The Peter Attia Drive - #286 ‒ Journal club with Andrew Huberman： the impact of light exposure on mental health and an immunotherapy breakthrough for cancer treatment

00:00:00.000 Hey, everyone. Welcome to the Drive podcast. I'm your host, Peter Atiyah. This podcast,

00:00:16.540 my website, and my weekly newsletter all focus on the goal of translating the science of longevity

00:00:21.520 into something accessible for everyone. Our goal is to provide the best content in health and

00:00:26.740 wellness, and we've established a great team of analysts to make this happen. It is extremely

00:00:31.660 important to me to provide all of this content without relying on paid ads. To do this, our work

00:00:36.960 is made entirely possible by our members, and in return, we offer exclusive member-only content

00:00:42.700 and benefits above and beyond what is available for free. If you want to take your knowledge of

00:00:47.940 this space to the next level, it's our goal to ensure members get back much more than the price

00:00:53.200 of the subscription. If you want to learn more about the benefits of our premium membership,

00:00:58.020 head over to peteratiyahmd.com forward slash subscribe. Welcome to another special episode

00:01:06.660 of The Drive. This episode is actually a dual episode with my good friend, Andrew Huberman,

00:01:12.240 where we are going to be releasing our conversation on both the Huberman Lab podcast and The Drive.

00:01:17.780 In this special episode, Andrew and I team up again for another round of Journal Club,

00:01:23.980 and you may recall this is the second time we've done it, having done it back in September of 2023.

00:01:29.680 We enjoy this so much that I suspect we're going to continue to do this, potentially at the cadence

00:01:34.500 of about once a quarter, but of course, we'll see. In today's Journal Club, we start by looking at a

00:01:40.100 paper that Andrew highlighted, which looks at how light exposure and dark exposure can affect mental

00:01:45.800 health. After that, I present a paper, which is kind of a landmark study on a class of drugs that

00:01:52.080 I believe are some of the most relevant classes of drugs in cancer therapy over the past 20 years,

00:01:58.040 the so-called checkpoint inhibitors. The hope here is not only that this conversation gives you

00:02:03.340 insights in the specific papers that we're discussing, both of which I think are highly fascinating,

00:02:09.600 but equally importantly, that you can learn something about how to read scientific papers,

00:02:14.080 what to look for, and what the papers say, and what's being reported, and how that doesn't

00:02:19.120 necessarily match with what the news is telling you. That's a really common issue, as many of you know,

00:02:24.080 and I certainly rail against this, where I'll comment on a paper that the media has picked up on

00:02:29.920 and completely misrepresented. And again, there's really only one antidote to this, and the antidote

00:02:35.000 is learning how to read the papers yourself. And unfortunately, there really is no better way to do

00:02:40.860 that than practice. And so what we really hope is that people will sit with us and maybe take a look

00:02:47.100 at the papers before they watch the podcast or listen to the podcast, and try to get a sense of

00:02:52.240 what they notice about these papers, what questions arise for them, and see if we touch on similar

00:02:56.920 topics. As a brief reminder to anyone who's been up in the Himalayas hunting Yeti for the past six years

00:03:02.500 and doesn't know who Andrew is, he is an associate professor of neurobiology and ophthalmology at the

00:03:07.000 Stanford University School of Medicine and the host of the very popular Huberman Lab podcast. He's also

00:03:11.880 a former podcast guest on episodes 249 and 270. So without further delay, please enjoy my conversation

00:03:20.020 with Andrew Huberman. Andrew, great to have you here for journal club number two. I'm already confident

00:03:31.540 this is going to become a regular for us. I'm excited. I really enjoy this because I get to pick

00:03:37.200 papers I'm really excited about. I get to hear papers that you're excited about, and we get to

00:03:43.420 sharpen our skills at reading and sharing data, and people listening can do that as well.

00:03:49.920 So last time I went first, so I think I'm going to put you on the hot seat first and have you go first,

00:03:54.840 and I'll follow you. Okay. Well, I'm really excited about this paper for a number of reasons. First of all,

00:04:00.860 it, at least by my read, is a very powerful paper in the sense that it examined light exposure

00:04:08.400 behavior as well as dark exposure behavior. And that's going to be an important point in more

00:04:13.500 than 85,000 people as part of this cohort in the UK. I'll just mention a couple of things to give

00:04:20.160 people background, and I'll keep this relatively brief. First of all, there's a longstanding interest

00:04:25.880 in the relationship between light and mental health and physical health. And we can throw up

00:04:30.660 some very well-agreed-upon bullet points. First of all, there is such a thing as seasonal affective

00:04:37.560 disorder. It doesn't just impact people living at really northern locations, but basically there's

00:04:43.400 a correlation between day length and mood and mental health, such that for many people, not all,

00:04:51.200 but for many people, when days are longer in the spring and summer, they feel better.

00:04:55.820 They report fewer depressive symptoms. And conversely, when days are shorter,

00:05:01.900 significantly more people report feeling lower mood and affect. There's a longstanding treatment

00:05:07.860 for seasonal affective disorder, which is to give people exposure to very bright light,

00:05:13.620 especially in the morning. The way that that's normally accomplished is with these sad lamps,

00:05:19.280 seasonal affective disorder lamps. And those lamps are basically bright, meaning more than 10,000

00:05:24.880 luxe lights that they place on their kitchen counter or at their table in the morning or in their

00:05:31.200 office. So they're getting a lot of bright light. That has proven to be fairly effective for the

00:05:37.140 treatment of seasonal affective disorder. What's less understood is how light exposure in the middle

00:05:42.260 of the night can negatively impact mood and health. And so where we are headed with this is that there

00:05:49.420 seems to be based on the conclusions of this new study, a powerful and independent role of both

00:05:57.300 daytime light exposure and nighttime dark exposure for mental health. Now, a couple of other key

00:06:03.740 points, the biological mechanisms for all this are really well established. There's a set of cells in

00:06:09.680 the neural retina, which aligns the back of your eye. They're sometimes called intrinsically

00:06:13.800 photosensitive. Retinal ganglion cells are sometimes called melanopsin retinal ganglion cells. We'll talk

00:06:19.860 about those in a bit of detail in a moment. It's well known that those cells are the ones that respond

00:06:25.720 to two different types of light input, not one, but two different types of light input and send

00:06:31.640 information to the hypothalamus where your master circadian clock resides. And then your master

00:06:36.180 circadian clock sends out secretory signals. So peptides, hormones, but also neural signals to the brain

00:06:42.920 and body and say, Hey, now it's daytime. Now it's nighttime, be awake, be asleep. But it goes way

00:06:48.940 beyond that. These melanopsin intrinsically photosensitive retinal ganglion cells, we know

00:06:53.640 also project to areas of the brain like the habenula, which can trigger negative affect, negative mood.

00:06:59.940 They can trigger the release of dopamine or the suppression of dopamine, the release of serotonin,

00:07:05.180 the suppression of serotonin. And so they're not just cells for setting your circadian clock.

00:07:10.240 They also have a direct line, literally one synapse away into the structures of the brain that we

00:07:16.260 know powerfully control mood. So the mechanistic basis for all this is there. So there's just a

00:07:21.240 couple of other key points to understand for people to really be able to digest the data in

00:07:26.020 this paper fully. There are basically two types of stimuli that these cells respond to. One is very

00:07:33.540 bright light, as we just talked about. That's why getting a lot of daytime sunlight is correlated with

00:07:39.300 elevated mood. That's why looking at a 10,000 lux artificial lamp can offset seasonal affective

00:07:44.620 disorder. By the way, just a couple of questions on that. How many lux does the sun provide on a

00:07:50.180 sunny day at noon? Okay, great question. So if you're out in the sun with no cloud cover or minimal

00:07:57.360 cloud cover in the middle of the day at noon, chances are it's over a hundred thousand lux.

00:08:03.700 On a really bright day could be 300,000 lux. Most indoor environments, even though they might seem

00:08:12.720 very bright, department store with the bright lights, believe it or not, that's probably only

00:08:18.320 closer to 6,000 lux maximum and probably more like 4,000 lux. Most brightly lit indoor environments

00:08:26.440 are not that bright when it comes down to total photon energy. Now, here's the interesting thing.

00:08:32.400 On a cloudy day, when you're outside, it can be as bright as an average of 100,000 lux, but it won't

00:08:42.040 seem that bright because you don't quote unquote see the sun. But it's also because when there's cloud

00:08:47.900 cover, a lot of those long wavelengths of light, such as orange and red light aren't coming through.

00:08:53.340 However, and this is so important, the circadian clock, the suprachiasmatic nucleus,

00:08:58.180 it sums photons. It's a photon summing system. So basically if you're outside in 8,000 lux,

00:09:08.400 very overcast UK winter day, and you're walking around hopefully without sunglasses, because

00:09:15.000 sunglasses are going to filter a lot of those photons out, your circadian clock is summing the

00:09:20.420 photons. So it's an integration mechanism. It's not triggered in a moment. And actually the experiments

00:09:26.620 of recording from these cells first done by David Burson at Brown were historic in the field of

00:09:31.520 visual neuroscience. When shown bright light on these intrinsically photosensitive cells, you could

00:09:36.140 crank up the intensity of the light and the neurons would ramp up their membrane potential and then

00:09:42.260 start spiking, firing action potentials or long trains of action potentials that have been shown to

00:09:47.700 go on for hours. And so that's the signal that's propagating into the whole brain and body.

00:09:52.900 So the important thing to understand is this is not a quick switch. That's why I suggest on

00:09:58.040 non-cloudy days, we'll call them, that people get 10 minutes or so of sunlight in their eyes in the

00:10:04.380 early part of the day, another 10 minimum in the later part of the day, as much sunlight in their

00:10:09.760 eyes as they safely can throughout the day. But since you're a physician and you had a guest on

00:10:13.880 talking about this recently, when the sun is low in the sky, low solar angle sunlight, that's really the

00:10:18.340 key time for reasons we'll talk about in a moment. And when the sun is low in the sky, you run very,

00:10:24.020 very little risk of inducing cataract by looking in the general direction of the sun. You should

00:10:27.780 still blink as needed to protect the eyes. It's when the sun is overhead, there's all those photons

00:10:32.760 coming in quickly in a short period of time. You do have to be concerned about cataract and

00:10:37.980 macular degeneration if you're getting too much daytime sunlight. So the idea is sunglasses in the

00:10:43.300 middle of the day are fine, but you really should avoid using them in the early and later part of the day,

00:10:47.140 unless you're driving into the sun for safety reasons. Another question, Andrew, if a person

00:10:52.300 is indoors, but they have large windows, they're getting tons of sunlight into their space. They

00:10:59.280 don't even need ambient indoor light. How much of the photons are making it through the glass

00:11:03.740 and how does that compare to this effect? In general, unless the light is coming directly

00:11:10.540 through the window, most of the relevant wavelengths are filtered out. In other words, if you can't see the

00:11:16.280 sun through the window, even if sufficient light is being provided, that's insufficient to trigger

00:11:21.640 this phenomenon? That's right. However, if you have windows on your roof, which some people do,

00:11:27.640 skylights, that makes the situation much, much better. In fact, the neurons in the eye that signal

00:11:34.360 to the circadian clock and these mood centers in the brain reside mainly in the bottom two-thirds of the

00:11:39.640 neural retina and are responsible for looking up. Basically, they're gathering light from above.

00:11:45.620 These cells are also very low resolution. Think of them as big pixels. They're not interested in

00:11:51.160 patterns and edges and movement. They're interested in how much ambient light there happens to be.

00:11:55.620 Now, keep in mind that this mechanism is perhaps the most well-conserved mechanism in cellular organisms.

00:12:01.880 And I'll use that as a way to frame up the four types of light that one needs to see every 24 hours

00:12:08.480 for optimal health. And when I say optimal health, I really mean mental health and physical health,

00:12:13.200 but we're going to talk about mental health mainly today in this paper. There's an absolutely beautiful

00:12:18.180 evolutionary story whereby single-cell organisms all the way to humans, dogs, rabbits, and everything in

00:12:25.620 between have at least two cone options, one that responds to short wavelength light, aka blue light,

00:12:32.500 and another one that responds to longer wavelength light, orange and red. So your dogs have this,

00:12:38.140 we have this, and it's a comparison mechanism in these cells of the eye, these neurons of the eye.

00:12:44.100 They compare contrast between blues and orange, or sometimes blues and reds and pinks, which are also

00:12:49.360 all long wavelength light. There are two times of day when the sky is enriched with blues, oranges,

00:12:57.080 pinks, and reds, and that's low solar angle sunlight at sunrise and in the evening. These cells are

00:13:05.600 uniquely available to trigger the existence of those wavelengths of light early in the day and in the

00:13:12.840 evening, not in the middle of the day. So these cells have these two cone photopigments and they say,

00:13:17.100 how much blue light is there? How much red light is there or orange light? And the subtraction

00:13:21.780 between those two triggers the signal for them to fire the signal off to the circadian clock of the

00:13:27.420 brain. And that's why I say, look at low solar angle sunlight early in the day. What that does

00:13:32.140 is it, what we call it is phase advances the clock. This can get a little technical and we don't want

00:13:36.600 to get too technical here, but think about pushing your kid on a swing. The period of that swing,

00:13:41.020 the duration of that swing is a little bit longer than 12 hours. So when you stand closer to the

00:13:49.520 kid, so your kid swings back and you give it a push, you're shortening the period. You're not

00:13:54.400 allowing the swing to come all the way up. That's what happens when you look at morning sunlight,

00:13:58.700 you're advancing your circadian clock. Translate to English or non-nerd speak. You're making it such

00:14:04.500 that you will want to go to bed a little bit earlier and wake up a little bit earlier the next

00:14:09.360 day. In the evening, when you view low solar angle sunlight, the afternoon setting sun or evening

00:14:16.040 setting sun, you do the exact opposite. You're phase delaying the clock. It's the equivalent of

00:14:21.340 your kid being at the very top of the arc. And so it's gone, you know, maybe let's say 12 and a

00:14:25.900 half hours is the duration of that swing. And you run up and you push them from behind and give them a

00:14:31.240 little more push. That's the equivalent of making yourself stay up a little later and wake up a

00:14:35.660 little later. These two signals average so that your clock stays stable. You don't drift, meaning

00:14:41.500 you're not waking up earlier every single day or going to sleep later every single day. This is why

00:14:46.620 it's important to view low solar angle sunlight in the morning and again in the evening as often as

00:14:52.920 possible. And it's done by that readout of those two photopigments. Now, midday sun, it's bright

00:15:00.140 light, but you see it as white light, contains all of those wavelengths at equal intensity.

00:15:05.120 So the middle of the day is the so-called circadian dead zone. In the middle of the day,

00:15:10.040 bright light triggers the activation of the other opsin, the melanopsin, which increases mood,

00:15:16.260 increases feelings of well-being, has some other consequences, but you can't shift your circadian

00:15:20.040 clock by viewing the sun in the middle of the day because it's in the circadian dead zone. It's the

00:15:24.260 equivalent of pushing your kid on the swing when they're at the bottom of the arc. You can get a

00:15:29.120 little bit more, but not much. And in biological terms, you get nothing. So this is why looking

00:15:35.120 at sunlight in the middle of the day is great, but it's not going to help anchor your sleep-wake cycle.

00:15:39.620 And if you think about it, this is incredible, right? Every organism from single cells to us

00:15:45.380 has this mechanism to know when the sun is rising and when the sun is setting. And it's a color

00:15:49.580 comparison mechanism, which tells us that actually color vision evolved first, not for pattern vision,

00:15:56.380 not for seeing beautiful sunsets and recognizing that's beautiful or paintings or things of that

00:16:00.960 sort, but rather for setting the circadian clock. Now, what if you only do one of these, Andrew? So

00:16:06.140 what if you've got constant exposure to low morning light, but your job prevents you from doing the

00:16:13.500 same in the evening or vice versa? Better to get the morning light because if you have to pick between

00:16:19.440 low solar angle light early or later in the day. And keep in mind, if you miss a day, no big deal.

00:16:24.120 It's a slow integrative mechanism averaging across the previous two or three days. But if you miss a

00:16:29.800 day, you'll want to get twice as much light in your eyes that next morning. The reason it's better

00:16:35.240 to do in the morning as opposed to the evening, although best would be to do both, is that most

00:16:40.540 people are getting some artificial light exposure in the evening anyway. And here's the diabolical thing.

00:16:46.020 Your retina is very insensitive to light early in the day. You need a lot of photons to trigger this

00:16:51.540 mechanism early in the day. As the day goes on, retinal sensitivity increases and it takes very

00:16:56.120 little light to shift your circadian clock late in the day. Keep in mind also that if you do see

00:17:01.480 afternoon and evening sunlight, there's a beautiful study published in Science Reports two years ago

00:17:07.360 showing that that can partially offset the negative effects of artificial light exposure at night.

00:17:12.660 I think of this as your Netflix inoculation. The amount of melatonin suppression from nighttime light

00:17:17.920 exposure is halved by viewing evening setting sun. Now, keep in mind, you don't need to see the sun

00:17:24.060 cross the horizon. It can just be when it's low solar angle. So you're looking for those yellow,

00:17:28.300 blue or blue, pink, blue, red contrasts. And on cloudy days, believe it or not, they're still there.

00:17:34.220 Just you don't perceive as much of it coming through. So that's three things that we should all strive to do.

00:17:40.120 View low solar angle sunlight early in the day. View solar angle sunlight later in the day and get as much

00:17:45.140 bright light in our eyes as we safely can, ideally from sunlight throughout the day.

00:17:49.880 And if you can't do that, perhaps invest in one of these satellites so that they can be a bit

00:17:54.580 expensive. There are a couple of companies that are starting to design sunrise simulators and evening

00:17:59.500 simulators that are actually good, that actually work. But right now, my read is that aside from one

00:18:05.460 company out there, which by the way, I have no relationship to, it's called the Tuolite, T-U-O.

00:18:10.140 And that light bulb was developed by the biologists of the University of Washington who

00:18:15.500 basically discovered these color opponent mechanisms. Those lights are not particularly

00:18:21.060 expensive, but they do seem to work. In fact, the study that is emerging, again, unpublished data

00:18:26.340 seems to indicate that if you look at it for more than five or six minutes, it can induce a mild euphoria.

00:18:31.880 That's how powerful this contrast is. And what they did there in that light, I'll just tell you the

00:18:35.920 mechanism, is they figured out that when most people look at low solar angle sunlight in the

00:18:39.880 morning, they're getting 19 reversals of blue-orange per second. So when you look at this

00:18:45.080 light, it looks like a barely flashing white light, but it's reversals of orange and blue,

00:18:50.620 orange and red and blue, and it's happening very fast.

00:18:53.720 What does the person looking at it perceive?

00:18:55.820 Well, I've used one of these. It just looks like a flickering light. And of course, there's always the

00:19:00.740 potential of a placebo effect.

00:19:02.040 Well, that's what I was going to say. Is there a way to control for that by having something that

00:19:06.320 looks the same to the user, but of course, is not producing the same photo effect?

00:19:12.420 Yeah, well, they've done that with the 10,000 luxe sad lamps, which most people use to try and

00:19:18.240 induce sunrise simulation in their home. But keep in mind that sunrise gives you this comparison

00:19:24.040 of short and long wavelength light. Just a bright 10,000 luxe light triggers one of the options,

00:19:30.280 but it won't set your circadian clock. So most of the sad lamps that are out there

00:19:34.660 are activating only one of the mechanisms in these cells that's relevant and not the one that's most

00:19:41.060 relevant. So I'm excited about what 2.0 is doing. I think that, and again, I have no relation to them,

00:19:46.020 except that I know the biologists who did the work that provide the mechanistic logic for that

00:19:50.140 engineering. I still think we're in the really early days of this stuff. What should be done

00:19:56.280 is to have this stuff built into your laptop. It should be built into your phone, and hopefully

00:20:00.800 it will be. Now, I mentioned this color contrast thing in sunrise and sunset. I mentioned the bright

00:20:06.820 light throughout the day, but there's a fourth light stimulus that turns out to be really important,

00:20:11.440 and this will provide the segue into the paper. It turns out that dark exposure at night,

00:20:16.800 independent of light exposure during the day, is important for mental health outcomes.

00:20:21.280 Most people think dark exposure. How do I think about that?

00:20:23.580 Absence of light exposure? It's the absence of light, but what this paper really drives home

00:20:29.360 is that people who make it a point to get dark exposure at night, aka the absence of light at

00:20:34.900 night, actually benefit even if they're not getting enough sunlight during the day. And this is especially

00:20:39.340 true for people with certain mental health issues. So I don't think we can overstate the value of

00:20:45.260 accurately timed light exposure to the eyes in the context of mental health. I think there's so much

00:20:51.880 data by now. I will say, however, that some people seem more resilient to these light effects than

00:20:58.220 others, meaning some people also don't suffer from jet lag too much. Some people can stay up late,

00:21:03.760 get a lot of bright light exposure in the middle of the night, and during the day,

00:21:06.620 they've got their sunglasses on all day, and they're in a great mood all the time.

