Dave Feldman with new data on LDL and….ALL CAUSE MORTALITY!
Does Cholesterol kill you? Well look and see…!
(full talk from Dave at Keto Salt Lake here: https://www.youtube.com/watch?v=UZv00mMiB9M&feature=youtu.be)
INDEX:
00:00 Mining NHANES Database for LDL and Mortality data
02:30 Stratifying the data – to get the answer
04:10 And the answer? Does higher LDL shorten your life?
06:24 A discussion on “Reverse Causality” – could it confound?
10:38` The confounding of it all – LDL is riddled with correlation to bad things
13:15 What’s really going on with LDL
15:51 More opportunities like this in the future – with more mining!
TRANSCRIPT:
Ivor Cummins 00:00 I’m here at Keto Salt Lake for a conference, which just started today. And I caught up again with Dave Feldman, who just gave a fascinating talk again on LDL but with some very new data. Great to see you again, Dave.
Dave Feldman 00:14 Likewise, Ivor. Thanks for having me.
Ivor 00:16 Not at all. Always a pleasure. Now, you had new data today, because I know that we’ve all wondered about LDL being a problematic compound in the body. And it seems there’s been lots of data over the last five or six years where when LDL is high but other things like HDL are high and triglycerides are low, most of the data suggests there’s no problem with LDL being high. But today was different because you pulled databases of 50,000 people, and you did a bit of analysis, and the results were quite stunning. So maybe if you start going through some of those.
Dave 00:50 Yeah. Well, of course, as you know, this has been a bit of a long history. I’ve not been in the space quite as long as you have but in the time that I have, I’ve been interested in how these markers play out, particularly as you were kind of mentioning outside of LDL, HDL cholesterol and triglycerides. And it’s always been a bit frustrating because as you know, all we pretty much can do is trade studies between each other. If we ourselves are not, you know, researchers with that data on hand, it’s hard for us to look through raw data. But we’re engineers, right? So engineers, we like to get under the hood of things and start taking it apart.
01:28 I was fortunate to have the help of Tommy Wood in being able to get ahold of the NHANES (National Health and Nutrition Examination Survey) data. I can’t say right off hand, but it’s basically the North American… it’s kind of survey data, but also includes a lot of blood work, which was pretty fascinating. And there are different kinds of sets, but the set we got ahold of was basically everything from 1999 to I want to say 2015. So it was really quite a lot of data. I think once we determined everybody who is eligible, they came in at around, I want to say 40,000. And once I had that in hand, I just went crazy. Had a good time, because of course, I wanted to know for myself without anybody curating the data for me and their study, and by using adjustments they wanted to use, what would I see? Right? So naturally, without wasting any time at all, I went, “All right, I want to see LDL, and of course, my favorite market, which you know, is all-cause mortality – live longer or not.
02:30 So immediately I went into LDL and stratified it against all-cause mortality. And guess what? Higher LDL, higher all-cause mortality. This actually makes it fairly. And it was stepwise, it looked really like a strong case for those who believe in the lipid hypothesis that, no in fact, this is a risk factor – high LDL equals high all-cause mortality.
02:55 Well then I went to check to confirm that indeed all of these people in the stratifications were nearly identical to each other in many of the things that also matter for mortality such as age. They’re not. So I then had to go, “All right, where is it that I could find where the ages will have parity to each other?” And I found that that could be in age 50 and up. So, once we had age 50 and up and then we stratified by LDL again, and then on top of that you apply follow up time, because the follow up time was also different between these different stratifications. So, we just changed it to be simply a mortality per year, a follow up. Simply put, if you have an LDL of say zero to 80, what percent of those people in your group are going to die per year?
Ivor 03:49 So standardized by age, you must do because age is a huge risk factor for mortality.
Dave 03:54 It sure is.
Ivor 03:55 So standardizing by age and standardizing by the follow up periods, to get the per year.
Dave 04:00 Correct.
Ivor 04:01 And that’s perfect data now to look at LDL linkage to all-cause mortality.
Dave: 04:08 That’s right.
Ivor: 04:08 What did you see with this treatment?
Dave: 04:10 The graph flipped. All of a sudden we see the mortality per year of follow up is highest at the lowest levels of mortality. And as you can see, as you move upwards in LDL, it actually gets lower, which does complement existing studies that we have seen of those that show higher LDL has lower all-cause mortality. But now I actually got to see it for myself with the data directly, I got to actually be playing with the data, do very minimal adjustments, and all those adjustments are in the slideshow. All I did was I basically removed anybody who didn’t have any of the major lipid numbers, removed anybody that had triglycerides over 400, since we’re using the Friedwald equation.
