that increases your risk of consuming, like, maple syrup, or like... JUSTIN: Saying “eh”. ALICE: Yeah. TOM: Yes. So, they ignored, they basically, because they settled the first incident out of court, they didn't... They were settled out of court. You can't, you know... We settled this out of court. ALICE: “You’re not allowed to get mad at us.” TOM: “You can’t get mad at us. We settled.” AECL intended to keep the obscure and undocumented error codes, which the FDA... ALICE: They liked them. ALICE: I'm not sure why, but they did. TOM: The FDA demanded they change to provide more explicit warnings. And then they did not plan to do any additional QA to ensure exact copying of the updated software to the installed devices, or test any further updates to software to ensure safety. ALICE: Just remarkably, like, lackadaisical about this. JUSTIN: “It’s fine, don't worry about it.” TOM: So a revised corrective action plan was published on December 22nd, 1986. This implemented suspension of treatment if a dose error was detected, and a plan to implement meaningful error messages. ALICE: Oh. Good. We're gonna make a plan... TOM: ...Make a plan to have a committee to sort of work out what is meaningful in the context of an error message. ALICE: What was so difficult about error messages at this point? I mean, I guess I understand that now, in an age of everything where it's like, “oopsie whoopsie, the computer made a fucky-wucky,” you might not want to have to delve back through error in block 0x whatever, but like... this is so... it's such a weird point to be insistent about. JUSTIN: Why not give the operators a printout of the error code documentation? I feel like that's a nice, quick interim solution here. TOM: Yeah, this was the revised plan, and everything was fine at this point. All the major safety issues had been addressed, and the federal oversight had been successful. ALICE: Please don't say a date now— JUSTIN: January 17th, 1987. ALICE: Fuck. JUSTIN: [laughs] TOM: Yes, this has a fake ending. The Therac-25... ALICE: It comes back from the dead? To kill someone else? TOM: It was not done yet. This is... so, on this case, an unnamed patient attends the Yakima Valley Memorial Hospital for treatment of a carcinoma. Again, no further information about the cancer itself. Second patient of the day, and they are intended to have two film exposures, so basically they do a... low-dose X-ray with a film underneath the patient to say, have we got the right place? And then a high-dose X-ray treatment, with an intended total dose of 86 rads. It's a very low dose. So this is the 80s, and so taking an X-ray requires you actually placing a physical X-ray photographic film under the patient, perform the first... ALICE: Still does in some places. TOM: Still does in some places. ALICE: You can get, like, Russian, like, well, used to be able to get Russian X-ray film relatively readily, that you can make some very interesting film exposures with, because it doesn't have a halation layer, or whatever it is. TOM: So they did the first, returned to the treatment room, set the beam for the second, upon returning they sort of set the beam up for the second X-ray, did the second X-ray, rotated the treatment head into the field light position, to check the positioning for therapeutic treatment. And at this point they inadvertently left the second X-ray—the second... exposed film under the patient. You know, these things happen, you know, you're treating sort of multiple patients, you've kind of got two films, you've taken one film, they left the X-ray film under. So we, so we have the hand command on the terminal, in the operator room, and you can also position the treatment head using a hand control as well. So in this case they set the dose using the handheld unit, went back to the terminal room, displayed "beam ready", having entered all the prescription information, the operator pressed the "B" key for beam ready. I get it, "B" for beam, "P" for pause, "E" for, you know... ALICE: Literalists. TOM: Literalists. ALICE: Canadian literalists, TOM: Canadian literalists, all the way through. The beam came on, but no dose information was displayed, the operator in deposition stated the error message may briefly have been displayed, however the treatment was simply paused. And as the operators are used to doing at this point, they pressed the P key to resume treatment. The machine paused again, and the–so we have the upgraded error messages here. This one was "flatness". ALICE: [laughs] What? What could that possibly mean? And we have, in like the past couple of hours, have we heard of anything that you could express in terms of, like, topographical flatness? Like... TOM: I think this is something to do with, so we sort of talk about beam flattening in... I suspected something to do with that. But again, this is the supposedly upgraded error system, and this is the sort of errors that we're getting. But the thing is, the machine only displayed dosing information for two X-ray films. 