00:21:09.760 Other people are more susceptible to these sorts of things, and we don't know whether or not

00:21:13.140 polymorphisms underlie that. I personally am very sensitive to sunlight in the sense that if I don't get

00:21:19.880 enough sunlight, I don't feel well after a couple of days, but I'm less sensitive to light exposure

00:21:25.340 at night, for instance. But I think it is perhaps, this is a big statement, but it is perhaps the most

00:21:32.080 fundamental environmental stimulus for levels of arousal and alertness, which correlate with all

00:21:38.820 sorts of neuromodulator and hormone outputs. None of this should come as any surprise. I will mention one

00:21:44.840 last thing. There was a study published, gosh, over 10 years ago now from Chuck Zeiser's lab at

00:21:49.040 Harvard Medical School. It's a phenomenal lab exploring circadian human health behavior. He's

00:21:55.720 just considered, no pun, a luminary in the field. But there was a study that was in error where they

00:22:02.600 had published in Science Magazine that light shone behind the knee could shift circadian rhythms,

00:22:07.780 and that paper was retracted. And a lot of people don't know that it was retracted. Light exposure to

00:22:12.360 the eyes is what's relevant here. And as far as we know, the color of one's eyes, darkness or

00:22:17.140 lightness of one's eyes bears no relevance on their sensitivity to these types of mechanisms.

00:22:22.480 So one question, one comment. The question again is going back to the morning, evening,

00:22:26.180 light. And I spend a lot of time looking at those types of skies, for example, just because of the

00:22:31.800 nature of my hobbies, because I'm always doing archery in the morning and rucking in the afternoon.

00:22:35.660 So it's not uncommon that I'm seeing both of those. How relevant is it that the sun be above

00:22:41.540 the horizon? So for example, it begins to get light about in 30 minutes before sunrise. So if sun rises

00:22:49.540 at 730, first light is seven, and then 715 to 730 is actually quite bright. I mean, you can see

00:22:57.280 anything and everything. And the same is true at sunset. So does that 30 minutes when sun is beneath

00:23:03.000 the horizon constitute part of that 10 minutes? It does. In an ideal circumstance, you'd get

00:23:08.480 outside and see the sunrise every day, and you'd see the sunset every day, even on cloudy days.

00:23:13.800 Some people like myself wake up before the sun comes up. And I get this question all the time.

00:23:18.460 Well, in the absence of powers to make the sunrise faster, which I'm not aware anyone has,

00:23:22.880 certainly not me. I think the best thing to do is simply to turn on as many bright lights as you

00:23:27.040 can indoors to trigger that melanopsin mechanism. If you want to be awake, if you want to stay asleep,

00:23:32.920 or sleepy, then keep them dim, and then get outside once the sun is starting to come out.

00:23:37.840 Some people wake up after the sun has risen, in which case get what you can. And some people wake

00:23:43.720 up 10 a.m. or noon, in which case you can still get the bright light exposure, but you won't shift

00:23:48.800 your circadian clock. Now, in the evening, especially in the winter months, it's important to look west

00:23:55.520 and try and get some sunlight in your eyes in the evening. If you've ever gone into the clinic,

00:24:00.060 for instance, at two o'clock in the afternoon, after lunch, and then in the winter, and then come

00:24:05.760 out and it's dark when you're walking to your car, it's an eerie feeling. That sort of eerie feeling

00:24:11.880 may correlate with the fact that you missed a signal. Your brain is trying to orient your brain

00:24:16.220 and body in time. And that's what all of this is. It's trying to orient in time. And again,

00:24:20.800 some people are more susceptible to that than others. Some people might like that feeling of,

00:24:24.580 oh, I went in when it was bright and I come out when it's dark. But the vast majority of people

00:24:29.580 feel better when they're getting this morning and evening sunlight exposure. And this is especially

00:24:34.460 important in kids. This is one of the things that this paper points out and there are good data that

00:24:39.080 people are spending approximately 90% of their time indoors nowadays, daytime time indoors. And those

00:24:47.140 indoor environments are simply not bright enough. You think, oh, there's all these bright lights. And

00:24:51.380 some people are putting blue blockers on in the middle of the day, which is the worst thing you

00:24:55.680 could possibly do. If you're going to wear blue blockers, and I don't think they're necessary,

00:24:59.100 but if you're going to wear them, you'd want to wear them at night. And in the evening,

00:25:02.560 you don't need to wear blue blockers. You just simply should dim the lights and ideally have

00:25:07.180 lights that are set a little bit lower in your environment, which the Scandinavians have been doing

00:25:11.860 for a long time. So kill the overhead lights and don't obsess about bright light exposure in the

00:25:18.200 middle of the night. In fact, for a long time, I and some other people were saying, oh, you know,

00:25:22.340 even just a brief flash of light in the middle of the night can quash your melatonin. That's true.

00:25:27.180 But the other time in which you're in this quote unquote circadian dead zone is in the middle of

00:25:32.560 the night. You can't shift your circadian clock in the middle of the night. But all of this gets down

00:25:37.320 to interweaving rhythms of light sensitivity, temperature, hormone output, cortisol. I mean,

00:25:45.100 there's a whole landscape of circadian biology. This paper, which was published in a new journal,

00:25:50.460 I'm really excited about called nature mental health. This journal was just launched recently

00:25:54.900 is entitled day and night light exposure are associated with psychiatric disorders and

00:25:59.840 objective light study in more than 85,000 people. Now I have to say that I think the title of this

00:26:05.800 paper is terrible. Sorry, folks at nature mental health, because if one just read the title,

00:26:10.680 it sounds like day and night light exposure are associated with psychiatric disorders,

00:26:15.260 right? If this were a newspaper headline, you'd be like, oh my goodness, well, what are you supposed

00:26:18.360 to do, right? But that's not the conclusion. The conclusion is that getting a lot of sunlight

00:26:24.920 exposure during the day and getting a lot of dark exposure at night is immensely beneficial for

00:26:31.240 psychiatric health in a number of ways. Now, I'm not one to bring up another paper unannounced,

00:26:36.440 but I will say that this paper built off a previous study entitled time spent in outdoor light is

00:26:42.180 associated with mood, sleep, and circadian rhythm related outcomes. And that was a cross-sectional

00:26:47.500 longitudinal study in 400,000 biobank participants. So this UK biobank is an incredibly valuable resource.

00:26:55.300 And there are now multiple studies establishing that one's pattern of light exposure is extremely

00:27:00.600 important. Now, the previous study in 400,000 participants basically nailed home the idea that

00:27:06.860 the more time you spend outdoors, the better is your mood, the better is your sleep, the better is

00:27:13.720 the rhythmicity of your sleep-wake cycles, and on and on. Something that I think, even though people

00:27:18.980 say we've known that for thousands of years, needed scientific substantiation. This new study

00:27:24.780 essentially looked at the relative contributions of daytime light exposure and nighttime dark

00:27:31.520 exposure. And they did that on a background looking in particular people who had major depressive

00:27:36.340 disorder, generalized anxiety, PTSD, bipolar disorder. Here's the basic takeaway. I'll quote them

00:27:43.000 here. I'll tell you my interpretation. Here I'm quoting, avoiding night at light and seeking light during

00:27:48.160 the day. I love that word seeking. Maybe a simple and effective non-pharmacologic means for broadly

00:27:54.720 improving mental health. So that's a pretty bold statement. And I love that they say seeking

00:27:59.100 because it implies that people aren't reflexively getting the light exposure that they need, that

00:28:03.720 this needs to be a practice, much like zone two cardio or resistance training. So what did they

00:28:08.360 do in this study? So basically they gathered up a hundred thousand people or so. It eventually

00:28:14.540 was pared down to about 86,000 participants because some just didn't qualify or didn't report

00:28:21.160 their data back. They equipped them with accelerometers on their wrists and those wrist devices also could

00:28:28.420 measure ambient light. Now that's not a perfect tool because what you'd love to do is measure

00:28:32.580 ambient light at the level of the eyes. By the way, will somebody design an eyeglass frame that

00:28:37.980 changes color when you've gotten sufficient light from sunlight during the day? And then at night

00:28:42.960 is a different color. And then if you're getting too much light exposure, we'll go to a different

00:28:46.760 color frame. This has to be possible so that you don't have to wonder if you got enough light during

00:28:51.240 the day. And of course, if it's at the level of the eyes, then you know, that's what's landing at

00:28:55.680 the eyes. So I was going to ask you about that. Do these wrist based devices potentially get covered

00:29:01.040 by clothing and some turned over, you have your sleeves down. I have my sleeves. Yeah. They had it

00:29:05.420 on the outside of the sleeve, but they asked that people just keep it on their dominant hand.

00:29:09.000 It's not perfect, but in some ways it's kind of nice that it's not perfect. We could turn that

00:29:13.600 disadvantage into advantage by thinking when the person is out and about, they're not often looking

00:29:18.600 right at the sun. If you're talking to a colleague under an overhang, for instance. So it's not

00:29:24.520 perfect. It's directionally right. Okay. And then they had two hypotheses, two primary hypotheses. One

00:29:30.460 that greater light exposure in the day is associated with lower risk for psychiatric disorders. And two

00:29:34.560 second hypotheses, greater light exposure at night is associated with higher risk for psychiatric

00:29:39.700 disorders and poor mood. This is oh so relevant for the way we live now. People on screens and

00:29:44.860 tablets in the middle of the night. Then they collected information about how much light exposure

00:29:49.040 people were getting as well as their sleep and their activity and so on. I should mention this was

00:29:54.100 done in males and females. It was a slightly older cohort than one is used to seeing people in their

00:29:59.060 fifties and sixties. They had psychiatric diagnosis information and then they divided people into

00:30:05.020 essentially two groups, but they had a lower. So a Q1 and a Q2, a lower quartile. That meant people

00:30:10.880 that were getting less daytime light as opposed to the third and fourth quartile, more daytime light.

00:30:17.000 They also had a nighttime light exposure evaluation and they had people with the low Q1 and Q2. So

00:30:24.300 these people are getting less nighttime light versus Q3, Q4, more nighttime light. Nicely,

00:30:30.900 they also looked at sleep duration and they looked at photo period, meaning how long the days were for

00:30:37.700 those individuals, how active they were, like 10 hours a day, 14 hours a day, because the more active

00:30:42.420 you are, the more opportunity for light exposure you have during the day or night, for instance. Okay.

00:30:49.060 So they had, I would say, fairly complete data sets then, and I'm just going to hit the top contour of

00:30:55.560 what they did. And sorry, sleep duration, sleep efficiency, et cetera, was determined off the accelerometer.

00:31:00.900 That's right, as well as self-report. Not ideal, right? You'd love for people to be wearing a whoop band

00:31:05.760 or a ring or something of that sort, but this was initiated some time ago. So they either didn't have

00:31:11.780 access to that technology or for whatever reason, didn't select it. Then what they did is they have

00:31:18.260 information on who has major depressive disorder, who has PTSD, generalized anxiety, bipolar, psychosis,

00:31:24.680 et cetera. And then they ran three models. You can tell me what you think about the power of these

00:31:29.960 models, but you know, as somebody who thinks about the mechanistic aspect of all of this a lot, but not

00:31:35.580 somebody who's ever run this type of study, I'd be really curious. Model one examined the unadjusted

00:31:41.600 association between day and nighttime light exposure and psychiatric outcomes. So just basically asking,

00:31:47.700 is there a relationship between how much light you get during the day and how much light you get at

00:31:51.960 night and how bad your depression is or anxiety is, et cetera. Looking at just a standard ratio of

00:31:59.480 the probability that you have a certain symptom or set of symptoms versus you don't given a certain

00:32:06.460 amount of light exposure. Model two adjusted for the age of the person, their sex and ethnicity

00:32:13.000 and photo period. So they looked at how long the days were in that given person's region of the

00:32:18.660 world. These people were all in the UK or were they around? They were all in the UK as far as I

00:32:22.980 know. And then model three adjusted for employment. So employed versus unemployed, which if you think

00:32:28.540 about it is pretty important, like you say, well, an unemployed person has a lot more time to control

00:32:32.180 these variables, but an employed person who's doing shift work does not. They incorporated

00:32:37.560 information about employed versus unemployed physical activity, which turns out to be very

00:32:43.520 important. And then things like shift work, et cetera. We can say very safely the outcomes with

00:32:49.980 each of these models, the results were very similar. So we don't want to discard the differences

00:32:56.620 between those models entirely. But in my read is in every figure of the paper, it doesn't seem like

00:33:02.060 model one, two, or three differ from one another in terms of total outcome.

00:33:06.100 That's an unusual aspect of this paper. So these adjustments are very standard. This is a classic

00:33:12.360 tool that's used in most epidemiology because you don't have randomization. So once randomization is

00:33:18.380 out the window, so for example, the paper I'm going to present is based on an RCT, there will be no

00:33:22.960 models. It's just here are the data. Yeah. Here they're asking people, what do you do? Report back

00:33:27.840 to us. We're going to measure your light exposure, but no one was assigned to any groups or swap,

00:33:32.040 whatever quote unquote controls are there. They were really not there. It's just comparisons

00:33:35.900 between groups. So what is interesting to me is that it's exactly as you said, and we'll make all

00:33:41.940 these figures available in addition to the papers, but it's very unusual that there's no difference

00:33:46.880 between the unadjusted and the adjusted models. And as you say, there's probably two places out of

00:33:55.120 30 when you look at all the different quartile comparisons where you might creep from statistically

00:34:02.480 significant just out of it or just into it. But yeah, you could simplify this figure two completely

00:34:08.720 by just showing one of the models and you would be getting 95% of the information. I think in one way

00:34:14.960 that suggests that there's less dependency on those variables. Of course, it still doesn't address

00:34:23.840 probably the greatest question I have here, which I'm sure we'll get to at some point as you continue.

00:34:28.960 I'm very curious what that question is. I'll suppress my curiosity for the moment.

00:34:33.820 So if we look at figure two of this paper, and I realize a lot of people are listening and they're

00:34:37.200 not able to look at this, although we have posted the figures on the YouTube version of this,

00:34:41.880 just want to make clear what's going on just for those that are listening. Essentially what they're

00:34:46.880 looking at is what they call the odds ratio, which is the probability of something happening in one

00:34:53.020 group divided by the probability of something happening in another group. I guess by way of example,

00:34:57.220 would be if you were going to look at the odds ratio of the probability of somebody getting lung

00:35:01.520 cancer if they smoke versus probability of somebody getting lung cancer if they don't smoke.

00:35:06.060 So odds ratios and hazard ratios are often confused. They're very similar and odds ratios generally refer

00:35:12.800 to a lifetime exposure, whereas a hazard ratio is defined over a specific period of time. But the math

00:35:19.260 is still effectively the same. And using the example you gave, if you took the odds ratio of death,

00:35:26.580 so let's talk all-cause mortality for a smoker versus a non-smoker, and the answer were 1.78.

00:35:31.600 I'm making that up, but that's directionally correct. 1.78 as an odds ratio means there's a

00:35:37.160 78% chance greater of the outcome of interest, in this case, death by any cause in the affected

00:35:44.580 group, which would be the smokers. So odds ratio of two is 100%, and odds ratio of three is 200%. So

00:35:51.380 the math is take the number, subtract one, and that's the percent.

00:35:55.100 Thank you for that.

00:35:56.280 Right. So figure two of this paper is one of the key take-homes. Essentially look at the odds ratio

00:36:02.260 of people who are in the, let's just look at the nighttime light exposure.

00:36:08.060 And just remind me, Andrew, and everybody else watching, every one of these is showing second,

00:36:13.020 third, fourth as your x-axis, meaning they're all being compared to the first quartile.

00:36:18.240 That's right.

00:36:18.660 And the first quartile is lowest light exposure or highest light exposure.

00:36:23.900 That's right. Lowest.

00:36:24.340 Well-

00:36:24.620 With the differentiate between day and night.

00:36:26.040 That's right. That's right.

00:36:27.160 Restate it.

00:36:28.040 Sure. So if we look at what is your risk of a psychiatric challenge, broadly speaking,

00:36:34.920 well, panel A is major depressive disorder. If you are in the second quartile, third quartile,

00:36:41.040 or fourth quartile of nighttime light exposure. So second being the least amount of nighttime light

00:36:46.740 exposure, third being more nighttime light exposure, and fourth, the most nighttime light exposure

00:36:52.780 relative to the first quartile.

00:36:55.000 This is just a stupid thing. Like if I were doing this figure, if you were doing this in a lecture,

00:37:00.260 you know what you would do to make it so easy? You would draw arrows on it that say increasing

00:37:04.500 light exposure at night, decreasing light exposure in the day. It's the same information. It just

00:37:09.800 makes it easier for the reader to understand. Maybe the teaching point I think is for people

00:37:14.060 when they review articles, like don't be afraid to do that and just-

00:37:18.100 Oh, I've got scribble all over this.

00:37:19.600 Yeah, exactly. So it's like, I draw the arrow. That's increasing light,

00:37:22.580 that's decreasing light. And that's how I can pay attention to what's actually happening.

00:37:25.660 Right. And I'm actually in touch with the editorial staff at Nature Mental Health,

00:37:29.380 although they don't know that I'm covering this paper until after this comes out. You know,

00:37:32.400 I think one thing that scientific journals really, really need to do is start making the

00:37:37.260 readability of the articles better for non-experts. I mean, chances are, if you can't understand a

00:37:43.800 graph, and this is true for everybody, chances are there's a problem with the way it's presented.

00:37:49.780 Put it on them, but then of course, try and parse it because rarely, if ever, is it all spelled out

00:37:55.140 clearly. But anyway, that's what we're trying to do here. So yeah, the way I would have done is say

00:37:58.660 second quartile is low amounts of nighttime light exposure and define what that is. Third quartile

00:38:03.960 is more light exposure and then fourth maximum amount of light exposure at night. And basically

00:38:08.180 what you see is that the probability of having worse major depressive symptoms linearly increases

00:38:16.480 as you go from the second to third to fourth quartile. So more nighttime light exposure, worse

00:38:22.340 for you. And there's a dose response, if you will, of the effect. Now we can march through or describe

00:38:31.060 figure two pretty quickly by saying the same thing is true. Now we're just talking about nighttime light

00:38:37.080 exposure for generalized anxiety disorder. So that's panel C. Bipolar disorder, although the difference

00:38:43.380 between the second and third quartile and bipolar disorder isn't as dramatic. Once you get up to

00:38:47.280 the fourth quartile, bipolar symptoms get much worse when people are getting nighttime light

00:38:53.260 exposure. I really want to emphasize that point because they go on in the discussion of this paper

00:38:58.540 to reemphasize that point several times. In fact, they say that while light exposure during the day,

00:39:04.760 of course, we will go into the data, is beneficial for mental health. For people with bipolar disorder,

00:39:10.780 it seems that light exposure at night is especially problematic, independent of how much sunlight

00:39:16.460 they're getting during the day. The person with bipolar disorder who's struggling with either a manic

00:39:20.760 or a depressive episode, who's making a point to get sunlight during the day, who's also getting

00:39:25.520 light exposure at night, is making their symptoms worse. And keep in mind, they couldn't completely

00:39:30.180 control this, but this is largely independent of things like sleep duration. So that doesn't necessarily

00:39:36.940 mean that the person's sleeping less, although in a manic episode, presumably they are. It's independent

00:39:41.680 of exercise. It's independent of a bunch of other things because any logical person will hear this

00:39:46.580 and say, okay, well, they gain more light at night because they're doing a bunch of other things,

00:39:49.440 but it's largely independent of those other things. Likewise, the symptomology of PTSD gets far worse

00:39:55.660 with increasing light exposure at night. Self-harm really takes a leap from being fairly,

00:40:02.340 I don't want to say minimal, at the second and third quartile. So low and let's say medium,

00:40:07.100 I'm taking some liberties here, but low and medium amounts of artificial light exposure at night

00:40:10.940 than for people who get quite a lot of nighttime light exposure. Self-harm goes up, probability of

00:40:17.280 psychotic episodes goes up or psychotic symptoms. Now, what's nice about the data is that the exact

00:40:22.980 inverse is basically true for daytime light exposure. Although not across the board, we can

00:40:28.280 generally say that for major depressive disorder, generalized anxiety, bipolar symptoms, there it's a

00:40:34.260 little more scattered, PTSD and self-harm, the more daytime light exposure, ideally from sunlight,

00:40:40.720 because that's actually what's being measured in most cases. We can talk about how we know that

00:40:45.020 is going to approximately linearly drop the probability or the severity of these symptoms.

00:40:52.860 And we could just explain again that the odds ratios now seem to be going down. So an odds ratio

00:40:57.540 of 0.7 now refers to a 30% reduction in the variable of interest here.

00:41:03.680 Exactly. Now the psychosis panel F, which focuses on psychosis, I think is also worth mentioning

00:41:09.180 in a bit more detail. There's a fairly dramatic reduction in psychotic symptoms as one gets more

00:41:14.960 daytime light exposure, independent of nighttime light exposure. There's a well-known phenomenon called

00:41:21.140 ICU psychosis, which is that people come into the hospital for a broken leg or a car accident,

00:41:27.280 maybe they were getting surgery from Peter back when for something totally independent.

00:41:31.820 They're housed in the hospital. And as anyone who's ever been in a hospital as a patient or visitor

00:41:35.960 knows, the lighting environment, the hospital is absolutely dreadful for health. Just dreadful.

00:41:41.420 People often complain about the food in the cafeteria as being unhealthy. That's often not always true,

00:41:47.020 not always true, but the lighting environments in hospitals is absolutely counter to health.

00:41:52.280 Especially in the intensive care unit. I think the intensive care unit at Hopkins, the main one

00:41:57.280 main SICU didn't have windows. People who go into the hospital with a brain injury or

00:42:02.780 with a stroke or something, I get contacted all the time, even though I'm not a clinician. What should

00:42:06.680 I do for my kid, my parent? I always say, get them near a window and start to the best of your

00:42:11.320 abilities, controlling their sleep-wake cycle. Now, oftentimes there's nurses coming in and taking

00:42:15.940 blood tests and measuring pulses in the middle of the night. That's disruptive. There's bright light,

00:42:20.140 not just blue light. That's disruptive. It's noisy. That's disruptive. ICU psychosis is when

00:42:25.440 non-psychotic individuals start having psychotic episodes in the hospital because of nighttime light

00:42:31.240 exposure and in some cases, lack of daytime sunlight. We can say that with some degree of confidence

00:42:37.560 because when those people go home, even though sometimes their symptoms for what brought them

00:42:41.700 to the hospital in the first place get worse, their psychosis goes away and it's independent of

00:42:47.320 medication. So let's just be really direct. There is a possibility that we are all socially jet-lagged,

00:42:56.040 that we are all disrupting these mood regulation systems by not getting enough daytime light and by

00:43:02.180 getting too much nighttime light. If we want to look at just some of the bullet points of the takeaways,

00:43:06.960 and then Peter, thank you. You highlighted a few of these. Can we just go back to this figure too

00:43:11.220 for a second? Oh yeah, sure. There's a handful of things that really jump out. I had a feeling Peter was

00:43:14.200 going to want to dig in the data more. Let's do it. Let's do it. And again, I normally wouldn't make

00:43:18.020 so much hay out of this except for the fact that they're so tight. There are a few that really stand

00:43:23.300 out. And again, I love this figure. I would have labeled it a little differently to make it

00:43:27.320 completely user-friendly, but nevertheless, the increasing light at night and the impact on

00:43:34.220 depression, let me be really technical in what I say, and the relationship or correlation to

00:43:39.120 depression is very strong. The relationship to light and self-harm in the upper quartile,

00:43:46.640 so when you take those 25% of people with the most nighttime light, that relationship to self-harm

00:43:52.900 is interesting and completely uncoupled from the other 75%. That's interesting.

00:43:57.720 By uncoupled, you mean that at the lower levels of light exposure at night, you're not seeing an

00:44:01.680 increase in self-harm. Not whatsoever.

00:44:03.500 And then once you get to that fourth quartile-

00:44:05.720 It's a big step.

00:44:06.380 It's like a 30% greater risk of self-harm.

00:44:09.580 That's right. Yeah. So it's totally flat. The first, second, third quartile,

00:44:12.920 no different. And then fourth, big jump. And then the inverse relationship. As light increases during

00:44:18.500 the daytime, you see this reduction in self-harm. Interesting. The PTSD relationship based on

00:44:24.980 nighttime light and the psychosis relationship based on daytime light. Those are the ones that really

00:44:31.180 jumped out to me. I think anxiety, relatively less impressive here. And bipolar disorder didn't seem

00:44:39.060 as strong as well. So I think those are the big ones that jumped out to me.

00:44:43.960 Yeah, I agree. There's a bit more scatter on generalized anxiety and the degree of significant

00:44:48.000 change is not as robust. In other words, getting a lot of daytime light, ideally from sunlight,

00:44:53.420 is not necessarily going to reduce your levels of anxiety. Getting a lot of nighttime light exposure

00:44:58.860 is not increasing nighttime anxiety that much. Although 20% is not nothing for nighttime light

00:45:05.760 exposure. But yeah, the psychosis, major depression, and self-harm, they leap out.

00:45:11.780 Maybe we can just drill a little bit deeper on major depression. And basically when you go from

00:45:15.280 the second to third quartile of nighttime light exposure, so more nighttime light exposure,

00:45:19.060 you basically go from no significant increase to almost a 20% increase. And then as you get

00:45:25.320 up to the fourth quartile, so the most nighttime light exposure at about 25% increase in major

00:45:33.260 depressive symptoms. And that's no joke. I think that we don't have the data right here,

00:45:38.000 but if we were to look at standard SSRI treatment for major depression, people debate this pretty

00:45:43.140 actively, but light is a very potent stimulus. And the timing of light is critical because the inverse

00:45:49.140 is also true. As you get to the fourth quartile of daytime light exposure,

00:45:52.120 you get about a 20% reduction in major depressive disorder.

00:45:56.860 Yep. What I like about a study like this is that it puts the error bars so easy to see on the data.

00:46:03.720 And why is that interesting? Well, there's a belief that bigger is always better in sample size.

00:46:10.340 And we often talk about that through the lens of power analysis. So how many subjects do we need

00:46:16.020 to reach a conclusion that is powered to this level? That's true. But what I don't think gets

00:46:22.300 discussed as often is the opposite of that, which is what if you overpower a study? In other words,

00:46:28.500 what if the power analysis says to have a level of power at 90%, you need a thousand subjects and you

00:46:35.060 say, great, we're going to do 10,000 subjects. Well, you're clearly powered for it, but you might be

00:46:40.160 overpowered and people might say, well, why would that be a bad thing? It could be a bad thing

00:46:44.240 because it means you are very likely to reach statistical significance in things that might

00:46:49.720 not be actually significant. And so one thing about this study that is just a quick and dirty way to

00:46:56.400 tell that it's probably not overpowered is that you have varying lengths of error bars. And what that

00:47:04.120 tells me is that, and again, this is not like a formal statistical analysis. It's just kind of like a

00:47:10.520 back of the envelope statistical analysis. If you look, for example, at self-harm in the top quartile,

00:47:15.620 you actually have pretty big error bars. In fact, all the self-harm have sort of slightly bigger error

00:47:19.700 bars. And yet when you look at, for example, the depression, even though the error bars aren't all

00:47:23.920 the same size, they're tighter. In fact, when you look at the relationship between depression and

00:47:27.700 daytime light, the error bars are really, really small. So that just gives me confidence that there

00:47:33.380 is variability in this, which paradoxically you kind of want to see, because it tells me that

00:47:38.140 this wasn't just done. There was, I think you said 8,000 subjects were in this.

00:47:43.020 86,000.

00:47:43.740 86,000. Sorry. Yeah. You realize that it wasn't that, oh, this should have been done with a 10th

00:47:48.500 of that or a half of that. And we're picking up signal that is statistically relevant, but

00:47:53.060 clinically irrelevant.

00:47:54.560 Thanks for that point. I didn't pay attention to that. I mean, I paid attention to the error

00:47:59.460 bars, but I didn't know that. So thank you. I'm learning too. I suppose for people that are

00:48:04.260 listening, we can just give them a sense of what the error bar ranges are. For self-harm,

00:48:08.700 you know, they're running as much as 20% either side of the mean, the average. And for major

00:48:14.080 depression, it looks like it's more like, if that, maybe more like five. Yeah. I see what

00:48:18.680 you're saying. So when you get a very large sample size, you're going to have some outliers

00:48:24.520 in there and you can mask those outliers just by having so many data points.

00:48:29.900 Because these error bars directly tell you whether or not you're statistically significant.

00:48:34.120 So what's really nice about this type of graph, and you see these in, there's going to be a graph

00:48:39.620 in my paper where you see the same analysis. They're always drawing the 95% confidence interval

00:48:45.180 on the data point. And if the 95% confidence interval does not touch the line of unity, which

00:48:52.960 in this case is the Hod's ratio of 1.0 or the x-axis, then you know it's statistically significant

00:48:58.120 to the confidence interval they've defined, which is almost assuredly 95%. Sometimes they'll make it

00:49:02.880 tighter at 99. And so that's why you can just look right at these and go, oh, look, in depression,

00:49:08.140 the second quartile didn't reach statistical significance because the error bars are touching

00:49:12.640 the line, just as the case for the second and third quartile for self-harm. But when you look at

00:49:18.860 the fourth quartile, you can see that the lower tip of the error bars isn't anywhere near unity.

00:49:23.900 And so we know without having to look up the p-value that it's smaller than either 0.05

00:49:29.520 or 0.1, however they've defined it. And it's really amazing when you see these overpowered

00:49:34.480 studies, which are easier to do epidemiologically, where the p-value ends up being microscopic.

00:49:40.360 They can drive their p-values down to anything low because sample size can be infinite, but

00:49:44.880 you can see that it's just, the error bar is just skimming above the unity line, but it's so,

00:49:51.360 so, so tight. Interesting. One thing that I hope people are taking away from this study is that

00:49:57.920 imagine you're somebody who has a very sensitive circadian mood system. Well, that would mean you

00:50:06.240 need less daytime light exposure to feel good or less bad, but it also means that you might need

00:50:14.080 very little light at night in order to negatively impact your mood systems. And in fact, they make

00:50:22.260 this argument in the discussion as an interesting point. Again, what I like about the study is that

00:50:27.520 they've separated day and nighttime light exposure. It turns out that many of the drugs that are used

00:50:33.100 to treat bipolar disorder are effective, perhaps in part because they reduce the sensitivity of the

00:50:39.520 light sensing circadian apparati. Now that's interesting, right? If you think about this, okay, so these

00:50:45.840 are drugs that can ameliorate some of the symptoms of bipolar, perhaps in part by reducing the extent to

00:50:51.460 which nighttime light exposure can exacerbate bipolar symptoms. Conversely, there's evidence that people

00:50:57.600 who take certain antidepressants may suppress the ability for daytime light to positively impact the mood

00:51:05.100 systems of the brain. Now, of course, we don't want people halting their medication on the basis of that

00:51:09.900 statement alone, please don't talk to your psychiatrist. But if we know one thing for sure, it's that if you want

00:51:16.180 a significant outcome and a paper, as a scientist, give a drug, any drug, and look at the amount of rapid eye

00:51:25.940 movement, sleep, or the circadian cycle, pretty much any drug alters a circadian rhythm, for better or worse. If we start

00:51:33.060 to think about which medications might adjust our overall sensitivity to light, sometimes this could

00:51:37.760 be a good thing. You think less sensitivity to light. Well, for people who have bipolar disorder, the amount

00:51:42.580 of daytime light exposure isn't that important for their overall mood regulation, but the amount of

00:51:47.280 nighttime light exposure really is. In other words, darkness for eight hours every night should be

00:51:52.200 viewed, in my opinion, as a treatment for bipolar disorder, not the only treatment. But it's also clear

00:51:58.080 that we should all be avoiding really bright, extensive, really bright nighttime light exposure. I mean, if

00:52:04.080 anything, my takeaway from this study is that darkness at night is the fourth key light stimulus. A couple of

00:52:10.620 things, very bright moonlight, very bright candlelight is probably only like, gosh, three to 50 lux. When you go

00:52:20.780 outside on a brightly lit full moon night, I encourage people to download this free app. I have no relationship to it

00:52:26.960 called light meter. It gives you a pretty good read of what the lux are in that environment. By the way,

00:52:31.760 a lot of people don't realize this. They think you just tap the button and then it tells you how many

00:52:35.180 lux. You hold it down. It's kind of fun. You can scan around the room and see how many lux are on

00:52:39.520 average coming from that location or outside. Go out on a really bright, moonlit night.

00:52:44.520 We should have a full moon tonight.

00:52:46.060 Yeah. Let's do it. You're not going to get above 100 lux.

00:52:48.800 That's incredible.

00:52:49.680 You're sitting at a candlelight dinner with your spouse or with friends, and it's clearly bright enough to

00:52:54.420 see them. Put that lux meter right up, not too close to the flame, 50 to 200 lux. Maybe 400 lux.

00:53:03.280 I had an interesting experience a couple of months ago on an elk hunt where it was a full moon,

00:53:08.200 which actually makes the hunting not so great. But it was the first time I've ever noticed my shadow

00:53:15.240 in relation to the moon. That's how bright it seemed the light was.

00:53:19.540 This is Halloween appropriate. We're recording this close to Halloween.

00:53:22.340 Super interesting to think it could be that dim.

00:53:25.580 Campfire, firelight. You think, okay, gathering around a campfire, then okay,

00:53:29.840 everyone's circadian rhythm must have been disrupted for ages before the development of

00:53:33.540 electricity. No. Those campfires are extremely bright, but they're not that bright compared

00:53:39.020 to a very densely overcast day.

00:53:41.820 And what is your phone if you don't use any sort of light mitigating tech on it?

00:53:46.980 Well, distance matters.

00:53:48.880 At the distance we're holding it?

00:53:50.080 Yeah. So with all the wavelengths cranked up, there is a nice feature intrinsic to the phone

00:53:55.620 where you can eliminate the blues at night or this kind of thing. But if you crank it up to maximum

00:53:59.340 light intensity, probably something like 500 to 1000 lux. Now, keep in mind though, it's additive.

00:54:05.640 So it's over time. So lux is a measure of, I think it relates back to candelas is the amount of light

00:54:13.220 shown. And I think it's like the one meter away and there's a squaring and a falling off of distance.

00:54:18.080 We can look it up. These are old school measurements converted to lux. But keep in mind that if you're

00:54:22.540 looking at your phone or tablet at 800 lux or 500 lux in the evening, you do that for two hours,

00:54:28.240 where you're summing quite a lot of photons. Now it is true. I do want to be fair to the biology

00:54:33.680 and it'd be dishonest to say anything different. We've hammered on people about not shifting their

00:54:39.000 circadian rhythm with light at night, but we know that the middle of the day and the middle of the

00:54:42.740 night are circadian dead zones. You can't shift your circadian rhythm that well in the middle of

00:54:47.600 the day, in the middle of the night, but you can provide a wake-up signal for your body and brain.

00:54:52.280 It's really that sunrise and sunset that are critical. That's why I said there are four

00:54:56.000 things. See sunrise or sun rising. You don't need to see across the horizon. Sunset, bright light

00:55:01.620 during the day, minimize light exposure at night. You don't need pitch black. In fact, pitch black

00:55:07.620 probably just increases the frequency of injury. You know, I get up in the middle of the night to use

00:55:11.020 the bathroom probably once. I think it's normal. I go back to sleep. If it were pitch black, I'd probably

00:55:16.540 injure myself. So just dim it down. Some people use red lights.

00:55:19.940 Our mutual friend. Yeah. Rick Rubin is, uh, can we tell a funny story about Rick?

00:55:24.120 Yeah, of course. Of course. You know this story, but just in case Rick's listening,

00:55:27.620 he'll appreciate this. Rick was here staying last summer. He's up in our guest house and he came

00:55:33.400 down after the first night and he was like unacceptable. You know what I'm talking about?

00:55:37.620 Unacceptable accommodations. What did he do? He removed all the lighting that existed in that

00:55:43.180 room and replaced it with red light bulbs, which I later used when I stayed here and then later stole

00:55:48.480 when I laughed. The teenage me, I took them. I took them. That's why Derek, when he stayed here,

00:55:55.020 Elaine or anyone else when he stayed here, Conti or anyone else didn't have those. I took them. I

00:55:59.300 love them. So funny. Jill is like, you know, Rick changed every light in the guest house to red.

00:56:05.840 No, I didn't know that, but I'm not surprised. Yeah. Well, in his place, he has mostly either no

00:56:11.740 lighting or red lighting. So during the day, it just goes by ambient light and then red light in the

00:56:16.240 evening or candlelight. And it's great. And you know, people hear red lights and they think they

00:56:20.860 have to buy these expensive red light units. That's not what we're talking about. You can

00:56:24.740 literally buy red party lights or just a red bulb. Some people say, well, can I just use a red film

00:56:29.940 or can I put a t-shirt over the lamp? I worry about people putting t-shirts over lamps because

00:56:33.520 of the fire hazard. But I'll be honest, I dim the lights in my home at night. When I travel,

00:56:38.680 sometimes I will bring one of the stolen from Rick Rubin red lights.

00:56:42.300 Rick Rubin stash. Here's something where I've sort of softened my tune. So I used to be kind

00:56:46.660 of a hard liner, no blue light in the evening guy. Everything was red light at night. As far as my

00:56:53.280 phone using flux on the computer, you know, whatever it was, I suspect that that matters somewhat,

00:56:59.700 but I think what matters more is the stimulation that may come from those things. And what I've

00:57:06.260 come to realize, at least in me, which means it probably is true in others as well. And at least

00:57:11.880 some others is that what I'm doing on my phone matters more than how bright my phone is. In other

00:57:20.220 words, if I've got the best blue light filter in the world on my phone, but I'm doom scrolling social

00:57:25.800 media and getting lit up on email, that's way worse for me than if I've got my phone on maximum light

00:57:34.440 and I'm watching YouTube videos of F1 cars and driving around having fun. It's a totally different

00:57:42.420 experience. So the context matters. And I think for that reason, I would want people to be mindful

00:57:48.000 of the whole picture going to bed under a period of intense duress brought on by something that's an

00:57:55.520 equally dangerous component to all of this. That's distinct from what we're talking about,

00:57:59.540 but I want people to be able to think of this in the context of everything.

00:58:02.840 It's a really important point. One thing I'll say is that if you're going to stay up

00:58:07.940 past your normal bedtime, if you're going to get a lot of light in your eyes,

00:58:11.560 I would hope that it would be for fun reasons and for reasons you enjoy. You should definitely

00:58:15.480 spend some nights out. You should definitely do some all-nighters studying if it's going to help

00:58:19.400 you get the grade that's permanent, right? I'd certainly have done all-nighters studying and grant

00:58:23.520 writing for years. There are going to be the inevitable all-nighters due to trip to the hospital or

00:58:29.420 you heard something on the news that really amped you up or you just simply can't sleep. That stuff

00:58:33.880 is going to happen. So I think the goal should be to minimize light exposure at night. And I think

00:58:38.480 what you just said is especially true because we don't know, for instance, people talk about the

00:58:43.720 negative impact of social media. Is it the fact that people are looking at this little box for so

00:58:48.140 many hours per day? Is it all the things they're not doing? Is it what they're looking at per se?

00:58:53.540 All of those things interact and are really important. We know based on studies from the Stanford Sleep Lab

00:58:58.800 that if you wake up in the middle of the night, looking at what time it is, can be very disruptive

00:59:04.220 to your ability to fall back asleep and to your sense the next day. It's a placebo effect, but it's

00:59:09.860 a powerful one of how tired you are the next day. They've done this where they wake people up in the

00:59:15.000 middle of the night and then they say it's 4 a.m. versus 2 a.m. versus 6 a.m. and people's perceived

00:59:20.860 levels of energy during the day in some ways correlate with how much sleep they think they got.

00:59:26.520 Likewise, and this is one of the concerns, potential concerns with sleep trackers.

00:59:31.400 Ali Crumman talked about this when she came on our podcast. If people see a poor sleep score,

00:59:36.360 they often feel worse than if they see a good sleep score. Now, of course, physiology matters. You

00:59:41.500 can't lie to yourself and say you got a great night's sleep simply by virtue of a sleep score. But

00:59:45.880 I worry more about the false, if it's a false negative, we don't want to put valence on this.

00:59:50.820 Seeing a bad sleep score and then deciding that you're going to have a terrible day.

00:59:53.740 You know, I think a bad sleep score is an indication that you might need to dial some

00:59:57.380 things in a bit better. Getting a great sleep score is an indication that you might be doing

01:00:01.700 a number of things right and start looking at these things as averages. Would you agree?

01:00:05.920 Yeah, completely. I don't think it's that different from CGM. I think that CGM is an amazing tool to

01:00:11.240 provide insight and you pretty much know the insights after a relatively short period of time.