7 rads. The operator went to unpause the treatment for a third time, but heard the patient moaning from the treatment room over the intercom and went to check on them. They reported a sensation of pain and burning in the chest, and over the next several days that developed into a burn with the characteristic striped pattern, matching slots in the blocking tray. Very similar to that seen at the previous Yakima Valley incident in December of 1985. So sort of a year or so before. AECL initially recommended users should manually confirm the position of the treatment. Again, “this is a you problem, not an us problem”, before positioning of the treatment. An engineer was dispatched from Ottawa, in the meantime medical physicists at Yakima Valley conducted their own investigations, and they managed to achieve production of X-ray beams in the field light position. ALICE: Oh good. So there's nothing attenuating that. TOM: No, so there's no electron beam controller, there is no tungsten, this is just coming through the... ALICE: The diagnostic light. TOM: The diagnostic light. So they tested this onto X-ray film, and they compared it with... Again, another fantastic bit of... Fortunately they actually left the X-ray film under the patient, because they were able to compare the X-ray film that inadvertently left under the patient affected, with the simulation films that they ran. And they noticed a match in the patterns displayed on the two. So again, another huge win here for medical physicists knowing what they're doing. ALICE: Radiation detective work. TOM: Yes. There's some fantastic detective work here. The patient was determined to have delivered a dose up to 10,000 rads, and they died in April of ‘87, relating to complications from the radiation exposure. While they had advanced cancer, the family brought a lawsuit again to AECL that indicated that the radiation exposure had hastened their death, and again, this was settled out of court. JUSTIN: Naturally. TOM: The AECL determined this could not be entirely due to hardware error, and determined the unit was awaiting an update to the software following the conclusion of the FDA corrective actions. So it hadn't had yet the software update from the East Texas instance. However, this fault was an entirely separate quirk of the operator's software. JUSTIN: Incredible. ALICE: A third one! Third fault! TOM: —Second, second. ALICE: Okay. TOM: ...because they think that the Hamilton one was probably the same thing as well, because in Hamilton, Ontario, that's where the patient had the line pattern. So they suspect the similar thing may have happened, but that was never determined. So, once again, we're going back to this diagram, it determines the position by the microswitches. They'd already had the hardware revision following the incident in Hamilton. They made modifications to the hardware, software was unchanged, and a different failure mechanism was the culprit here. So this is, again, I think we're onto hard mode now. Because what–next slide please– happens here is, what goes wrong in computer 2, more hard mode. The software flaw here is that, is what's referred to as a race condition error. So there's multiple competing processes occurring at the same time. Position of– ALICE: Different timing problems. TOM: A different timing problem. ALICE: It's all time-related, it seems. TOM: Yeah. TOM: This one relates to the position of the head, matching that in the prescription data, and it's got a housekeeping process called lmtchk, lmt check, which sets up a single bit in one shared variable called F$mal, and if this is set to zero, the code enters the setup done stage of the treat algorithm. It does this by working through an 8-bit counter called Class3, it increases by one each time the setup test portion of Tphase runs. This is going through multiple setup runs, as we talked about earlier. It's an 8-bit counter, so the maximum it can count to is from zero to 255. So if someone were, coincidentally, to press the set button on the hand control, at the exact moment that variable rolls over again to zero, the software wouldn't enter the head position check mode, and would instead display that the beam was ready to fire. ALICE: That's so weird. And it's purely dumb luck that you have to do this simultaneously. TOM: Yes. You have to do this simultaneously, which is why this machine was used– It happened on two, possibly three occasions out of what is likely hundreds and hundreds of uses, just because someone happened to hit the button at exactly the right moment at which that bit flips over from 255 to zero. JUSTIN: I mean, statistically though, statistically though, that's like 0.3% of times, which is a lot. That's a lot of times for the software to run in such a way that it kills the patient. TOM: Yeah. So... I think we're post three hours on this point now. I think we can sort of wrap things up a little bit. There was a back and forth with the FDA, essentially at this point. Multiple lawsuits that were settled out of court, which is why we have very little information. The extensive hardware changes were made to the device. ALICE: But it's too expensive to just throw them all out. I presume, you have to find a way to fix them and keep them running. TOM: Yeah, exactly. So they had to make multiple sort of alterations. What they basically did was... retrofit this machine back to being... essentially the Therac-20. So, what they had to do was they introduced hardware interlocks on the therapy head positioning system. They implemented a turntable potentiometer to independently monitor the therapy head position. I think we're on next slide if we haven't got it–yes, we have. Okay, good. Great. The other thing that had to be implemented was a dead man switch in the operator room. So to actually make the thing work, you had to have your foot on the pedal. Otherwise it would not fire. A hardware shutdown circuit was introduced. Independent hardware control(?) shutdown circuit. A