01:00:15.800 30 days, maybe at the outside 90 for a person with a very complicated life and you know all you need

01:00:22.860 to know about how the inputs affect the output. Thereafter, if you choose to use it, it's a

01:00:28.500 behavioral tool. In other words, you're using this to build in a Hawthorne effect. I think the same is

01:00:33.420 largely true with sleep trackers. Most people have this profound sense of learning when they first

01:00:39.820 encounter one of these things. And it's again, you've heard it all a hundred times. Oh my God,

01:00:43.180 I can't believe what alcohol does to my sleep. Or caloric trackers. Exactly. Like I think

01:00:47.980 Lane Norton's app Carbon, I have no financial relationship to it. I use it. It's taught me,

01:00:54.440 wow, like I consume a lot of calories in the form of certain things at certain times of day. And

01:00:58.680 there's just a lot of good learning in that. But it's the act of tracking that helps you manage

01:01:03.120 it. And similarly, I think it's the act of knowing you're going to be looking at that score that

01:01:08.440 gamifies it, that kind of helps people do the right things. Oh, you know what? I'm not going to

01:01:12.300 have that drink tonight, or I'm not going to eat that snack before bed because I've now been

01:01:16.280 conditioned to see how that impacts score. That said, I think that recovery scores and things like

01:01:22.280 that are just notoriously poor at predicting performance. And I think there's a reason that

01:01:29.000 serious athletes would never use things like that. They would tend to rely on the more tried and true

01:01:34.120 methods of predicting behavior, such as heart rate, maybe heart rate variability, but morning resting

01:01:39.520 heart rate, probably more predictive than anything else. And then in workout things such as heart

01:01:44.400 rate, heart rate recovery, lactate threshold, things like that. So yeah, I agree. I think we have to,

01:01:49.460 and I say this as a guy who's generally perceived to be the most pro device guy in the world, people

01:01:54.880 would be surprised how sparingly I use things like that. I mean, I do some tracking, not as much as you.

01:02:02.220 I love things that seem to work the first time and every time in terms of our natural biology,

01:02:09.300 based on a couple of criteria, there's an established mechanism. It's been explored in

01:02:14.300 the context of pathology, like mental health disorders, as well as pro health in healthy

01:02:19.040 individuals, that it make really good sense at the level of kind of wellness. And let's just say

01:02:25.500 ancient health. You know, when you're talking about getting a lot of sunlight during the day,

01:02:29.200 a lot of people will say, well, of course, get outside and play, not getting too much light at

01:02:33.600 night. Of course, this is just good old quote unquote, good old fashioned advice. People spend

01:02:38.260 90% of their time indoors. Now their daytime environments are too dim. Their nighttime

01:02:43.800 environments are too bright. And this misleading aspect of artificial light that when you see a

01:02:51.080 bright bulb, you think I'm getting a lot of photons is part of the problem. And the fact that when

01:02:57.080 you're out on a overcast day and it, you know, you think there's sun quote unquote, isn't out,

01:03:01.720 well, it's hidden by cloud cover, but just think about how well you can navigate that environment

01:03:07.440 without a flashlight versus at night where you would require a flashlight. We evolved under this

01:03:14.180 traumatic difference in day, night availability of photons, independent of whether or not you can

01:03:21.100 see the sun. And it's just very clear that all the mechanisms in our brain and body that regulate mood

01:03:27.420 are just powerfully regulated by this stuff. So I've made it a point to really reduce the amount

01:03:32.280 of nighttime light that I'm getting, but I am less concerned about flipping on the light switch to use

01:03:37.420 the bathroom as I used to be. I used to think, oh, I'm like quashing all my melatonin. This is

01:03:41.680 terrible. I know I can't shift my circadian clock. Then I know that that light, yes, while it's bright,

01:03:46.880 if it's brief, I'm not going to worry about it too much. Would it be better to have a dim light on

01:03:51.800 as opposed to a bright light? Sure. But I'm not going to stress it in a hotel bathroom or something.

01:03:55.320 I'm not going to walk around, you know, shielding my eyes. People will sometimes ask me, by the way,

01:03:59.240 is it different to look at the phone directly versus if you tilt the phone away? Well, it absolutely

01:04:04.620 is. I mean, think about a flashlight shown on the ground in front of you, very few photons getting

01:04:09.440 in your eyes versus shown directly into your eyes. Think about ambient light from the sun going

01:04:14.140 everywhere versus looking in the general direction of the sun. So east in the morning,

01:04:18.240 west in the afternoon, of course, the directionality of the light matters. So I'm not saying it,

01:04:23.580 you know, that you need to like peek at your phone as if you're looking over the edge of a bowl or

01:04:28.720 something into it. But my friend, Samer Hattaru is head of the chronobiology unit at the National

01:04:33.840 Institutes of Mental Health. We used to room together at meetings. We stopped because he's a terrible

01:04:37.860 snorer. There were a few times when I considered suffocating him in the middle of the night,

01:04:41.500 since he was already suffocating himself. Now we just, we don't stay in the same rooms anymore.

01:04:46.340 We're no longer postdocs, but I caught him looking at his phone in the middle of the night

01:04:51.040 and he would tilt it away. Like he's holding a platter for those that are just listening and

01:04:55.280 kind of like looking over at the screen there. I'm like, what are you doing? This is ridiculous.

01:04:58.760 He said, I'm trying not to get so much light in my eyes. That's a little extreme, but I think it

01:05:02.080 illustrates the point, which is how much direct light exposure you get at night matters. How much

01:05:06.660 direct sunlight exposure you get, especially early morning, late afternoon,

01:05:10.940 and throughout the day, it really matters. Now remind me, Andrew, what is the wavelength

01:05:15.000 of sunlight? Great. So sunlight is going to include all visible, visible, which runs from how many

01:05:20.740 to what? Well, let's answer two questions there. This wrist sensor detected 470 nanometer to 650

01:05:27.780 nanometer light. So that's going to be blue and ultraviolet. Ultraviolet's kind of blue to orange.

01:05:34.340 Yeah. Blue to orange. That's what this was measuring. So red light is going to be more like 680.

01:05:39.180 Far red is getting out to 7720 and up. Upwards of that blue light is going to fall somewhere in the

01:05:45.560 low fours. Ultraviolet is getting down into the high threes and lower. And so these spectra of

01:05:52.080 light. So during the day, you know, midday light, you're getting what looks like white light. You'll

01:05:55.520 see, oh, the sky is blue and the sun is bright white light. It's not even yellow to your eye. And of

01:06:01.100 course, don't stare at it, especially in the middle of the day, you're getting all visible spectra. So

01:06:05.340 you're getting everything from UV all the way out to red light. It's just coming in at equal

01:06:09.500 intensities. So is that a potential limitation of this study in that it didn't have a sensor that

01:06:14.580 could pick up the full spectrum of light? Potentially. We don't think of humans as UV

01:06:20.500 capable. Like we can't perceive UV light. A ground squirrel, for instance, has UV sensors in its eyes.

01:06:27.060 Turns out, you know why they use this? It's crazy. You know, when the ground squirrels sit up

01:06:30.480 on their haunches, they're actually signaling one another. They rub urine on their belly and

01:06:34.860 it reflects UV. The New York Times, for some reason, has been running a series of articles

01:06:39.640 about naturally occurring fluorescence at night and all sorts of scorpions and monotremes like

01:06:46.020 the platypus. No one really knows the reason for these odd wavelength of light emissions for all

01:06:51.120 these animals. But, you know, we view things in the blue, violet, and up to red. And we're not pit

01:06:57.900 vipers. We can't see far red. But we can see lower than 470 nanometers. And we can see higher

01:07:03.840 than 650. Is there a technology reason why they had such a narrow band in these sensors? Is it not

01:07:09.860 possible that they could have used a wrist sensor that was wider? This study was initiated in 2013.

01:07:15.280 The tech was probably far worse than it is now. Again, I would love for somebody to design an eyeglass

01:07:21.380 where it's measuring how many photons you're getting across the day. I'm not a big fan of having

01:07:27.600 everything be amplified. So I would love it if the frame would just shift color across the morning.

01:07:32.440 Like you go outside on a cloudy day, you know, you wear these glasses. And by the way, it's fine to

01:07:36.560 wear eyeglasses or contacts for sunlight viewing for setting your circadian rhythm. People always

01:07:40.940 say, well, why is that okay when a window is not? Well, corrective lenses are actually focusing the

01:07:45.300 light onto your retina. The windows and windshields are scattering the light and filtering.

01:07:49.900 And how much are sunglasses filtering this out?

01:07:52.320 Too much. We can safely say way too much, probably causing a tenfold decrement in the total luxe count

01:08:00.240 that's landing on your retina. But of course, sunglasses are important driving into sun. And some

01:08:04.920 people have very sensitive eyes. I can't sit at a cafe with a bright, brightly reflective table in the

01:08:10.760 afternoon. I just squint like crazy. I can't do it. My dad, who's darker eyed and, you know, he's

01:08:15.340 South American descent. He can just sit there just fine. My mom, who's got light eyes like me and,

01:08:20.240 you know, we're like, oh, it's really tough. You just have a terrible time. People differ in their

01:08:25.160 light sensitivity. So there's one other macro question I have here, and it's not answerable

01:08:31.440 because without randomization, we can't know it. But it's the question of how much reverse causality

01:08:39.200 can exist in these observations. So again, these observations demonstrate very tight correlations,

01:08:46.300 very strong associations, especially in the five areas that we highlighted. But it's possible that

01:08:53.120 part of what we're seeing is reverse causality brought on by both the treatments, which you've

01:08:59.040 already kind of alluded to, and also the condition itself. Do you want to explain reverse causality

01:09:03.880 for people? And maybe could you mention for those that missed the Hawthorne effect?

01:09:08.040 The Hawthorne effect is an effect that is named after an observation of what took place in a factory

01:09:13.880 where they were actually studying worker productivity with light of all things. What it

01:09:18.520 refers to is the idea that people will change their behavior when they are observed. So if I said,

01:09:25.020 well, I really want to know what a day in the life is like for Andrew Huberman, I'm going to follow him

01:09:30.040 around for a day. It's very unlikely that his behavior that day will be exactly as it was if I wasn't

01:09:36.440 there. Which is why you probably will never see a day in the life of Andrew Huberman.

01:09:40.300 Although it's pretty scripted unless I'm traveling. It's morning sunlight, hydration,

01:09:45.180 and some cardio or weight training, and then a lot of time reading papers. It'd be the most boring

01:09:49.500 video in the world because it'd be mostly me reading and underlining things.

01:09:53.080 But it's why gamifying things can be beneficial, right? It's why a CGM can be beneficial because

01:10:00.820 it's sort of like somebody's watching you and you're going to modify what you eat in response to it,

01:10:05.280 or why tracking can really be an effective way to reduce input because there's a sense of being

01:10:11.700 monitored by doing that, especially if someone literally monitors it. In other words, you can

01:10:15.420 set up an accountability partner where your health coach or someone is actually seeing the data. So

01:10:20.460 that's what it is. Now, as far as reverse causality, when you look at variables, so let's just pick a

01:10:27.940 common one that's unrelated to this. So there's an association that more diet soda consumption is

01:10:34.660 associated with greater obesity. It's a bit paradoxical. Is that true? It is. Yeah, it's been

01:10:40.980 demonstrated in many series. The greater the consumption of diet soda, the greater the prevalence

01:10:46.840 of obesity. And that has been postulated by some to suggest that non-nutritive sweeteners,

01:10:53.780 such as aspartame or sucralose or things like that, are actually part of what's causing obesity.

01:10:59.800 And while there are probably some arguments you could make around the impact that those things

01:11:03.480 might have on the gut microbiome, and maybe there's some way and that's happening, it's also equally

01:11:08.020 likely, if not probably more likely, that there's reverse causality there. That a person who is obese

01:11:13.460 is therefore contemplating how much they're eating or thinking, hey, what's an easy way that I can

01:11:19.560 reduce calories? How about instead of drinking a Coke, I drink a diet Coke?

01:11:23.780 And so there, the causality which you would impute to mean the drink is causing the obesity,

01:11:30.720 it might be no, the obesity is causing the choice of drink. So here the question is,

01:11:36.000 how much of the effect we're seeing is a result of the condition that's being studied? How much of the

01:11:43.560 disruption in both day and night light exposure is the result of the depression? It's dysregulating

01:11:50.900 the sleep. Maybe they're sleeping more during the day and more awake at night because of depression.

01:11:56.060 Again, you can't know this. This is where epidemiology never allows us to determine this.

01:12:02.340 And sadly, these questions can only be answered through either direct randomization or Mendelian

01:12:07.900 randomization, which by the way, I was going to also ask you, do you know if anyone has examined this

01:12:12.660 from a Mendelian standpoint? That would be very interesting. I don't know enough about the

01:12:17.560 biology to know what SNPs would be studyable, but that would be interesting.

01:12:22.740 Yeah. What Peter is saying is, you know, if you knew something about the genomes of these people,

01:12:28.320 you would be in a great position to perhaps even link up like sensitivity genes with genes for

01:12:34.920 pathways involved in major depression, bipolar. Getting to this issue of reverse causality,

01:12:39.720 I mean, I think it's very straightforward to imagine that the person who's experiencing a

01:12:43.900 manic episode is going to be up for two weeks at a time, sadly, and getting a lot of nighttime light

01:12:50.780 exposure. Now, nighttime dark exposure as a treatment for bipolar is something that people

01:12:55.640 are starting to talk about. So making sure that those people are awake, that they're at least blue

01:12:59.860 blocking at night, reducing their online activities, but people with severe manic episodes have a hard

01:13:04.960 time regulating their own behavior, of course.

01:13:07.420 And it's not one or the other. Like, I don't want the question to come across to the listener

01:13:11.600 that it has to be one or the other. It's only A can cause B or B can cause A. No, it's actually a lot

01:13:18.700 of times these things feed off each other. Going back to the soda example, I actually think there's a

01:13:23.240 bit of both. I actually think there's a real clear body habitus dictates beverage choice, but I also am

01:13:30.320 starting to think that in susceptible individuals, non-nutritive sweeteners will alter the gut biome,

01:13:36.840 and that alters metabolism.

01:13:39.320 What about just hunger? I remember Lane telling me, I've seen at least one of the studies that

01:13:45.280 water is probably better for us than diet soda, but that for some people, diet soda is a great

01:13:51.800 tool for reducing caloric intake. I also know some individuals, not me, who drink diet soda. I drink

01:13:59.160 diet soda from time to time, mainly stevia sweetened sodas. But what I'm referring to here are people

01:14:04.060 besides myself who drink diet soda and it seems to stimulate their appetite. There's something

01:14:09.140 about the perception of sweet as driving hunger, whereas not eating or drinking anything with any

01:14:16.280 sweetness doesn't seem to. This is one of the things I wonder if it impacts why some people like

01:14:21.240 intermittent fasting, because for some people, I wonder if the perception of sweetness, even just the

01:14:27.700 smell of food we know can stimulate appetite. So you can imagine the perception of sweetness in the

01:14:31.700 mouth, even if there's no calories there. I don't think it necessarily makes people hypoglycemic,

01:14:36.460 but perhaps it makes them think about like sweet means food. For instance, for years,

01:14:41.560 I love the combination of a diet Coke and a slice of pizza whenever I was in New York,

01:14:46.540 ideally two slices of pizza. Now, every time I have a diet Coke, which isn't that often,

01:14:51.060 but I like diet Coke, especially with a little bit of lemon in it. I just think about a slice of cheese

01:14:56.080 or mushroom or pepperoni pizza. It's like, I want it. I crave it more. So there's a

01:15:00.080 paired association there that I think is real. And we know based on Dana Small's lab at Yale,

01:15:05.340 that there's this paired association between the sweetness from sucralose and that there's an

01:15:10.620 insulin response. They actually had to cease the study in kids because they were becoming

01:15:14.760 pre-diabetic, which unfortunately meant the study was never published. Have you talked to Dana on this

01:15:19.500 podcast? I really want to get her onto my podcast.

01:15:21.580 No, we wrote a premium newsletter on this several months ago. It's got to be like,

01:15:28.060 I don't know, 10 to 20,000 words on all things related to sugar substitutes.

01:15:33.400 Okay. I'll read that.

01:15:34.100 So yeah, folks who are interested in this topic, I would refer them to the premium newsletter on

01:15:38.340 sugar substitutes. I think it was our September edition. The short of it is the data are a little

01:15:43.700 bit noisy, but there is indeed some sweeteners in some studies do result in that phenomenon you

01:15:48.960 described, the cephalic insulin response. And I came away from the research that went into that,

01:15:54.320 which was Herculean effort on the part of the team, a little bit more confused than when I went

01:16:00.920 in, but being even more cautious around artificial sweeteners than I was going in. And not for the

01:16:08.120 reasons, I didn't find any evidence that these things are cancer causing. So that's the headline

01:16:12.920 stuff people worry about.

01:16:13.900 You have to ingest like 10 grams of it or something crazy like that.

01:16:17.320 I came away more confident that from a long-term safety perspective in terms of cancer and

01:16:23.200 catastrophic outcomes like that, that wasn't the issue. But I came away much more cautious

01:16:27.200 around these things can really be mucking around with both your brain chemistry and your gut

01:16:32.640 chemistry, which can pertain to your metabolism. And therefore my takeaway was buyer beware,

01:16:39.880 use limited amounts only.

01:16:41.800 The one, by the way, that still emerged to me as a reasonable one, it's the only one that I use,

01:16:46.820 and I've talked about a lot is xylitol, xylitol for chewing. So for gum and allulose as an

01:16:53.240 additive safer. You're saying, yes, those are basically the only two I will consume.

01:16:59.020 I'll drink a diet Coke every now and again, if I'm on a plane or something, this law that got

01:17:03.320 passed a few years ago, you couldn't bring liquids of your own into the airport and on the plane.

01:17:07.380 What a great scheme to get people to buy overpriced fluids in the airport. I mean,

01:17:13.940 there are more important issues in the world, but this one really gets me. But I drink things

01:17:18.420 with a little bit of stevia in it, the occasional diet Coke. I generally avoid sucralose. I don't

01:17:24.000 like the way it tastes. Monk fruit's too sweet, but maybe we'll do a podcast on that in the future.

01:17:28.420 Okay. So I think we can wrap this paper.

01:17:30.560 But tell me what you think about that point. You know more about this stuff than I do. But

01:17:34.800 if you had to just lay on your judgment, if it were a hundred to zero, you would say the light is

01:17:42.000 100% causal in the effects we're seeing. If it were zero to a hundred, you'd say, nope,

01:17:49.220 the behavior is 100% causal of the exposure to light. Again, you can't know it, but what does your

01:17:55.640 intuition tell you?

01:17:56.460 Okay. There's my intuition. And then there's my recognition of my own bias because I started

01:18:02.000 working on these circadian pathways originating in the eye back in 98 as a graduate student at

01:18:08.580 Berkeley. The cells, these melanopsin intrinsically sensitive retinal ganglia cells were discovered

01:18:12.880 in the early 2000s by a guy named Iggy Provencio, Dave Burson, Sam Rattar, Sachin Pant, and others.

01:18:18.900 And it was like one of the most important discoveries in all of biology, clearly. So I've been very

01:18:24.060 excited about these systems, but if I set that aside, so bias disclosure made, I think 65 to 75%

01:18:32.560 of the effects are likely due to light directly. Now it's impossible to tease those apart, as you

01:18:39.180 mentioned, but to play devil's advocate against myself, you could imagine that the depressed

01:18:44.240 individual is laying around indoors with the curtains drawn. They didn't sleep well the night

01:18:49.360 before, which gives you a photosensitivity that isn't pleasant. Like it sucks to have bright light

01:18:53.800 in your eyes first thing in the morning, especially if you didn't sleep well. And then they're, you

01:18:58.780 know, making their coffee in a dimly lit, what they think is brightly lit environment. And then they're

01:19:03.160 looking at their phone and the state of the world sucks and their state of their internal landscape

01:19:08.080 is rough. And maybe they're dealing with a pain or injury or something. And their likelihood of

01:19:14.880 getting outside is low. And when they do get outside, they're going to shuffle and up. So I

01:19:19.500 could see how the behaviors could really limit the amount of light exposure. And then evening rolls

01:19:23.840 around. They've been tired all day and a common symptom of depression, you fall asleep. And then

01:19:28.000 two or three in the morning, they're wide awake. What are you going to do at two or three when you're

01:19:31.600 wide awake, sit in the dark? No, you're going to get online. You're going to listen to things you

01:19:36.880 might have. I'm not recommending this, but an alcoholic drink in order to try and fall asleep. I mean,

01:19:41.500 this is the pattern. Shaking up that pattern is really what so much of my public health work

01:19:47.440 these days is about and trying to get people onto a more natural daylight, night, dark rhythm. But

01:19:52.800 yeah, it's impossible to tease apart. We do know this, and this is really serious. We know that in

01:19:58.240 almost every instance, almost every psychopathology report of suicide in the weeks, but especially in

01:20:05.500 the days preceding suicide, that person's circadian rhythms looked almost inverted from their normal

01:20:12.420 patterns. And that's true of non-bipolar individuals as well. Circadian disruption and

01:20:19.600 disruption in psychiatric health are inextricable. Conversely, positive mood and affect and circadian

01:20:26.540 behavior seem very correlated. I mean, I think it's clear that if you want to become an early riser,

01:20:32.800 get light in your eyes and get activity in your body early in the day, you entrain to those rhythms

01:20:38.340 so that you start to anticipate that morning workout. You start to anticipate the morning

01:20:41.780 sunlight. Just one more scientific point. We know that when you view bright sunlight in the morning

01:20:47.200 or just sunlight that's illuminating your environment, as you said, you don't even have

01:20:50.120 to see the sun itself, that there's a 50, 5, 0% increase in the amplitude of the morning cortisol

01:20:56.840 spike, which is a good thing because the amplitude of the morning cortisol spike is inversely

01:21:02.200 related to the amplitude of the evening cortisol spike and high evening cortisol is associated with

01:21:06.680 middle of the night waking and on and on. So, you know, I'm very bullish on these mechanisms.