hardware interlock on the bending magnet, so it ensured that it can't fire when the tungsten... ...until it’s in the correct position. Limiting the editing keys to cursor up, backspace, and return. And, hardware interlocks again, from preventing the beam from turning on in the field. Hey, but you can put your up arrow key back on the keyboard. TOM: Dig it out of the cupboard, and you can take the tape off, and you can stick the up arrow back on. JUSTIN: [laughs] TOM: They eliminated the software errors in the East Texas, Yakima Valley instance. They changed the error messages to be meaningful, although no description has been made of what “meaningful” means in the circumstance. They had an additional software shutdown mechanism. And what were described by Leveson as 23 other changes to improve reliability of the software. They finally, after six instances, acknowledged that any errors resulting in a treatment pause would instead suspend the treatment. And they prevent– ALICE: Sometimes you've got to inconvenience your user a little bit more. TOM: Yes. And prevented on-site copying of code, so all the code... they had to centralise the code so that all the updates were applied simultaneously. What Ed Miller, the Director of the Division of Standards Enforcement at the Center for Devices and Radiological Health at the FDA wrote in 1987 was that the “FDA has performed extensive review “of the Therac-25 software and hardware safety systems. “We cannot say with absolute certainty that all software problems “that might result in improper dose have been found and eliminated.” ALICE: Oh cool, okay. TOM: “However, we are confident that the hardware and software safety features recently added “will prevent future catastrophic consequences of failure.” That was... ALICE: “Inshallah?” TOM: “Inshallah”. JUSTIN: Yeah. TOM: Uh, no further Therac-25 incidents were reported. ALICE: 30 years, with everyone watching it like a hawk, these things presumably are now at the end of their lives anyway. TOM: Yeah, I mean, the likelihood is these were probably all, again, these cost a million dollars each, but then... in terms of litigation costs, that probably reaches a point whereby it's cheaper to replace it with a unit that doesn't cause catastrophic radiation injury rather than... JUSTIN: We always gonna have... one or two patients that are gonna walk in a room and they're gonna see “Therac-25” printed on the unit and they're gonna be like, nope, nope. ALICE: Oh yeah. ALICE: [laughs] JUSTIN: I want a second opinion. ALICE: No, no thank you. JUSTIN and ALICE: [laugh] ALICE: I mean, the good news is that this, like, is now enough of a case study for, like, software engineering. I would hope, I would hope and pray, right, that this leads to sort of like, a change in mindset of how you write software for applications of something that can kill a person. TOM: And basically, the software was largely on... all of the tests were on the hardware. So they sort of wrote the software, but the errors weren't documented– the error conditions weren't documented, they didn't go through the same QA procedures as the hardware, and they, this, one of Leveson's critiques of the software is that it's massively over-engineered. So it's too complex for what it's supposed to be doing. What it's supposed to be doing is set up the electron beam, change the mode, fire the beam. And what it's trying to do is kind of keep juggling sort of multiple sort of variables that are sort of calling back on each other, and it's, and sort of, you know, you're creating the conditions for which this can fail by introducing these sort of race condition errors, because you're relying on the fact that people don't do weird things with software, which people do all the time. We talked about speedrunning earlier, and effectively speedrunning is sort of ways of just like, you know, very... reproducibly breaking software in accidental ways that are not supposed to happen. And that's sort of, kind of the thing that sort of happens here, in that these, you know, these errors occurred because, as is sort of the case, let's say, with speedrunning, that, you know, okay, if you are an average person, if you're the average engineer who came in to sort of inspect the device in East Texas, they couldn't reproduce the error. You know, the person who could reproduce the error first was the operator, and they could really reproduce the error. And then the medical physicist, and then eventually the engineer, having sort of been shown how to do it, but it sort of takes a lot of practice. ALICE: Once they believe them. TOM: Once they believe them. ALICE: Which it seems they're very reluctant to do, as well. It's just like, refusal to actually trust the operator of your product when they say that it has a failure condition, that they didn’t anticipate. TOM: And let's face it, every medical physicist that was involved, and every radiation protection officer that was involved with these incidents, cottoned on to what was going on. And it was really a sort of race to reproduce the errors before, you know, there was legal consequences or sort of, you know, things were sort of squirreled away because things were settled out of court. And it's only thanks to the extensive documentation that was made by these people that we've got any idea of what went on in the first place, because the existence of this thing has been... scrubbed, it has been scrubbed from existence. ALICE: Barely has a Wikipedia article. ..it has like three