01:21:13.080 I also love that they're so deeply woven into our evolutionary history, you know, that we share with

01:21:18.000 single cell organisms. It's so wild. But of course, there's going to be a bidirectionality there and

01:21:23.920 it's impossible to see where one thing starts and the other one stops.

01:21:27.060 Here's my take, Andrew. First of all, I actually, with far less authority than you,

01:21:32.760 agree with your assessment and might even be a little bit more bullish, might even put it at 80-20.

01:21:38.180 And here I'll give you my explanations, which stem more from my fastidious battles with

01:21:45.420 epidemiology in general. Because so much of the world that I live in still has to rely on

01:21:50.940 epidemiologic data. And so how do you make sense of it? The truth of it is most of it is really

01:21:56.420 pretty bad. I tend to find myself looking at the Austin Bradford Hill criteria all the time.

01:22:02.960 And for folks who don't know, he was a statistician who basically proposed a set of criteria.

01:22:09.020 I believe there are eight of them. And I can't believe I don't know every one of them off by

01:22:12.200 heart. I certainly used to. But the more of these criteria that are met within your correlations,

01:22:18.520 the more likelihood that you will find causality. So when I think of your data here,

01:22:23.860 the data in this paper, I'll tell you what makes these correlations seem to have causality

01:22:30.480 within them in the direction that's being proposed. Look at the dose effect. So dose effect

01:22:36.420 matters. And this is done in quartiles. And that's a very elegant thing. If they just did it as

01:22:41.260 on off, it would be harder.

01:22:43.240 High low.

01:22:43.820 That's right. But the fact that they did it in quartiles allows you to see that every example

01:22:47.860 in figure two, I don't believe there is an exception to this. There's only one exception

01:22:52.040 to what I'm about to say. Sorry, two out of like God knows how many. They're all monotonically

01:22:57.240 increasing and decreasing. In other words, the dose effect is always present. Another thing is

01:23:02.920 biologic plausibility. You've spoken at length about that today. So in other words, sometimes you have

01:23:08.740 to look at epidemiology and ask, is there a biologic explanation? And here there is. You've added

01:23:13.540 another one, which is evolutionary conservation to the biologic plausibility. Then you can talk

01:23:18.260 about animal models or experiments in humans over short durations that generally support these

01:23:24.340 findings. And so those are just a couple of the Bradford Hill criteria that lead to my belief

01:23:31.280 that, yeah, there's reverse causality here, but it's not the full explanation and that more

01:23:37.120 of the explanation is probably the direction that's being proposed. At the end of the day, like

01:23:41.380 what's the purpose of the discussion? The purpose of the discussion is if you are under the influence

01:23:46.840 of any of these psychiatric conditions, in addition to the treatments you're doing now, what else can

01:23:52.660 you do? And to me, the takeaway is follow these light behaviors. That's a relatively low lift. When you

01:23:59.440 consider some of the things, like I'm over here asking people to do zone two for three hours a week

01:24:03.920 and VO2 max workouts and all this other stuff. And I think all those things matter for mental health

01:24:09.160 as much as physical health, but this strikes me as on the spectrum of low asks, if it's only even 30%

01:24:18.320 causality, 70% reverse causality, like I'll take those. I would still instate that.

01:24:23.700 Yeah. And it's taking your coffee on the balcony. It's people will often say, well, how do you do this

01:24:28.220 with kids? The kids should be doing it too. You know, it means popping your sunglasses off. It means getting

01:24:33.760 out for just a few minutes. And the fact that it's additive, that these photon counting mechanisms, they sum.

01:24:39.640 This paper also says, and I should have stated this earlier, if you missed your daytime light

01:24:43.720 ration, get your nighttime dark ration. They are independent and additive. I mean, that's a really

01:24:50.340 something, but of course, ideally you get both, but I appreciate your take on it. Thanks for your expertise in

01:24:56.180 parsing epidemiology. I look at fewer studies of that sort, but I learned from you. And that's one

01:25:01.660 of the reasons I love doing these journal clubs is I learn. So along those lines, tell us about the

01:25:07.160 paper you selected. I'm really eager to learn more. Well, I wanted to pick a paper that was

01:25:14.100 interesting as a paper. This paper I think is interesting in that it is the landmark study of a

01:25:21.260 class of drugs. But in the same way that you kind of picked a paper that I think has a much broader

01:25:26.740 overarching importance, the reason I picked this paper, which is from the New England Journal of

01:25:32.600 Medicine, it's about 13 years old, is because it is kind of the landmark study in a class of drugs that

01:25:39.920 I believe are the most relevant class of drugs we've seen so far in cancer therapy. And even though the

01:25:47.980 net effect of these drugs has only served to reduce mortality by maybe 8 to 10 percent, which is not a

01:25:56.560 huge amount, it's the manner in which they've done it that gives me great hope for the future, even if

01:26:03.240 it's through other means. So I'll take a step back before we go into the paper for, again, just the

01:26:08.100 context and background. So the human immune system is kind of a remarkable thing. It's hard when you're

01:26:15.920 trying to imagine what's the most amazing part of the human system. And maybe it's my bias as well,

01:26:21.200 because just as you spent your time in the light system and the photosensing system, I spent my time

01:26:25.900 in the immunology world. But it is remarkable to me how our immune systems evolved. And they have this

01:26:32.620 really brutal task, which is how can they be tuned to detect any foreign pathogen that is harmful

01:26:44.560 without knowing a priori what that could be? So in other words, how can you tune a system to be so

01:26:52.320 aggressive that it can eradicate any virus or bacteria billions of years into the future without

01:26:58.620 knowing what it's going to be? But at the same time, it has to be so forgiving of the self that it

01:27:05.740 doesn't turn around and attack the self. It's remarkable. And of course, we can always think of

01:27:10.960 the exceptions. There are things called autoimmune conditions. So clearly the system fails and the

01:27:15.900 immune system turns around and attacks the self. If you see a person with vitiligo, I have a little

01:27:20.800 bit of vitiligo on my back, a couple of spots. Clearly the immune system is attacking something

01:27:26.580 there and destroying some of the pigment. I didn't realize vitiligo was autoimmune.

01:27:30.900 Yeah. There are lots of more serious autoimmune conditions. Of course, you know, somebody that has

01:27:34.980 lupus or immune system can be attacking the kidney. The immune system can be attacking any

01:27:38.900 autoimmune conditions can be deadly, but fortunately they are very rare. And for the most part, this

01:27:44.340 immune system works remarkably well. So how does it work and why is it that cancer seems to evade it

01:27:53.680 virtually all of the time? This is the question. Let's first of all, talk about how it works.

01:27:59.240 And then when I tell you how it works, you'll say, that sounds amazing. Clearly it should be able to

01:28:05.040 destroy cancer. I'm going to simplify it by only talking about one system, which is how T cells

01:28:11.420 recognize and get activated, how T cells recognize antigens. So we have something called an antigen.

01:28:18.900 So an antigen is an antibody generating peptide. So it's a protein, almost always a protein. They can

01:28:26.700 be carbohydrates, but they're almost always proteins and they're very, very small peptides. Like we're

01:28:31.060 talking as little as nine amino acids, maybe up to 20 amino acids. So teeny tiny little peptides,

01:28:38.120 but it's amazing that in such a short peptide, the body can recognize if that's Andrew or not Andrew.

01:28:46.540 We talk about proteins in kilodaltons. We're talking about proteins in terms of thousands of amino acids

01:28:52.980 that make up every protein in your body. And yet if it samples a protein and sees that, Hey, this little

01:28:59.380 nine, 10, 15 peptide amino acid is not part of you. I know it's bad. And therefore I'm going to generate

01:29:08.040 an immune response to it. So we have what are called antigen presenting cells. You have cells that go

01:29:15.120 around sampling peptides and they will on these things called MHC class receptors, bring the peptide

01:29:24.000 up to the surface and serve it up to the T cell. There are two types of these. There's MHC class one

01:29:30.200 and MHC class two. This is major histocompatibility complex. That's correct. We refer to them that way

01:29:37.920 because of the context in which they were discovered, which was for organ rejection.

01:29:42.360 So not surprisingly, when you need to put a kidney into another person, if that kidney is deemed foreign,

01:29:49.780 it will not last long. And the early days of organ transplantation were rife with immediate

01:29:54.780 rejections. The immediates are the ABO incompatibilities, but the sort of next layer of

01:30:00.380 incompatibility was MHC incompatibility, which would lead to within weeks, the organ is gone as opposed to

01:30:05.680 within hours. So you have these two classes of MHC. You have class one and class two. Class one is what we

01:30:12.780 call endogenous. So this is basically what happens when a protein or an antigen is coming from inside

01:30:21.440 the cell. So let's consider the flu. So if you get the flu, the influenza virus infects the respiratory

01:30:28.700 epithelium of your larynx. And that virus, as folks listening might remember from our days of talking

01:30:35.940 about COVID, viruses can't replicate on their own. What they do is they hijack the replication

01:30:41.760 machinery of the host, and they use that either to insert their RNA or DNA to replicate. And in the

01:30:48.620 process, proteins are being made. Well, those proteins are the proteins of the virus, not of us.

01:30:53.900 So some of those peptides get launched onto these MHC class one gloves. Basically the glove comes up to

01:31:01.060 the surface and a T cell comes along. And in the case of MHC class one, it's a CD8 T cell. These are

01:31:08.620 what are called the killer T cells, right? And so this cell comes along and with its T cell receptor,

01:31:14.960 the T cell receptor meets the MHC class one receptor with the antigen in it. And if that's a lock,

01:31:22.380 it realizes that's my target. And it begins to replicate and proliferate and target those. And that

01:31:29.460 creates the immune response. And by the way, that's how it works when you vaccinate somebody. You're

01:31:33.580 basically pre-building that thing up. So would this fall under the adaptive immune response or the

01:31:39.120 innate immune response? So this is adaptive. Innate is just these pure antibody response on the B cell

01:31:44.240 side. I won't get into that for the purpose of this discussion. The other example is MHC class two,

01:31:51.040 and that's also part of the adaptive system or the innate system, which is more what we call the

01:31:57.260 exogenous form. So these are peptides that are usually coming from outside the cell.

01:32:01.880 So we're going to focus more on the MHC class one, because this is peptides that come from inside

01:32:07.360 the cell. Okay. So just keep in the back of your mind, if a foreign protein gets presented from inside

01:32:13.360 a cell to outside a cell, the T cells recognize that and they will mount a foreign response. And by the

01:32:18.740 way, that's why we basically can beat any virus. If you consider how many viruses are around us,

01:32:25.000 the fact that we almost never die from a viral infection is a remarkable achievement of how well

01:32:31.340 this immune system works. Constantly combating these viruses. Constantly. And by the way,

01:32:35.880 we don't really have very effective antiviral agents. It's not like antibiotics. Like we have

01:32:40.520 antibiotics up the wazoo. We're way better at fighting viruses than bacteria. I've always wondered

01:32:46.280 about this. To what extent is our ability to ward off viruses on a day-to-day basis as an adult,

01:32:52.280 reliant on us having been exposed to that virus during development. Like as I walk around today,

01:32:58.320 maybe I'll be exposed to a hundred thousand different viruses. Would you say that half of

01:33:03.080 those, I have already got antibodies too, because I was exposed to them at some prior portion of my

01:33:07.220 life? Yeah. Hard room to quantify. And the other ones, I'm just building up antibodies. Like I was on

01:33:11.600 a plane last night, someone was coughing. So I was hiding and I had COVID a little while ago. So I wasn't

01:33:16.200 too worried about that. And I feel great today, but I just assume that on that plane, I'm in a swamp

01:33:23.580 of viruses, no matter what. And that most of them I've already been exposed to since I was a little

01:33:29.740 kid. So I've got all the antibodies and they're just fighting it back, binding up those viruses and

01:33:33.940 destroying them. Yeah. I think it's part that. And I also think it's part of them that our body can

01:33:38.320 destroy without mounting much of an immune response. So therefore your immune system is doing

01:33:42.480 the work yet. It's not mounting a systemic inflammatory response that you're not sensing.

01:33:46.820 So is it also a physical trapping in, you know, in my nasal epithelium?

01:33:50.760 Yeah. You have huge barriers, right? So the skin, the hairs in your nose, all of these things are

01:33:56.420 huge barriers, but assuming that still a bunch of them are getting in, at least the respiratory ones,

01:34:01.460 that's the other thing to keep in mind, right? There were certain viruses that are totally useless

01:34:04.740 floating around the air. The viruses that most people are really afraid of hep C, hep B, HIV. Well,

01:34:10.800 if they're sitting on a table or floating around the air, they're of no threat to you. They have to

01:34:15.100 be transmitted through the barrier. But again, some of these viruses you're going to defeat without

01:34:20.360 an enormous response. And then some of them, why is influenza, quote unquote, such a bad virus?

01:34:26.320 Whereas the common respiratory cold kind of sidelines you for a day. It's the immune response

01:34:31.940 that you're feeling. The bigger the immune response to the virus, the more you're feeling that. You feel

01:34:37.620 your immune system going crazy. The interleukins that are spiking, the third spacing that occurs

01:34:44.080 to get more and more of the immune cells there, the spike of your temperature as your body basically

01:34:48.840 tries to cook the virus. All of that stuff is your body.

01:34:51.840 The fatigue.

01:34:52.760 Yeah. Yeah. You're being drained and all this happening. So one more point I'll mention just

01:34:57.100 to close the loop on the autoimmunity. How is it that we learn not to attack ourselves? That's

01:35:01.980 something called thymic selection that occurs in infancy. So you and I have a no good for

01:35:08.740 nothing tiny little thymus that would be, it's almost impossible to see these things. You know,

01:35:13.120 when we used to operate on people, the thymus is barely visible in an adult, in a healthy adult

01:35:18.480 outside of thymic tumors. And a child, the thymus is quite large. And the purpose of the thymus

01:35:23.160 is to educate T cells and basically show the T cells what self is. And any T cell that doesn't

01:35:31.760 immediately recognize it gets killed. It's a really clever system where we basically teach you

01:35:38.260 to recognize self at a very early age. And if you can't do that, you're weeded out. And then

01:35:44.180 the thymus involutes thereafter because it's sort of served its purpose.

01:35:47.560 Okay. Now let's talk about cancer. So what do we know about cancer? So we know that, again,

01:35:53.700 cancer is a genetic disease in the sense that every cancer has genetic mutations.

01:35:59.660 Most of those mutations are somatic, which means most of those mutations are mutations that occur

01:36:06.660 during the course of our life. They're not germline mutations.

01:36:10.000 The germline being the eggs and sperm, right? So it's all other cells. And I love that you pointed out

01:36:16.360 that cancer can be genetic, but isn't necessarily inherited. People hear genetic and they think

01:36:21.640 inherited. Inherited is always genetic to some extent, but genetic isn't always inherited.

01:36:27.820 Yeah. So there are a handful of cancers that are derived from inherited mutations. So Lynch

01:36:35.140 syndrome is an example of that. Hereditary polyposis is an example of that, where you have a gene

01:36:43.000 that gets passed through the germline and that gene codes for a protein, like all genes do. And it's

01:36:49.520 either you have too much of a gene or too little of a gene. So it's either a gene that promotes cancer

01:36:53.920 and you have too much of that, or it's a gene that prevents cancer and you have too little of it or a

01:36:59.160 dysfunctional version of it, right? So BRCA is an example of that. BRCA is hereditary. BRCA codes for a

01:37:05.880 protein. And the women and men, but mostly the women that we think about who have a BRCA mutation,

01:37:11.780 that in some cases almost guarantees breast cancer, it's because of a defective copy. So it's like

01:37:18.600 they don't get the protein that they need to protect them from breast cancer. So what do we know? Well,

01:37:23.500 we know that, and this is probably one of the most remarkable things I've ever learned, and it still

01:37:31.400 blows my mind every time. Well, actually, before I get to that point, I want to make another point.

01:37:35.520 Okay. So cancer, our cells become cancerous, but they're clearly hijacked because they have these

01:37:41.500 mutations. And as a result of these mutations, they make proteins that allow cancers to behave

01:37:47.780 differently. And cancers behave differently from non-cancers in two very critical ways.

01:37:54.260 The first way is that they do not respond to cell cycle signaling. So if you cut your skin,

01:38:00.560 it heals, but how does it know to heal just right and not to keep growing and growing and growing and

01:38:06.840 growing and growing? Well, it knows that because there are cell cycle signals that tell it time

01:38:11.220 to grow, time to stop. Believe it or not, this is an extreme example. If you donated to me half of your

01:38:16.840 liver, which I know you would. Absolutely. I'd give you more than half of my liver. If it meant that we

01:38:22.200 could keep doing these journal clubs, right? Yeah, yeah, yeah. Within months, you would regenerate a full

01:38:28.340 liver. Isn't that amazing? That's so wild. It's like a salamander. You cut off a salamander limb

01:38:32.520 and please don't do that experiment because other people are doing it anyway. And it grows back.

01:38:36.980 And it knows how much to grow back. When the cell is perfectly functioning, it knows how much to grow.

01:38:43.560 Well, cancer loses that ability. That is one of the hallmarks of cancer. It just keeps growing.

01:38:48.700 It doesn't grow faster, by the way. That's a misnomer. People think cancers grow faster than

01:38:52.420 non-cancers. There's no real evidence that that's the case. They just don't stop growing.

01:38:55.920 The second property of cancer is the capacity to leave the site of origin, go someplace else and

01:39:04.520 take up residence. So that's metastasis. That's the metastasis component. So if you think about it

01:39:08.620 for a minute, a cell that never stops replicating and has the capacity to up and leave and move and

01:39:15.480 take up residence is clearly different from the cell itself. So if I have a cell of colonic epithelium,

01:39:21.320 the cell that lines the inside of my colon, clearly got a set of proteins in it. But if all of a sudden

01:39:26.240 that thing can grow, grow, grow, grow, grow, not stop, not stop, not listen to the signal, and then

01:39:32.320 somehow wind its way into the liver and just keep growing and growing and growing, it must have

01:39:36.600 different proteins. So the question then becomes, why does cancer even exist? How has our immune system

01:39:43.880 not figured out a way to just silence this and eradicate it the way it does to virtually every

01:39:50.140 virus you encounter? And to me, this is one of the most interesting questions in all of biology,

01:39:54.900 and it really comes down to how clever cancer is, unfortunately, how evolutionarily clever it is.

01:40:02.460 It basically does a lot of things to trick the immune system. So it has its own secretory factors

01:40:10.440 that tamp down the immune system. It grows in an environment because of its nature. So one of

01:40:17.800 the things that's long understood about cancer is it's heavily glycolytic. And when something is

01:40:23.180 heavily glycolytic, it's going glucose to pyruvate to lactate nonstop. There are lots of reasons for

01:40:29.460 that. I think there's more than one. What does that afford it? Is that afforded a migratory

01:40:34.280 potential? No, so it's super interesting. So that's the effect that what I just described is called the

01:40:37.700 Warburg effect. And when Warburg proposed this, which God was probably in the 1920s, it was before

01:40:45.840 World War I, before World War II, he proposed it. He thought the mitochondria of cancer cells were

01:40:51.620 defective. So he proposed that the cancer cells mitochondria don't work. Hence, they have to

01:40:57.520 undergo glycolysis. They can't undergo aerobic metabolism. We now know that that's not the case.

01:41:03.180 So we now know that the Warburg effect or the Warburg effect, if I'll refer to him correctly by

01:41:08.140 his name, almost assuredly does not have to do with defective mitochondria. Others have proposed

01:41:12.900 several mechanisms. I think there's probably more than one thing going on. So a paper that came out

01:41:17.100 in 2009, very influential paper by a guy named Matt Vander Heiden, Craig Thompson, and Lou Cantley

01:41:23.780 proposed that the reason that cancer cells do the Warburg effect is that they're not optimizing for

01:41:30.200 energy. They're optimizing for cellular building blocks. And if you do the mass balance, it completely

01:41:34.500 makes sense. Dividing cells need building blocks more than energy. And glycolysis, while very

01:41:41.640 inefficient for generating ATP, is much more efficient at generating substrate to make more cells.

01:41:46.860 But another proposed mechanism is exactly at this one. Glycolysis lowers the surrounding pH

01:41:52.100 because of lactate. Lactate attracts hydrogen, pH goes down. And guess what that does to the immune

01:41:56.760 system? Detracts the immune system. So it's also a way to hide from the immune system.

01:42:02.540 Like a pH cloaking, leveraging pH to cloak the signal that the immune system would otherwise see.

01:42:08.340 Yep. And then when you layer on top of that, that it knows how to secrete things like IL-10,

01:42:12.960 TGF-beta, all of these other secretory factors that also inhibit the immune system. Basically,

01:42:19.140 it's figured out a way to kind of hide itself from the immune system.

01:42:21.680 The way you describe it, cancer sounds like a virus.

01:42:25.180 Yes.

01:42:25.520 It sounds a lot like a virus. And that leads me to ask,

01:42:28.220 are there any examples of contagious cancers? I recall seeing some studies about these little

01:42:33.340 critters down in Australia, Tasmanian devils, that they would scratch each other and fight,

01:42:38.840 as Tasmanian devils do. They're actually quite cute. And they would get cancers and tumors growing

01:42:43.520 on their faces. It was like a literal physical interaction that could transmit cancer from one

01:42:48.740 animal to the next. So it's less that there are viruses that cause cancer. So in that sense,

01:42:54.960 you could argue, yes, there are contagious cancers.

01:42:58.020 Well, HPV.

01:42:59.060 Sure. Yeah. HPV, hep B, hep C. But there are even cancers like cutaneous cancers that arise from viruses.

01:43:06.560 But I don't know if that's quite the same as what you're saying.

01:43:09.900 What you're saying is an important point. We don't want to go down the rabbit hole of the HPV,

01:43:13.780 but right, that's increasing susceptibility to cervical cancer. Now there's a vaccine against

01:43:18.740 HPV, right? There wasn't when we were in college, as we all knew, there was no vaccine. But direct

01:43:23.920 transmission of cancers from one organism to the next, more rare.

01:43:28.260 Yes. So now, a moment ago, I said there's this really incredible thing about cancer that blows my

01:43:34.300 mind and about our immune system, which is that at least 80% of solid organ tumors, and we're going

01:43:41.120 to mostly talk about solid organ tumors because that's where the field of oncology has made very

01:43:46.500 little progress. So if you go back 50 years, where has oncology made huge progress? It's made great

01:43:51.940 progress in blood tumors, leukemias, and some kinds of lymphomas. In fact, there's two kinds of

01:43:58.860 lymphomas where the progress has been remarkable. One has been in Hodgkin's lymphoma, and the other

01:44:04.440 has also been in immunotherapy, has been in a type of B-cell lymphoma, where that B-cell demonstrates

01:44:10.240 or presents something called a CD19 receptor. So in B-cell lymphomas with CD19, there's a very unique

01:44:17.240 niche immunotherapy, we won't talk about that today, called CAR-T therapy that has got rid of

01:44:21.820 those guys. And then leukemias have also been pretty good. But in solid organ tumors, there have

01:44:27.880 been only two real breakthroughs in the last 50 years. One has been the therapy for a certain type

01:44:35.140 of testicular cancer, and it's really just a chemotherapy cocktail that has been found to work

01:44:39.480 really well. And the other has been in this really rare kind of gastric cancer called the GI stromal

01:44:44.500 tumor, which happens to result from one mutation in a kinase pathway. And there's one drug that can now

01:44:50.980 target that, and it works. It's kind of amazing. Cures that cancer. Cures that. What I'm talking about

01:44:56.200 are the cancers that kill virtually everybody else. When you sort of line up, what are the big causes of

01:45:02.780 cancer death? Let's start at the top. It's lung. It's then breast and prostate in men and women.

01:45:09.600 It's colorectal. It's pancreas. Those are the big five. More than 50% of cancer deaths in Americans

01:45:15.160 come from those five. These are what we call the solid epithelial tumors. And you can march down the

01:45:20.160 list, and most cancers that most people are thinking of are those cancers. Well, here's the thing. More than

01:45:24.200 80% of those cancers have antigens that are recognized by the host's immune system. I will state it

01:45:31.440 again because it is so profound. 80% at least of those cancers actually generate an antigen, meaning

01:45:39.820 a little peptide in that cell gets presented to the T cell and it is recognizable. And now the question

01:45:49.200 is, why is that not sufficient to induce remission? And the short answer is there are not enough T cells

01:45:59.060 that are able to act and or they are being sufficiently inhibited from acting, which gets me to the point

01:46:06.120 of this paper. One of the ways in which the body inhibits the immune system, which we should remind

01:46:13.900 ourselves is an important thing, right? Is something called the checkpoint inhibitor. Okay. So go back to

01:46:20.760 that idea that I talked about before. You have an antigen presenting cell. It brings up an MHC receptor

01:46:27.960 with a peptide on it, and there is a T cell that is coming. And I actually brought a diagram, which

01:46:33.760 I'm going to link to this. I don't want to make this too complicated, but I really think that this

01:46:38.480 figure is helpful to understand how these drugs work. So the MHC receptor with the peptide is sitting

01:46:44.300 there and it binds to the T cell receptor on the T cell, but there is another receptor on the T cell,

01:46:51.060 a CTLA-4 receptor. And that binds to a receptor that I won't bother naming now. The names don't

01:46:59.440 matter. But there's another receptor on the antigen presenting cell that binds to that. And that acts

01:47:05.840 as the brakes in the reaction. So CTLA-4, which is on the T cell, binds to another CD receptor on the

01:47:15.800 antigen presenting cell. And it says, tamp down the response. And the reason for that is we want to

01:47:23.480 keep our immune system in check. This basically is a way of asking the immune system. Because remember,

01:47:27.660 when the immune system sees that antigen, it wants to go nuts. It wants to start replicating and

01:47:33.280 killing. This is a CD8 T cell. It is a targeted killer T cell. The checkpoint says, let's double check

01:47:40.580 that. Let's be sure. Let's tamp down the response. And as a result of that, a thought experiment emerged,

01:47:47.260 which was, what if we block CTLA-4? What if we block the checkpoint? Could we unleash the immune

01:47:55.140 system a little bit more? And I will say this, at the time it was proposed, it seemed a bit far-fetched.

01:48:03.780 Because of the complexity of the immune system, it seemed a little far-fetched that simply blocking

01:48:09.600 the checkpoint would have any effect. It's also worth noting that prior to this, one immunotherapy

01:48:17.360 had found some efficacy, which was trying the exact opposite strategy. Rather than blocking the

01:48:24.220 inhibitor, it was throwing more accelerant at the fire, which was giving something called interleukin-2.

01:48:30.700 So interleukin-2 is, for lack of a better word, candy and fuel for T cells. So the idea was,

01:48:38.340 if we have T cells that innately recognize a cancer antigen, can we just give high doses of interleukin-2

01:48:46.220 and have them undergo proliferation and response? And the answer turned out to be yes, but only in

01:48:53.460 two cancers, melanoma and kidney cancer, and only at very small levels. About 10% of the population

01:48:59.720 would respond to these things. Now, look, that's 10% of people who were going to be dead within six

01:49:04.360 months, because these are devastating cancers. And once they spread, there are no treatments that

01:49:09.800 have any efficacy whatsoever. In fact, I think median survival for metastatic melanoma at the

01:49:14.120 time was probably four months. So this was a very grim death sentence. But the idea now was,

01:49:20.380 what about doing the exact opposite approach? Instead of trying to throw more fire at the T cell,

01:49:26.960 what if we can take its brakes down? Instead of giving more gas, let's give less brakes.

01:49:31.380 And there were some phase one studies that demonstrated efficacy phase two. And the paper

01:49:35.700 I'm going to talk about today is the phase three study that compared the first version of these. So

01:49:42.460 the drug we're going to talk about today is an anti-CTLA-4 drug called ipililumab. There is another

01:49:49.280 drug out there that came along shortly thereafter that is an anti-PD-1 drug. So PD-1 turns out to be

01:49:57.740 another one of these checkpoints on T cells. And the Nobel Prize, by the way, I think it was 2018 or

01:50:05.660 2019 in medicine or physiology, was actually awarded to the two scientists who discovered

01:50:11.960 CTLA-4 and PD-1. So I believe this is the only Nobel Prize in medicine for immunotherapy. It's a very

01:50:18.920 big deal. So this study sought to compare the effect of anti-CTLA-4 to a placebo. And the placebo in this

01:50:30.520 case was not a real placebo. It was a peptide vaccine called GP-100 to ask the question, in patients with

01:50:39.260 metastatic melanoma, what would be the impact on median survival and overall survival? Let's talk a

01:50:48.900 So again, one of the funny things about this, I used to read these papers a lot, Andrew. These

01:50:54.080 used to be my bread and butter papers, reading these like it's my hobby. And I don't read them

01:50:58.700 that much anymore. So it was kind of amazing how long it took me to remind myself of stuff I used

01:51:04.340 to remember. But you do have to kind of go back and read the methods and figure out who were the

01:51:08.900 patients in this? What was the eligibility criteria? Why did they do it this way? And of course,

01:51:14.280 it all kind of came back to me, but it took a minute. So the first thing is these are all patients

01:51:20.160 who had progressed through every standard therapy. So these are patients for whom there were no other

01:51:26.600 options. These patients either had very advanced stage three melanoma, which means it was local

01:51:34.260 regional melanoma, but it couldn't be resected. So an example of that would be a cancer that was

01:51:42.380 completely engulfing. Let's say the primary site was the cheek and it had completely grown into all

01:51:49.540 of the surrounding soft tissue. It hadn't spread anywhere, but it was all the lymph nodes of the

01:51:54.860 neck. And I've seen patients like this and it's just completely disfiguring. And they'd already been

01:51:59.880 through the standard chemotherapy and nothing was working and the thing was growing. And then it was

01:52:04.500 mostly made up of patients with stage four cancer. Now melanoma has a very funny staging system.

01:52:10.100 In cancer, we typically talk about something called the TNM staging system. It is the standard way that

01:52:16.660 cancers are staged. T refers to the tumor size, N refers to the lymph node status, and M refers to

01:52:24.060 the presence or absence of metastases. And for most cancers, it is a very simple system. T is typically

01:52:31.280 a number one, two, sometimes up to three and four. N is typically zero, one, or two, and M is zero or

01:52:37.720 one. Either there's no mets or there are mets. So for example, in colorectal cancer, the T staging

01:52:43.080 determines the depth in the colon wall that it went. N is, did it go to mets? And I think in colon,

01:52:49.900 I'm a little rusty on this. I think colon has N zero, one, or two, depending on how many lymph nodes.

01:52:55.000 And then M zero, did it go to anything beyond that, like to the liver, lung, et cetera, or not?

01:53:00.040 Melanoma is a bit more complicated. It has M zero, meaning no mets, but it also has M one A,

01:53:07.420 M one B, M one C, and M one D. And within each of those, it has a threshold for high and low lactate

01:53:16.420 dehydrogenase or LDH. So it's both a staging based on imaging and biochemical. And the reason for that

01:53:22.900 is LDH level is such a strong prognostic indicator of survival. In addition to M staging,

01:53:28.620 higher LDH levels tend to reflect more acidity, which we talked about why that's problematic,

01:53:34.200 tends to reflect faster growing tumors, higher turnover, higher metabolic activity.

01:53:40.120 M one A, let me see if I can remember this. M one A's are cancers that have metastasized

01:53:47.520 to surrounding soft tissue or soft tissue anywhere in the body. So anywhere else on the skin.

01:53:54.080 And you might think, well, that's kind of crazy. Like how does that happen? And it's really bizarre.

01:53:58.100 You can have a patient who had a melanoma that showed up in one part of their body,

01:54:01.980 and then they have metastases on other parts of their skin. M one B, and I always get B and C

01:54:08.780 confused. I think B is the lung. So M one B is to the lung. M one C is to any internal organs. So

01:54:16.880 liver, et cetera. And M one D is to the CNS. And as those numbers increase, as those letters increase,

01:54:22.620 the prognosis gets lower and lower and lower. So one of the first things I always look at when I look

01:54:26.920 at a paper like this is, tell me about the patient population. What was the breakdown of patients?

01:54:32.860 And in table one, so that's again, in clinical papers like this, table one is always, always,

01:54:38.980 always baseline characteristics. Oh, I should mention one other thing, Andrew. This was done

01:54:44.440 as a three to one to one randomization. So again, in the simplest form, the study would have two groups.

01:54:51.700 We're going to just have a treatment group and a placebo group. But in this arm, you had three

01:54:57.340 groups with one of them being the placebo. The placebo got just GP100, which is just a cancer

01:55:04.180 vaccine. By the way, this is a cancer vaccine that never showed any efficacy. So it was a cancer

01:55:10.420 vaccine that had been tested both with interleukin two directly and as an adjuvant for patients who had

01:55:17.340 melanoma resected who were tumor free and then given the vaccine as adjuvants to see, did that have an

01:55:23.980 effect on outcomes? And it didn't. So it's kind of a known placebo. So you had that group, then you had

01:55:30.080 the anti-CTLA-4 group, and then you had anti-CTLA-4 plus GP100. What's the rationale?

01:55:38.020 For the three to one to one, it's basically, it increases statistical power. This total study

01:55:44.300 was a little under 700 people. They put 400 in the anti-CTLA-4 plus GP100 group, and then a little

01:55:52.940 over 130 in each of the other two groups. So you're always going to be able to make these two

01:55:57.900 comparisons. What you can check by doing this is, is there any effect of GP100 in this setting,

01:56:04.980 which had never been done before. So again, GP100 is a known protein expressed by melanoma.

01:56:12.400 And all of these people were haplotyped to make sure that their immune system would recognize it.

01:56:17.840 And the question was, would giving people anti-CTLA-4, i.e. taking the brakes off their

01:56:23.800 immune system with or without GP100 make a difference? Going through this, you can see it

01:56:28.800 sort of skews about 60% to 40% male to female. They talk about something called the ECOG performance

01:56:34.860 status. That refers to how healthy a patient is coming in. So ECOG zero is no limitations whatsoever,

01:56:41.240 which is kind of amazing when you really consider something. I think this speaks to just how

01:56:45.320 devastating this disease is. These are patients who all have like six months to live, a year max.

01:56:51.760 And yet look at this, 58 to 60% of them have no limitation on their quality of life at this very

01:56:58.240 moment. That's going to change dramatically absent a cure here. And then ECOG one has some limitation.

01:57:04.180 And you can see that ECOG one plus ECOG zero is basically 98% of the population. You can see the

01:57:10.680 staging there. So again, very, very few of these patients are the M zero category. M zeros are people

01:57:17.660 who have stage three disease that is so aggressive, it can't be resected. That's about 1%. But the majority

01:57:23.520 of these people are the M1As, M1Bs, M1Cs. So these are people with very aggressive cancers.

01:57:30.560 You can also see that about 10 to 15% of these people also have CNS metastases. Again,

01:57:37.580 the poorest prognosis of the poor. And then you can see about 40% of them have the LDH level above

01:57:45.120 cutoff. All of this is to say, we're talking about a group of patients who have a very high likelihood

01:57:52.900 of not surviving more than a year. It would be very unlikely that many of these patients would

01:57:58.940 survive more than a year. So basically more than 70% of these people have visceral metastases.

01:58:04.180 A third have high LDH and more than 10% have brain mets. They've also all progressed through

01:58:10.200 standard therapy. Radiation chemo.

01:58:12.940 Yeah. And the chemo for melanoma can be a toxic chemo that really just doesn't really do anything.

01:58:19.740 So is it commonplace to use a treatment that failed in clinical trials as a placebo in these

01:58:28.500 sorts of studies? Yeah, it's interesting. You're referring obviously to the GP100. I think the

01:58:33.600 thinking was, okay, it hasn't been effective in other treatments, for example, when combined with

01:58:40.860 IL-2 or as an adjuvant, but never before has it been tried with a checkpoint inhibitor, which is

01:58:48.020 the technical term for this type of drug. I think there was also some belief that it would be easier

01:58:55.000 to enroll patients. I don't think they stated this, but that's often the case. It would be easier to

01:58:59.720 enroll patients if they would know that even in the placebo arm, they're still getting an active agent.

01:59:05.520 Got it. And I suppose there's always the possibility that the combination of the

01:59:09.160 failed drug with a new drug would work. So you're increasing the probability for novel discovery.

01:59:14.980 For sure. And again, if you go back to the randomization of three to one to one, it's really

01:59:21.480 only one fifth or 20% of the participants that would get just the GP100. So in other words,

01:59:28.640 you're basically telling people when they come into this study, there's an 80% chance you're going

01:59:34.220 to get anti-CTLA-4. That's a much better set of odds than your typical study where you're going to be

01:59:40.080 50% likely to get the agent of interest. Right. And people who are literally dying of cancer,

01:59:47.060 they don't want to be in the control group. Right. So the primary outcome for this study

01:59:52.280 actually changed in the study. Now they have to get permission to do that. So the original primary

01:59:59.480 endpoint was the best overall response rate. So I have to explain how response rates are measured.

02:00:05.400 This is a bit complicated. Remember, all of these patients by definition have measurable,

02:00:10.580 visible cancer by visible, either on the surface of their body, but more likely on an MRI or CT scan.

02:00:16.500 So all of these patients had to be scanned head to toe within 12 weeks of enrollment.

02:00:21.020 Again, there's another thing I should point out here, which I know you understand,

02:00:23.560 but it's always worth reminding people when a study like this takes place,

02:00:27.340 it usually takes place over many years. And so it's not the case that all 700 of these patients

02:00:33.120 were enrolled on the same day and finished observing them on the same day. No, no, no. This took place

02:00:37.220 for a very long period of time. This took place across tens of centers. I can't remember if this

02:00:43.020 was just globally or across the world. It might've been across the world. And so every center really

02:00:47.640 needs to adhere to a very strict protocol. And you have a central organization that is running this.

02:00:53.800 So you have a drug company. I think this is Bristol Myers Squibb that makes the drug. They provide

02:00:58.000 the drug. And then you have a CRO, a clinical research organization that is basically managing

02:01:04.100 the trial. And the trial is being done at cancer centers all over the world or all over the country.

02:01:10.660 And enrollment, I think, began in 2008 for this. No, no. I think it completed in 2008. It probably

02:01:16.500 started in about 2004, 2005. And therefore, you had to kind of have real clear protocols around this.

02:01:22.600 So a complete response is the easier of these to understand. A complete response is everything

02:01:28.620 vanishes completely. That's very rare in cancer therapy. So instead, what we look for is a partial

02:01:37.040 response. A partial response, and there are really different ways to define this. There are different

02:01:42.080 criteria, but this is the most common way you define a partial response. A partial response is at

02:01:48.140 least a 50% reduction by diameter. Because remember, in this type of imaging, you're looking

02:01:55.620 at 2D versus 3D. So if you're looking at a lung lesion and it's this big, you know, if it's two

02:02:01.360 centimeters long, it has to go to at least one centimeter in diameter. So it's a 50% reduction at

02:02:06.880 least of every single lesion with no new lesions appearing and no lesions growing. So it's very strict

02:02:14.240 criteria, right? Again, CR means everything vanishes. PR means at least a 50% by diameter,

02:02:22.720 which by the way, is a much bigger reduction in terms of tumor volume when you consider the linear

02:02:27.180 versus the third power relationship of length and volume of every single lesion with nothing new

02:02:33.400 appearing regardless of how small and no lesion growing. So that's a PR. So you basically have

02:02:39.780 no response, progression, we talk about those together, and then partial response and complete

02:02:45.420 response. So initially, the authors of this study, the primary endpoint of this was going to be the

02:02:51.160 best overall response rate. So what was the proportion of patients that hit PR? What was the

02:02:56.460 proportion that hit a CR? That's very common in this type of paper where the outcomes are typically

02:03:02.900 so dire. I don't remember when the study ended, but the amendment was made to change the primary

02:03:09.700 endpoint to overall survival at some point during the study. And by the way, that tends to be the

02:03:15.880 metric everybody cares most about. So the overall survival for metastatic melanoma is zero, with the

02:03:23.560 exception of people who respond to interleukin-2, high-dose interleukin-2, and that will boost the overall

02:03:30.500 survival rate to somewhere between 8% and 10%. Very, very low. These patients, many of whom had

02:03:39.100 already taken and progressed through interleukin-2. Let me refresh my memory on what percentage of those

02:03:48.460 patients. About a quarter of these patients had already taken high-dose interleukin-2, and by

02:03:55.720 definition, the fact that they're in this study means they had already progressed through that. That

02:03:59.880 treatment had failed. Just reiterate the state these patients are in. So now let's look at figure

02:04:07.160 one. So again, I'll describe it because I realize many people are just listening to us. All of this

02:04:12.260 will be available both in the video, and then we'll link to the paper. So figure one is a figure that

02:04:18.480 probably looks really familiar to people who look at any data that deal with survival. It's called a

02:04:23.480 Kaplan-Meier survival curve. So on the x-axis for this curve is time, and time here is shown in

02:04:30.760 months, and on the y-axis is the overall survival. The very top, 100% at the bottom, 0%. And it has

02:04:39.540 three graphs or three curves that are superimposed on one another for each of the three groups. Again, the

02:04:46.480 control group, which is the GP100, the anti-CTLA-4 group by itself, and the anti-CTLA-4 plus GP100.