sections, of which it's sort of presented as a mild case study. TOM: I think– JUSTIN: [stammers] TOM: Sorry. JUSTIN: ...This incident always brought up like a– at least something I thought it did, which apparently it didn't– bring up the idea of professional licensing for software engineers, right? Where you work on like a really safety critical system like this. You know, if you're a civil engineer and you pulled something that did this, you would go to jail. ALICE: Sure. JUSTIN: But if you're a software engineer, it's like, okay, we made this very complex system that we said is 100% reliable, and it results in several deaths. Well, you know, whatever. And... I thought after this, you started to get like a movement for... professional licensing for software engineers who work in these sorts of safety critical, but otherwise relatively simple systems. This has not been the case. I think in total there have been 44 software PE licenses issued in the entire United States. So it's not common, it's still, you have... And I guess that's how you can get away with things like Tesla Autopilot, or the... self-driving, full self-driving stuff, where the car homes in on kids and kills them. And no one experiences any liability from this. TOM: Yeah. ALICE: I think it's easy to generalize, especially like, we've seen it here, where it's like, it's one person, one unnamed person, becoming “computer people”. Because it's like, a lot of people do not understand software engineering. I certainly don't. And therefore it becomes the province of “people”, the people who do that. And... that's their responsibility. It just comes out of the computer somewhere. JUSTIN: If they were a structure, there would be a stamp with a name on it and a number. ALICE: Yes. But no one ever wants to do more regulation apart from us. TOM: One thing that, in terms of this, and this is sort of an expanded discussion that can go on, but what Leveson says in her critique of everything is that “it often takes an accident to alert people to the dangers involved in technology.” I'm quoting directly here. “A medical physicist wrote about the Therac-25 incidents that, “‘in the past decade or two, “‘the medical accelerator industry has become a little complacent about safety. “‘We've assumed that manufacturers have all kinds of safety design experience “‘since they've been in the business a long time. “‘We know that there are many safety codes, guides and regulations to guide them, “‘and we have been reassured by the hitherto excellent record of these machines. “‘And except for a few instances in the 1960s, (e.g. at Hammersmith, Hamburg), “‘the use of medical accelerators has been remarkably free of serious “‘radiation accidents until now. “‘Perhaps though we've been spoiled by this success.’” And the comment that this problem seems to be common across all fields. So yeah, the sort of complacency about, you know, these previous devices were fine, and... the software people know what they're doing wasn't the thing. And just a brief thing on this, so the next slide is a sort of wrap up. The main question that a lot of people ask is, you know, this device was able to come to market with two software mechanisms that could deliver a fatal dose of radiation. So what sort of testing was involved? What sort of FDA licensing process was involved? And the licensing process involved was something called premarket notification. So this was introduced because it was designed, in principle, this is quoted directly from the FDA, that “the device to be marketed is safe and effective, “that is, substantially equivalent to a legally marketed device.“ So because the– ALICE: We've seen this happen as a loophole with the car industry! We've seen, like, dangerous cars get through because they've managed to convince the Department of Transportation that it's not a substantial, it's not a new model, it's like a slight refinement of an existing thing that you've already passed. JUSTIN: Or maybe an airplane. ALICE: Or an airplane, yes! JUSTIN: [laughs] TOM: So the Therac-20 and the Therac-25, because the Therac-6 was on the market and had already gone through the full approvals process, neither of the other devices were required to go through the same, and they entered the market because of the 510k procedure. ALICE: Because it speeds things up, it allows you to iterate, it's efficient... I mean, it's also something else that harkens to the 737 MAX, is, as we just said, it's the airline industry too, is one of these industries that has benefited from success because it is very safe. It knows it's very safe, it considers itself to have a safety culture, and that was precisely the way that allowed it to be so complacent about these things. Is because, when you have this sort of, like, evidence base of safety, it allows you to go, oh, well this thing that happened was probably just an aberration, because mistakes like that do not happen in our industry. They don't happen to us. TOM: No. And the thing is, is that while the FDA did change its procedures here, so they did implement a way by which end users were then able to record sort of errors and report errors with medical devices, and software, because this is obviously, you know, software is a thing that's in relative infancy at the time, they still have issues with medical devices. So subsequently, to the Therac-25, there's been other device scandals which have occurred. In sort of the specialty use of orthopedics, there was a particular design of total