02:04:55.140 And one of the characteristics of a Kaplan-Meier curve is by definition, they have to be decreasing

02:05:00.520 in a monotonic fashion because it's cumulative overall survival. That just means it can't come

02:05:06.220 down and go back up. Nobody comes back to life. So once a person dies, they are censored from the study,

02:05:12.600 and the curve drops and drops and drops. And you can see that they kind of highlight,

02:05:17.260 and I actually think it makes the graph a little harder to read when they put some of those marks

02:05:21.780 on there. But what really becomes clear when you look at this is that there's a clear distinction

02:05:28.100 between the curve for the placebo group, the GP100 group, and the other two, the two treatment groups.

02:05:37.140 Now, you'll note at the very end that the two treatment groups appear to separate a little bit.

02:05:42.220 I'll talk about that in a second. So when I look at these, Andrew, the first thing I always turn

02:05:47.340 my attention to, I can't resist. I have to look at the right-hand side of the graph because what is

02:05:52.080 that really telling me? The tale of this is showing me the true overall survival. And I want to sort of

02:05:58.560 figure out what is going on. So in the GP100 group, which is the placebo group, it is kind of amazing to

02:06:05.680 think that there is still one person who is alive at 44 months. It's amazing. I mean, it's both sobering

02:06:12.780 and amazing that like one person made it to 44 months. The next thing I ask myself is, well, how long

02:06:18.880 did half of the people make it? That's called median survival. And to do that, you go up to the

02:06:24.760 y-axis and you draw a little line from the 50 over and then you bring that down. That's awfully low.

02:06:32.680 Yeah. In fact, the table will tell us exactly what that is because I think it's really hard to eyeball

02:06:37.580 that stuff. So let's go to, so there's always a table that will accompany these things. And let's

02:06:44.080 pull up that table. I've got this paper spread out over so many things. That's adverse events.

02:06:49.920 Where's our survival table here? Two subgroup analysis of overall survival.

02:06:56.360 It would probably be helpful if I stapled these things together because it would be easier.

02:06:59.800 Well, this is always a trade-off since this is a journal club episode. I will say that stapling

02:07:04.060 helps, but it also prevents one from separating things out, writing in the margins. I like these

02:07:09.020 little mini clips, no financial relationship to the mini clips either. Just have to state that because

02:07:15.720 if you don't say that, people go, oh, you must have a stake in these mini clips.

02:07:19.120 I like these little mini clips. In fact, I'm such a nerd. I always have one of these

02:07:23.380 Pilot V5 V7s on my pocket or my hip. And then my pockets are always filled with these little mini

02:07:29.700 clips. But then again, I have a friend who's a musician. He's always raining guitar picks.

02:07:33.720 As far as occupational hazards go of being a nerd, these mini clips.

02:07:37.520 I'm a big fan of the mini clip as well. Yeah.

02:07:39.820 I went without it today. All right. So thank you. Yes. Table two. Let's look at table two while

02:07:44.840 looking at the Kaplan-Meier curve because now this allows us to see a couple of things. By the way,

02:07:49.120 remember how I said there's like that one person who is still alive in the treatment group? Well,

02:07:54.840 you can tell that he's not a complete responder. He or she is not a complete responder because

02:08:00.980 under evaluation of therapy in table two, it says best overall response and it says complete

02:08:07.440 responders zero. So there was zero complete responders in the placebo. There were two partial

02:08:13.540 responders. Again, a partial responder is some lesions got smaller, some got bigger. Stable

02:08:20.480 disease. It didn't really change that much. And progressive disease is obviously it went beyond.

02:08:27.980 And when you say partial response, like lesions got smaller, are they literally just

02:08:31.380 tracing the circumference of one of these skin lesions and saying, okay, it got bigger, smaller,

02:08:37.380 like just morphology. Yep. Yep. Gosh, this feels so crude in terms of like, I mean, it makes total

02:08:43.000 sense, but like in terms of modern medicine, oh, like your lesion grew from like three millimeters to

02:08:47.440 six millimeters. And then you're literally like drawing little boundaries around little blotches

02:08:52.100 on the skin. Yeah. You're putting a little measuring tape on them. Now, again, most of these are happening

02:08:56.120 in the radiology suite because most of the disease for these patients is inside the body. Remember more than

02:09:01.740 70% of these patients had visceral metastases. So liver, soft tissue, lung, brain. In fact,

02:09:10.140 if you include lung, liver, brain, and viscera, it's pretty much all the patients. So most of this

02:09:15.640 is looking at a CT scan or an MRI for the brain. Okay. So that's kind of the first thing that comes up.

02:09:20.480 The median response rate should be shown pretty prominently here. So I'm looking through this

02:09:26.880 and where is median response? Maybe it's shown in a different table, not disease control rate time

02:09:37.660 to progression. I remember it's about 10 months, but maybe that's just in the text. Yeah, here it is.

02:09:45.120 I thought this would be in a table, but it's on page 715 of the paper. It just reports it.

02:09:50.620 So, I'm sorry, I misspoke. The 10 months was for the anti-CTLA-4 plus GP100 and 6.4 months

02:10:01.220 for the GP100 alone. That's the control. And then 10.1 for the anti-CTLA-4 alone. And again,

02:10:09.360 I'm just always doing this. I'm kind of going back to the paper to be like, does that make sense?

02:10:12.980 And yeah, you called it, right? You said median survival was about eight. Well, it turns out it's

02:10:17.520 actually like six and change because it has that little ding in it and it's out to a little past

02:10:22.640 10 on the two others. So the net takeaway here is, again, just to put that in English because it's so

02:10:28.220 profound, 50% of the patients in the control group were dead in six months. 50% of the patients in the

02:10:37.820 treatment group, both treatment groups were dead in 10 months. So what that means in cancer speak

02:10:44.160 is these drugs extended median survival by four months. Now, that's an important concept.

02:10:52.640 When we think about how has cancer therapy changed over the past 50 years, median survival for

02:10:59.840 metastatic cancer has increased across the board. So a person today with metastatic colorectal cancer

02:11:06.360 or a woman today with metastatic breast cancer or a person with metastatic lung cancer,

02:11:11.880 these people will live longer with those diseases today. Thanks mostly to treatments. This is not

02:11:19.860 an early detection lead time bias issue. This is treatments are allowing people to live longer.

02:11:26.740 And that's an important part of the story, but it's only half of the story, yet it often gets touted

02:11:32.400 as the story. The other half of the story, and frankly, the story that I think is more important is

02:11:37.280 what is overall survival doing? And if you go back to those cancers, the answer is zero.

02:11:44.700 Overall survival hasn't changed for solid epithelial tumors. It was 0% in 1970, and it's 0% today.

02:11:53.700 Everyone dies. Everyone dies from metastatic solid organ tumors. There's those niche examples I gave you.

02:12:00.340 Testicular cancer is now an exception. GI stromal tumors would be an exception. And I'm not including

02:12:06.140 leukemias and lymphomas where now there are exceptions. Not to try and be overly optimistic,

02:12:12.380 but if I look at the graph in figure one, and I look out at the tail of the graph.

02:12:18.680 That's right.

02:12:19.480 And for those that are just listening, what I see, and I'm far less, far less familiar with this type of

02:12:26.400 work and this analyzing these type of data. But what I see is that people in the placebo group,

02:12:30.680 they're all dead, except that one. They're basically all dead at 44 months.

02:12:35.520 Yep.

02:12:35.880 But when I look at how long it takes for everyone to be dead in the true treatment groups,

02:12:41.500 looks like 53, 54 months or so.

02:12:45.200 And they're not dead. That's the point.

02:12:46.940 They're hanging in there, right? So, because, you know, an extra, somebody who lost both of my

02:12:52.840 scientific advisors, two of the three, the other one to suicide. We've talked about this before,

02:12:58.120 but the other two to different cancers, both had the BRCA2 mutation, by the way,

02:13:03.040 you know, an extra eight to 10 months with your kids or with your spouse, or to quote unquote,

02:13:09.900 get your affairs in order is a big deal. I mean, it's still depressing in the sense that nobody survives

02:13:14.860 long-term, but, you know, an extra 10 months, as long as one is not miserable in that time,

02:13:20.680 completely miserable. I mean, that's an extra 10 months of living.

02:13:24.680 And what's interesting here is the observation period stops and some of these patients are still

02:13:29.240 going. What you're highlighting is kind of the point I want to make, which is overall survival

02:13:34.840 is the most important metric and it's the highest bar, make no mistake about it. And it's certainly

02:13:41.280 not the bar any drug company is ever going to want to talk about for a cancer drug.

02:13:46.960 But why not?

02:13:47.800 None of them work.

02:13:48.440 They only want to talk about cures. They don't want to talk about median survival.

02:13:52.320 They only want to talk about extending median survival. And there are lots of people out

02:13:57.080 there that are on this platform. I don't need to get onto it, but we'll say, look, it's a real

02:14:01.640 racket in oncology today where drugs that are extending median survival by four weeks are being

02:14:08.540 put on the market at a tune of 50 to a hundred thousand dollars per treatment. That's not uncommon

02:14:13.880 in oncology. There was one drug that was approved for pancreatic cancer. I believe it extended

02:14:18.400 median survival by nine days and it costs $40,000.

02:14:22.280 And it's being advertised as significant.

02:14:24.460 That was a statistical significant improvement in median survival. It's really understandable why

02:14:28.700 people are very skeptical of the pharma industry. And I think a much more nuanced view is necessary.

02:14:34.120 Clearly, I don't think pharma is all bad, but I really understand why people lose faith in pharma

02:14:39.000 when these types of products somehow make regulatory approval.

02:14:43.440 Does insurance cover these kinds of drugs?

02:14:46.120 It can. In fact, it often does. It depends on the FDA approval, of course, and the indication,

02:14:50.740 but a lot of times they do. Yeah, there's a societal cost to these things,

02:14:54.880 but there's also a patient cost. So a lot of times insurance doesn't fully cover it and a patient has to

02:15:00.240 bear the cost difference. And on top of that, you alluded to this a second ago, which is what if your

02:15:04.700 quality of life is dramatically compromised as a result of this treatment? And yes, statistically,

02:15:09.760 you're going to live nine days longer or three weeks longer, but at what cost to your health

02:15:15.520 in those final remaining days? And by the way, you're potentially straddling your loved ones with

02:15:21.460 enormous debt in your absence. So it's a super complicated topic.

02:15:25.580 There's a dignity component too. I mean, I've seen this in people dying. At some point,

02:15:30.420 they become such a diminished version of their former selves that they don't want to be seen

02:15:36.400 by people that way. Yep. So what is exciting about this drug, this paper is not the one that

02:15:45.540 shows it. The reason I chose this paper, Andrew, is because it was the first approval. A second drug

02:15:52.380 came along that is an anti-PD-1 drug. That drug is called Keytruda. That drug turned out to be even

02:15:59.040 better, has even a greater response rate, both in terms of median survival and overall survival.

02:16:05.600 But this was the landmark paper. I also have a slight bias here, and I'll disclose in a moment why,

02:16:12.760 but I think it just talks about very interesting biology. Let's talk about a couple of things that

02:16:16.900 stuck out to me in this paper. The first thing that stuck out to me, and the authors didn't comment

02:16:21.360 on it, unless they did and I missed it, is look at figure two. So figure two is the subgroup analyses

02:16:31.700 where you're sort of showing a similar graph to the one you showed earlier. You show the response rate

02:16:38.900 or the change in response between the groups, and then you put the error bars on it. And this is where

02:16:44.940 we talk about how, well, it's a 95% confidence interval, so does it touch the unity line? So

02:16:50.180 these are called tornado plots typically. What you'll notice is that in the top, it's comparing

02:16:56.680 the anti-CTLA-4 with GP-100 versus the GP-100. And in the bottom, you're looking at the anti-CTLA-4

02:17:03.060 versus the GP-100. So at a glance, you can see GP-100 is not doing anything. That's the first takeaway

02:17:09.300 of comparing A to B. What I find most interesting is look at the subgroup analysis of females.

02:17:16.160 Notice that in females, while there's a trend towards risk reduction, and this is risk reduction

02:17:22.380 for overall mortality. So again, I just want to restate that. The primary outcome of this trial

02:17:27.320 was changed to overall survival, which I think is the better outcome, by the way. And overall,

02:17:32.560 for all patients, when you compare anti-CTLA-4 plus placebo versus placebo, there was a 31% risk

02:17:43.320 reduction in overall mortality. That's the mathematical interpretation of what you're

02:17:49.060 seeing at the tail end of that Kaplan-Meier curve.

02:17:51.780 Living longer.

02:17:52.900 Living longer.

02:17:53.600 And it sounds like a big difference.

02:17:55.180 Sounds like a big difference.

02:17:56.060 Sometimes it is a big difference.

02:17:57.360 It is for those people because you're really looking at basically 0% surviving in the placebo

02:18:04.440 group versus 20% of people are still alive at 56 months in the treatment group. But look,

02:18:13.000 that means 80% have died. But notice that, and sorry, when you just look at the anti-CTLA-4 plus

02:18:19.940 GP-100 in the subgroup B, that hazard ratio is even showing more compression. It's a 36%

02:18:27.120 reduction in risk of death. But notice that the females did not reach significance. So in the

02:18:34.900 first group, they barely do. And you can see that because the confidence interval runs from 0.55 to 0.92.

02:18:42.180 And notice the error bar almost touches the line. And in the second one, it does not reach

02:18:47.220 significance at all. So I actually went and kind of did a little reading on this after. And I said,

02:18:52.600 hey, was this an outlier study? And it turned out it wasn't. And that about half the studies

02:18:58.600 of anti-CTLA-4 did indeed find that the drug was less effective in women than men,

02:19:05.500 which I found interesting. Now, I couldn't find any great explanation for it, but the most

02:19:10.780 plausible explanations fit into two categories. The first are maybe there are differences in the

02:19:16.840 immune response to the drug if you're a man or a woman. The second comes down to dosing. I should

02:19:23.040 have said this at the outset, but of course, these drugs are not like a pill where it's like everybody

02:19:26.820 gets 50 milligrams of this. They're all dosed based on weight. So this study is dosed, I believe,

02:19:33.060 at three milligrams per kilogram. And because most men are heavier than women, men are getting a higher

02:19:39.980 dose than women. And weight and body surface area and immune system, like these things are not all

02:19:47.080 perfectly linear. I kind of wonder if this difference is simply explained by men on average

02:19:54.520 getting a higher dose than women. Last thing I want to talk about here is in table three. So table three,

02:20:04.760 always an important table to look at in any paper, is what are the adverse outcomes? What are the

02:20:09.640 adverse effects of the drug? I've spent a little bit of time with this and I confess, I definitely

02:20:14.120 don't want cancer to the extent that I can avoid it. But this table made me wonder whether or not I

02:20:19.420 would also want to just avoid cancer treatment given the life extension provided. I mean, these adverse

02:20:25.520 events are pretty uncomfortable. Just to put in perspective, and you always have to kind of be

02:20:31.820 mindful of how many of these adverse events are occurring in people just because their disease is

02:20:36.940 progressing. So the first thing I always want to look at is total adverse events in all three groups.

02:20:43.480 So grade three and grade four are real toxicities. Grade four toxicity is life-threatening toxicity,

02:20:48.900 by the way. Grade three is pretty significant toxicity. Grade one and two, that's not that severe.

02:20:54.020 A little rash, put some corticosteroids on it, it went away kind of thing. Okay, so in the treatment plus

02:20:59.800 GP100 group, 98.4% of people reported some event. So all but 1.6%. In the NTC-TLA-4 group alone,

02:21:10.100 it was 96.7%. So only 3.1% did not. But in the placebo group, it's 97%. So it's important to keep

02:21:17.560 in mind, everybody's having some adverse effect. Okay, well, what if you say, well, let's just limit

02:21:22.040 it to the most severe events? Well, let's just talk about grade four toxicities. There were 6.1%

02:21:29.540 of those in the placebo group, 8.4% in the NTC-TLA-4 group, and 6.8% in the combined group. So not a

02:21:38.140 huge difference in grade four toxicity. Meaning that whatever adverse events are occurring may

02:21:44.800 not be related to the treatment. It may not be related to the treatment. If you think about it,

02:21:49.200 and it's a very awful, sad, morbid thought to imagine, you're looking at the adverse responses

02:21:54.180 of people, more than 80% of whom died during the course of a very, very short study. And so it's

02:22:01.220 very difficult to disentangle what effects or what side effects a person is having just from that

02:22:06.420 process as they are from the actual treatment. But if there is an area where there's a really clear

02:22:12.600 difference, it's down in the autoimmune category. So if you look at any immune-related events,

02:22:19.820 you can see that in the NTC-TLA-4 plus GP100 group, it's about 60% in both of those treatment groups

02:22:28.400 versus 30%. And if you look at the grade three and four toxicities, it's 10% in the NTC-TLA-4,

02:22:37.660 15% in the NTC-TLA-4 alone group, and only 3% in the treatment. So that's a real difference.

02:22:46.860 But it makes sense that people getting this drug plus placebo or just the drug would have autoimmune

02:22:53.060 issues because this is an immunotherapy. It's an immunomodulator. In fact, what is it doing?

02:22:57.920 It is taking the brakes off the immune system. But then again, the things that they list out,

02:23:03.100 pruritus, is that an irritation of the skin? Yeah. Irritation of the skin. I'm not a physician,

02:23:08.320 but I know that any itis is going to be like an inflammation. And OMA, unfortunately, likely a

02:23:13.440 cancer or cell replication. Look at the gastrointestinal differences. And the vitiligo,

02:23:18.540 so 3.7%, 2.3%, 0.8%. The GI stuff is the most common stuff you're going to see there. Those are the

02:23:26.100 really big ones. And of course, there's diarrhea and there's diarrhea. Oh, yeah.

02:23:29.500 Like there's traveler's diarrhea. There's ate an overly spicy large meal the night before diarrhea.

02:23:34.920 And then there's like, can't really do anything besides make trips back and forth to the bathroom

02:23:39.700 diarrhea. Put it this way. There's colitis here is diarrhea so significant these patients require

02:23:44.080 IV fluids. Now, what you don't see here is how many of these patients actually required

02:23:48.200 corticosteroids to reverse the autoimmunity. So a lot of times what will happen here in these studies

02:23:53.080 or with these drugs is the autoimmunity becomes so significant that you have to stop the drug and give

02:23:59.020 corticosteroids. Do the exact opposite. You now have to shut the immune system down. So you just took

02:24:03.960 the brakes off it with the drug and now you need to shut it down with corticosteroids. When I was in

02:24:09.380 my fellowship, I wrote a paper about autoimmunity correlating with response rate in anti-CTLA-4

02:24:18.840 early on. This was during the phase 2 work. So the NCI was a very early adopter of participating in

02:24:26.600 these trials. And it was observed that, or at least hypothesized, this is what the paper basically

02:24:32.500 wrote about, which was, is there any correlation between autoimmunity and response? And it turned

02:24:38.740 out the answer was yes. There was a very strong correlation. So there was no difference in auto

02:24:44.480 immunity between the doses. So the paper we wrote was two dosing schedules. So it was basically the

02:24:51.400 full dose, the three milligrams per kilogram versus a low dose, one milligram per kilogram. This is a

02:24:55.460 phase two trial. Those are your two arms. There turned out to be no difference in autoimmunity between

02:25:00.320 them, but there was a big difference between the response rate that tied to autoimmunity. In other

02:25:07.840 words, autoimmunity predicted response. Now, I think over time, these investigators, the doctors who

02:25:16.140 administer these treatments are getting better and better at catching these things earlier because

02:25:19.860 these autoimmune conditions can actually be devastating. So on a very personal note, when

02:25:25.360 Keytruda came out, I want to say it was around 2000, 2013, 2014, thereabouts. Again, it was for

02:25:34.400 treatment of metastatic melanoma. I want to come back and explain why melanoma gets all of the attention

02:25:39.400 in immunotherapy conditions. I'll state that. But anyway, a friend of mine got pancreatic cancer

02:25:45.280 and he got the bad type of pancreatic cancer. So this is adenocarcinoma of the pancreas. This is a

02:25:52.300 non-survivable type of cancer. Furthermore, his was unresectable. Can you explain what that is for

02:25:57.180 people? About 20% of people who have pancreatic cancer technically have it in a way where you could

02:26:04.240 still take out the head of the pancreas. Right. The Whipple procedure. The Whipple procedure.