hip replacement with a metal-on-metal bearing that came to market because it was a modified version of a previous design, and the previous design had undergone all the previous safety procedures, even though it had an entirely different bearing surface. And that resulted in lots of patient morbidity where these things failed early because of the eccentric wear on the cup. This is, that's an episode in itself. And also, sort of transvaginal mesh implants is the other big medical scandal. So, these are sort of implantable meshes that were originally developed sort of to repair sort of hernias. But then they applied for a license based on the fact that we use these to repair hernias, uterine prolapse is kind of like a hernia, it's still going in the abdominal cavity. ALICE: “It’s the same...” TOM: "It's basically the same." And there was a sort of, you know, there was a BMA investigation as well, and they sort of found that there were sort of tens of different devices, all having spun off from this one parent device, all doing completely different things in different parts of the body, made of different materials. And in the case of the sort of vaginal, transvaginal mesh implants, a huge scandal that's left lots of people severely... disabled, incontinent, all sorts of horrific consequences as a result. ALICE: Just sort of an easy way to get your foot in the door of regulations. You get one thing approved, and then just anything else like it. You can just, like, sort of slide through. JUSTIN: “Alright, the last thing was good, this one's probably good too... approve it, Bill!” ALICE and JUSTIN: [laugh] TOM: And in medicine, it's this sort of, there can be an element whereby this occurs in one specialty, and everyone sort of goes, well, hmm, well, that would occur in that specialty. But– ALICE: “Those idiots over in the radiology department, “they might do that. This will never happen to me in the urology department.” TOM: So I think this is... where there is a good argument for sort of saying, well, you've gotta keep abreast of what is happening in other sectors of your profession, because what's happened in one has happened in another, and it's happened multiple times. And any new device, you need to have some sort of level of scrutiny of all new devices coming to market. And after every one of these, there is then the push for regulation, but in a specific sector. So... this hip implant, which is something called the DePuy ASR, so as a result of that, lots of places in the US and the UK did change the way that they sort of record medical devices, they're much more stringent about what devices they record. They've implemented basically a complete ban on metal and metal hip bearings being used. Then there's several other fields whereby there are loads of devices on the market with premarket notification. And... I think caution– yes, it does drive innovation, and I think it's important to have health innovation from a personal standpoint, but at the same time, be aware of sort of making huge dramatic changes to one device or... ALICE: Just an easy way of doing regulation on the cheap, because governments don't like regulating things, and manufacturers don't like being regulated, and the only people who benefit from regulation are clinicians and patients, you know, and who cares about them? TOM: [sighs] ALICE: [laughs] JUSTIN: [laughs] ALICE: And I think we've learned a lot about the Therac-25. TOM: Yeah, it's good. I'm glad to have written a thesis, it's fantastic. JUSTIN: [laughs] ALICE: [laughs] ALICE: We think we have time for Safety Third, or do we want to get an early night after a mere three and a half hours? JUSTIN: I wanna go get some food before everything closes. ALICE: Perfectly reasonable. TOM: Fair enough. JUSTIN: Yeah. ALICE: Tom, thank you so much for coming on, and for writing the slides for this. TOM: That's no problem. ALICE: It's been very entertaining, and we'd love to have you back on any time we get anything, even tenuously medically related. TOM: That's fine, I'm happy to come back. Yeah. I have... JUSTIN: [mumbles] TOM: Sorry. JUSTIN: I was gonna say, our next episode is on the Chernobyl disaster. ALICE: Yes. If people want more Tom, where can they find you? TOM: So I have nothing to plug. You can find me working in the NHS, although you'd struggle to find me because I'm sort of in a lab most of the day. The only things I would say at this point, certainly if you're in the UK, please support the nurses who are on strike at the moment, please support ambulance drivers on strike, the BMA is holding a strike ballot in two days' time for junior doctors, so support junior doctors when we go on strike. And at some point, it looks like there may even be a consultant strike, as an indication of how problematic things are in the NHS, because, yeah. We are losing good people at a rapid rate, and it is at all levels of the organisation now. JUSTIN: It's incredible how Jeremy Corbyn did this. ALICE and JUSTIN: [laugh] ALICE: I don't know why he did this, and I don't really support him doing that, but, y'know. Well, we have a Patreon, you can subscribe to it, you get a bonus episode when we do a bonus episode, we do them as quickly as we can, which is not very quickly. JUSTIN: Yes. ALICE: We are also falling apart, but without under any of the other institutional pressures, we're just also doing that for fun. And y'know, do please do that. We would be very happy to take your money, and we will see you next time. JUSTIN: Yes.