02:26:08.360 Now, tragically, most of those patients will still recur. My understanding is that pancreatic cancer

02:26:13.920 progresses from anterior to posterior in the pancreas and that the Whipple is a removal of

02:26:19.500 the front end, the anterior. That's the Whipple procedure. So if the cancer has progressed far

02:26:24.960 enough caudal into the posterior pancreas, then there's nothing left to cut out basically.

02:26:30.960 Can we survive without a pancreas for any amount of time? Oh yeah, absolutely.

02:26:34.200 So why don't they just remove the whole pancreas? Oh, that's my point. It's already

02:26:37.500 micrometastasized. The surgical procedure is not the challenge anymore. It used to be.

02:26:42.320 So at Johns Hopkins, which is one of the hospitals where this was pioneered,

02:26:45.820 the 30-day mortality for a Whipple procedure was, I don't know, 80%. Whoa. And the reason was

02:26:53.280 to figure out how to suture a pancreas to the bowel. So the pancreas is such an awful organ to operate on

02:27:01.020 because its enzymes are designed to digest anything and everything. So imagine now you have to cut the

02:27:07.880 pancreas in half, take out the head of the pancreas with the duodenum, and then somehow

02:27:13.180 sew that open half of a raw pancreas to the end of the jejunum and not let it digest itself.

02:27:20.640 Someone at Hopkins figured this out.

02:27:22.640 No, the first one was actually done by A.O. Whipple. But yes, at Hopkins is where they figured

02:27:29.320 out the way to put drains in, the surgical technique, how to do it in two layers, what type of stitches to

02:27:36.000 use. All of the nuances of this were worked out in a few places, but I would say Hopkins more than

02:27:41.720 any place else. And are there physicians who try this on non-human primates or something,

02:27:47.260 or is this always just done on patients?

02:27:48.980 Well, nowadays, I mean, put it this way, even 25 years ago at a major center like Hopkins,

02:27:55.340 the mortality of that procedure was less than 1%.

02:27:57.940 So there have been some victories.

02:27:59.760 Well, yes, but here's my point. That's no longer the bottleneck.

02:28:02.840 Taking out the pancreas safely, as complicated and challenging as that is, and if you need a

02:28:07.300 Whipple procedure, you only want to have it done by someone who just does that night and day.

02:28:11.580 You don't want weekend warriors doing it. That's not why people are living or dying.

02:28:16.020 They're dying because the cancer just comes back. It was already spread to the liver by the time you

02:28:21.120 did it. You just didn't realize it yet. So whether you took out the whole pancreas or the head of the

02:28:25.680 pancreas or the tail of the pancreas, the location of the tumor is predictive of survival only in the

02:28:32.980 extent that it basically is a window into how soon did symptoms occur. So pancreatic cancers in the tail

02:28:40.600 tend to be more fatal, even though they're way easier surgically to take out because by the time you

02:28:48.280 develop symptoms of a tail pancreas cancer, it's a big cancer.

02:28:53.820 I was going to ask this question later, but I'll just ask it now. Given the link between the immune

02:28:57.580 system and these cancers, is there an idea in mind that people who are, let's say, 40 and older or 50 and

02:29:06.440 older who don't yet, they're not diagnosed with any cancer, would periodically just stimulate their immune

02:29:13.120 system to wipe out whatever early cancers might be cropping up. You know, just take a drug to just

02:29:19.780 ramp up the immune system, even to the point where you start having a little diarrhea, maybe a few skin

02:29:23.880 rashes, and then come off the drug. Just basically to fight back whatever little cell growths are

02:29:30.240 starting to take place in skin or liver for three weeks out of each year. I mean, why not?

02:29:36.540 Yeah, it's an interesting question. I've never thought of it through that lens. I suppose the question is,

02:29:41.060 what can we do to keep our immune systems as healthy as possible as we age?

02:29:45.880 Stay on a normal circadian schedule. There's evidence for that.

02:29:49.220 Sure. No, there's evidence that certainly if it promotes sleep, anything that promotes better

02:29:53.540 rest is going to promote immune health. Because if you ask the macro question, which is like,

02:29:58.840 why does the prevalence of cancer increase so dramatically with age? There are certain diseases

02:30:04.660 where it's really obvious why the prevalence of the disease increases with age.

02:30:08.760 Yeah, like age-related macular degeneration.

02:30:11.220 Sure. Or cardiovascular disease is by far the most obvious because it's an area under the curve

02:30:16.120 exposure problem. The more exposure to lipoproteins and the more the endothelium gets damaged,

02:30:20.900 the more likely you are to accumulate plaque. And again, it totally makes sense why 10-year-olds

02:30:25.660 don't have heart attacks and 80-year-olds do. But when you sort of acknowledge that, well, hey,

02:30:32.180 anybody's accumulating genetic mutations. We're always surrounded and being bombarded by things

02:30:36.900 that are altering the genome of ourselves. Is it simply a stochastic process where the longer you

02:30:42.880 live, the more of these mutations you're going to occur until at some point one of them just

02:30:46.680 wins? I think that's got to be a big part of it. But I think another part of it, clearly I'm not

02:30:51.960 alone in thinking this, is that our immune system is getting weaker and weaker as we age. People become

02:30:56.760 more susceptible to infections as they get older. And I think that that's equally playing a role in

02:31:02.780 our susceptibility to cancer. So yeah, I think the question is how do you modulate immunity as you

02:31:08.460 age? And to me, that's one of the most interesting things about rapamycin potentially is that when

02:31:13.720 taken the right way, it seems to enhance cellular immunity, which again, that's potentially a really

02:31:19.480 big deal. Again, at least in short-term human experiments in response to vaccination, it's

02:31:24.320 enhancing vaccine response. So the question is, would that translate into cancer? Nobody knows.

02:31:29.580 Could that be one of the reasons why animals treated with rapamycin live longer and get less

02:31:35.260 cancer? Don't know. It could also be that it's at a fundamental level that's targeting nutrient

02:31:39.700 sensing. Where I was going with that story was that, and maybe I'll back up for a moment.

02:31:45.420 Why melanoma? So we didn't really know this 30, 40 years ago in the early days of immunotherapy.

02:31:53.400 But what we know now is that most cancers probably have about 40 mutations in them.

02:31:59.580 That's like ballpark. 40, 50 mutations is standard fare for a cancer. But melanoma happens to be one

02:32:06.840 of the cancers that has many, many more mutations. And the more mutations a cancer has, the more

02:32:13.260 likelihood that it will produce an antigen that's recognized as non-self. And that's why in the early

02:32:19.920 days of immunotherapy, the only things that worked were IL-2 against metastatic melanoma and kidney

02:32:26.020 cancer, because kidney cancer turned out to also be one of those cancers that, for reasons that are

02:32:30.660 not clear, produced hundreds of mutations. And so it's no surprise that the early studies of

02:32:37.660 checkpoint inhibitors were also done in metastatic melanoma, where you basically have more shots on

02:32:43.380 goal. Again, if I'm going to take the brakes off my immune system, I might as well do it in an

02:32:48.240 environment where there are more chances for my T cells to find something to go nuts against.

02:32:54.780 It's 2013, 2014. And this friend of mine who has something called Lynch syndrome, which is a,

02:33:01.920 one of those few hereditary or germline mutations that results in a huge increase in the risk of

02:33:07.980 cancer. He had already had colon cancer at about the age of 40 and had survived that. It was a stage

02:33:13.620 3 cancer, but he had survived it. Well, now five years later had developed pancreatic cancer. And

02:33:20.520 when he went to see the surgeon, they said, there's nothing we can do. It's too advanced. To put that

02:33:26.300 in perspective, that is a death sentence. That's a six month survival. And at around that time, there

02:33:33.060 was a study that had come out in the New England Journal of Medicine that had talked about how patients

02:33:37.300 with Lynch syndrome had lots of mutations. So we talked with his doctors about the possibility of

02:33:44.780 enrolling him in one of the Keytruda trials. There was one going on, I think at Stanford.

02:33:49.860 The thinking being, well, you would want to target a checkpoint inhibitor against somebody who has a

02:33:55.160 lot of mutations. And even though typically we don't see that in pancreatic cancer, his is a unique

02:34:00.020 variant of it because it's based on this. And so sure enough, he was tested for these mismatch repair

02:34:06.040 genes. He had them enrolled in the trial and amazingly had not only a complete regression of

02:34:12.560 his cancer, and he's still alive and cancer-free today, 10 years later, but the treatment worked

02:34:18.660 so well at activating his immune system that his immune system completely destroyed his pancreas.

02:34:24.580 So now he is effectively had a pancreatectomy based on his immune system. So now he actually has

02:34:30.500 type 1 diabetes. He has no pancreas.

02:34:33.720 Jack's insulin to deal with that or implant.

02:34:36.200 No, no. He has to use insulin just like someone with type 1 diabetes.

02:34:38.800 Had to pick being alive with type 1 diabetes.

02:34:41.480 Yeah, of course. No comparison. But it's just an interesting example of how remarkable this

02:34:46.960 treatment was able to work. You could completely unleash the immune system of a person and you eradicate

02:34:54.860 the cancer and the rest of the cells around it. There are many organs we could live without.

02:35:00.520 There are certain organs you can't live without. You can't live without your heart, lungs, liver,

02:35:04.260 kidneys. But many things that kill people arise from organs. The breast, you could live without

02:35:09.900 all breast tissue. Prostate, you can live without all prostate tissue.

02:35:12.560 I mean, no one would choose to live without these.

02:35:14.560 Right. But I'm saying if you had metastatic cancer and you had a bullet that could selectively

02:35:20.340 target a tissue, you would take it. And right now, the only tissue we can do that against

02:35:25.500 is a CD19 B cell. And that's what those CAR T cells are. So right now, these are not tissue-specific

02:35:31.180 treatments, but they're mutation-specific. The last thing I'll say about this paper that I found

02:35:35.740 interesting, I was looking for it and I was surprised they didn't at all comment on if there

02:35:40.780 was any correlation between autoimmunity and response. So they obviously acknowledge the

02:35:45.320 autoimmunity in table three, but I would have loved to have seen a statistical analysis that

02:35:50.880 said, hey, is there any correlation between response rate and autoimmunity? But they didn't

02:35:56.280 comment to that effect. So we're left wondering what the current state of that is. And I guess

02:36:02.460 in summary, I'll say that the reason I thought this was an interesting paper to present is that

02:36:08.600 I still believe that immunotherapy is probably the most important hope we have for treating cancer.

02:36:17.800 And while I think we're still only scratching the surface of it, collectively, the overall survival

02:36:23.620 increase for patients with metastatic solid organ tumors is about 8% better than it was 50 years ago.

02:36:31.040 And virtually all of that has come from some form of immunotherapy, I think is promising. And I think the

02:36:38.020 holy grail, meaning the next step, if you go back to where we started the discussion,

02:36:43.400 is coming up with ways to engineer T cells to be even better recognizers of antigens.

02:36:53.140 And there's many ways to do that. One is to directly engineer them. Another is to find T cells that have

02:36:59.240 already migrated into tumors. Those are called tumor infiltrating lymphocytes or TIL. And expanding those

02:37:06.360 and engineering them to be better and younger. Is it possible to engineer our own T cells to be more

02:37:12.980 pH variant tolerant? Meaning since this cloaking of the local area by changing the pH, could we pull

02:37:22.120 some T cells? I'm always thinking about the inoculation stuff, like pull some T cells as part

02:37:27.260 of our standard exam when we're 30, you know, and grow some up in an environment that the pH is slightly

02:37:33.240 more acidic than normal, and then reintroduce them to the body. I mean, after all, they are T cells.

02:37:40.080 In other words, give them a little opportunity to evolve the conditions they can thrive in,

02:37:46.440 or even just keep them in the freezer in case we need them.

02:37:49.700 Yes. So the interesting thing is, I don't know that if you just got them to be comfortable in a

02:37:54.420 lower pH, it would be sufficient. Because there are still so many other things that the cancer is

02:38:01.940 doing as far as using other secreting factors. It seems that by far the most potent thing comes down

02:38:09.240 to expanding the number of T cells that recognize the antigen and making sure that you can get that

02:38:17.740 number big enough without aging them too much. So in some senses, it has become a longevity problem

02:38:23.720 with T cells. The way to think about it is, you want an army of soldiers who are wise enough to

02:38:31.600 recognize the bad guys, which comes with age, but young enough to go and kill. And right now,

02:38:38.900 both extremes seem to be unhelpful. When you go and find tumor infiltrating lymphocytes in a tumor,

02:38:45.280 they're very wise. They've demonstrated that they can do everything. They can outmaneuver the cancer,

02:38:50.280 but they're too old to do anything about it. And when you take them out to try to expand them by

02:38:55.160 three logs, which is typically what you need to do, expand them by a thousand fold,

02:38:59.400 they can't do anything. And what about avoiding melanoma altogether? I mean,

02:39:02.780 obviously avoiding sunburn. Somehow I got couched as anti-sunscreen, and that is absolutely not true.

02:39:09.020 I said some sunscreens contain things that are endocrine disruptors, and we're going to do a whole

02:39:15.000 episode on sunscreen. Maybe we could do some journal clubs on them.

02:39:18.040 I'm actually planning something on that as well. I wanted to do a deep dive on this.

02:39:21.600 And some dermatologists reached out, some very skilled dermatologists reached out and said

02:39:26.120 that indeed some sunscreens are downright dangerous, but of course melanoma is super

02:39:30.480 dangerous. No one disputes physical barriers for sunscreen. Everyone agrees that that is unlikely

02:39:35.560 to have endocrine disruption. So physical barriers are undisputed, but aside from limiting

02:39:41.120 sunlight exposure to the skin, what are some other risks for melanoma?

02:39:45.600 I mean, I think that's the biggest one. I do not believe that smoking poses a risk for

02:39:49.480 melanoma, and if it does, it's going to be very small. There are hereditary cases,

02:39:53.880 so one needs to be pretty mindful when taking a family history. And by the way, there are really

02:39:58.360 weird genetic conditions that link melanoma to other cancers, such as pancreatic cancer,

02:40:04.100 by the way. So whenever I'm taking somebody's family history and I hear about somebody that had

02:40:08.740 melanoma and someone that had pancreatic cancer, there's a couple genetic tests we'll look at

02:40:12.740 to see if that's a person that's particularly sensitive from a genetic predisposition.

02:40:17.800 And by the way, I think with melanoma, although it's not completely agreed upon, I think it's

02:40:23.480 less about sun exposure and more about sunburn. And again, I'm sure there's somebody listening to

02:40:28.800 this who will chime in and apply a more nuanced response to that. But there's a fundamental difference

02:40:33.920 between I'm out in the sun, getting sun, making some vitamin D versus I'm getting scorched and

02:40:40.620 undergoing significant UV damage. There might also be something to be said for the time in

02:40:45.780 one's life. And I've certainly seen things that suggest that early repeated sunburns would be

02:40:50.880 more of a risk. So I think that's not a controversial point in the sense that who wants to be sunburned,

02:40:56.900 right? So it's like whatever one needs to do to be sunburned, whether it's being mindful of what the

02:41:01.800 UV index is, wearing the appropriate cover, wearing the appropriate sunscreen. I also find

02:41:07.080 the whole anti-sunscreen establishment to be a little bit odd.

02:41:11.440 Well, the anti-sunscreen establishment is odd. I'm trying to open the door for a nuanced discussion

02:41:16.320 about the fact that some sunscreens really do contain things like oxybenzenes and things that

02:41:21.360 are real endocrine. And you're spraying them on kids.

02:41:23.880 Yeah, yeah, yeah. But when you just look at the straight, you know, the good old-fashioned

02:41:26.600 mineral sunscreen is perfectly safe. Yeah. As far as we know, dare we cross the seed oil debate into

02:41:33.240 this? Some of the folks who are really anti-seed oil also claim that seed oils increase risk for

02:41:37.880 sunscreen. Peter and I are smiling because we have teed up a debate soon with some anti-seed oil and

02:41:46.640 less anti-seed oil experts. So that's forthcoming. That's going to be a fun one. We'll be doing all of

02:41:52.700 that with our shirts on. I really appreciate you walking us through this paper, Peter. I have never

02:41:58.360 looked at a paper on cancer and certainly not one like this. I learned a lot and it's such an

02:42:04.600 interesting field, obviously because of the importance of getting people with cancer to

02:42:09.240 survive longer and lead better lives, but also because of the interaction with the immune system.

02:42:14.920 So we learned some really important immunology too.

02:42:17.300 And this was great. I feel much more confident now in the belief that the exposure to light early

02:42:24.180 and late in the day can actually have benefits. And as I said, I think that there's some causality

02:42:29.440 here and I think it shouldn't be ignored. Well, this was our second journal club. I look forward

02:42:34.440 to our third. Next time you'll go first. We'll just keep alternating. And we've also switched venues,

02:42:39.460 but we both wore the correct shirt. I hope people are learning and not just learning the information,

02:42:46.060 but learning how to parse and think about papers. And I certainly learned from you,

02:42:49.620 Peter. Thank you so much. Yeah. Thanks, Andrew. This is great.

02:42:52.940 Thank you for listening to this week's episode of The Drive. It's extremely important to me to

02:42:57.320 provide all of this content without relying on paid ads. To do this, our work is made entirely

02:43:02.300 possible by our members. And in return, we offer exclusive member only content and benefits above and

02:43:08.940 beyond what is available for free. So if you want to take your knowledge of this space to the next

02:43:13.260 level, it's our goal to ensure members get back much more than the price of the subscription.

02:43:18.440 Premium membership includes several benefits. First, comprehensive podcast show notes that detail

02:43:24.540 every topic, paper, person, and thing that we discuss in each episode. And the word on the street is

02:43:30.380 nobody's show notes rival ours. Second, monthly ask me anything or AMA episodes. These episodes are

02:43:38.260 comprised of detailed responses to subscriber questions typically focused on a single topic

02:43:43.120 and are designed to offer a great deal of clarity and detail on topics of special interest to our

02:43:48.480 members. You'll also get access to the show notes for these episodes, of course. Third, delivery of our

02:43:54.380 premium newsletter, which is put together by our dedicated team of research analysts. This newsletter

02:43:59.760 covers a wide range of topics related to longevity and provides much more detail than our free weekly

02:44:06.000 newsletter. Fourth, access to our private podcast feed that provides you with access to every episode,

02:44:12.880 including AMA's sans the spiel you're listening to now and in your regular podcast feed. Fifth,

02:44:19.800 the qualies, an additional member only podcast we put together that serves as a highlight reel featuring

02:44:26.000 the best excerpts from previous episodes of the drive. This is a great way to catch up on previous episodes

02:44:31.740 without having to go back and listen to each one of them. And finally, other benefits that are added

02:44:36.540 along the way. If you want to learn more and access these member only benefits, you can head over to

02:44:42.080 peteratiamd.com forward slash subscribe. You can also find me on YouTube, Instagram, and Twitter,

02:44:49.040 all with the handle peteratiamd. You can also leave us a review on Apple podcasts or whatever podcast

02:44:55.620 player you use. This podcast is for general informational purposes only and does not

02:45:00.820 constitute the practice of medicine, nursing, or other professional healthcare services, including

02:45:05.200 the giving of medical advice. No doctor patient relationship is formed. The use of this information

02:45:11.300 and the materials linked to this podcast is at the user's own risk. The content on this podcast is not

02:45:17.700 intended to be a substitute for professional medical advice, diagnosis, or treatment. Users should not

02:45:23.260 disregard or delay in obtaining medical advice from any medical condition they have, and they should

02:45:28.000 seek the assistance of their healthcare professionals for any such conditions. Finally, I take all conflicts

02:45:34.340 of interest very seriously. For all of my disclosures and the companies I invest in or advise, please visit

02:45:40.680 peteratiamd.com forward slash about where I keep an up-to-date and active list of all disclosures.

02:45:53.260 for all of my disclosures.

The Peter Attia Drive - January 22, 2024

#286 ‒ Journal club with Andrew Huberman： the impact of light exposure on mental health and an immunotherapy breakthrough for cancer treatment

Episode Stats

Length

Words per Minute

Word Count

Sentence Count

Misogynist Sentences

Hate Speech Sentences

Summary

Transcript