Ivor: 04:54 You removed people who were treated with lipid-lowering?
Dave 04:57 That’s right, and that was the one of their adjustments is we removed everybody who had statins.
Ivor 05:00 Yeah, which of course you’d have to do, because that’s going to change every natural relationship between biomarkers and death by having a drug. So your data is perfectly clean.
05:10 Now another interesting thing is, NHANES data has been criticized because a lot of it has food questionnaires, food frequency questionnaires. So the Harvard School of Public Health know this, but they still try and make correlations with diet and disease. But that’s different because there you can have confounders. But just to stress for anyone listening, this is LDL values and all cause on cardiovascular mortality. There is no confounding. This is just the reality of the data.
Dave 05:39 More specifically, I didn’t even bother with any of the dietary data yet. I’m not even sure how interested I would be in it. I was interested in going straight to the blood work.
Ivor 05:47 Yeah.
Dave 05:48 What would it matter if you were getting any kind of data from diet if at the end of the day, that’s not even what you were wanting to look at? That was the case for me.
Ivor 05:57 Absolutely. Diet is only for the people trying to make correlations between diet and disease. But that’s exactly not what you’re interested in here. It’s the LDL versus mortality.
Dave 06:07 Right.
Ivor 06:08 So, when you corrected for age, which was crucial and corrected for follow up period, which is crucial also, you got a graded higher LDL, lower mortality all-cause but also lower cardiovascular mortality, I think.
Dave 06:24 Well, here’s what gets funny. I didn’t spend that much time. So as was a big portion of my talk, while I do find the question of cardiovascular disease important, no question, right? It’s not as important as the mother of all of them, which is all-cause mortality. Do you live longer, or do you live shorter by how much LDL you have, right? Because the problem is, if we’re looking only at one marker, as I bring up with my kind of humorous example but it’s the punctuated point, the cyanide diet, right? If I put you on a cyanide diet, guess what? I guarantee I’ll reduce your chance of dying of a heart attack by 100% and I’ll reduce your chance of dying of Alzheimer’s disease by 100%.
07:10 That was much of my talk. The much of my talk was to introduce those things that LDL can also be potentially beneficial toward. And the only way you can really know if you look at the balance at the end of the day. Is if it turns out that people who have low LDL have a higher association with cancer, how can you know if having low LDL is in fact, in some way causal towards cancer? You can’t really easily know that, can’t as easily know that with LDL and heart disease if in fact it turns out that that’s part of the body’s strategy for containment, right?
07:45 So that’s why I want to see, look, looking at all of the numbers together, particularly for mortality, do you just live longer if you have higher LDL? And the NHANES data with very minimal adjustments… let me just add one more thing. This is the beginning of a conversation, not the ended one. So, other people have access to the NHANES data, and if they think they have a better, more accurate way to show of it, I absolutely welcome that. But this was the most minimal way that I could see, that you could adjust it and look cleanly at all-cause mortality. But I’m definitely interested in any other ways to do this. For now though, it looks like, typically after age 50, the one time we could get parity with age, the higher your LDL (generally speaking) the lower your all-cause mortality.
Ivor 08:33 Yup. There you have it. And again, we kind of knew elements of this from older datasets and studies, but the beauty of this is it’s straight from the raw database with minimal manipulation so we can really trust it. Fantastic!
08:48 Now another thing that sometimes said as well, the lower LDL people could have higher mortality, but this reverse causation… in other words, people who are sick and have other problems, their LDL could go down, and then you get an association with low LDL non cause mortality. But to those people, I always sent the BMJ paper from around six, seven years ago, and a guy looked at the Framingham data or team, and they went back 20 years of LDL measurements. And they showed with no question that the low LDL link to mortality was present for the full 20 years prior to the death. So in other words, there was no “following disease the LDL went down” situation. The LDL was low for 20 years of tracking before the outcome, without any exception. So the reverse-causality does not apply specifically for cancer, which was the question – but I would say also for everything else.
Dave 09:43 This is something that I also like the NHANES data for. The age that’s listed in the NHANES data is the time at which you took the test. Then you have follow-up. And follow-up could be 5, 10 years longer, right?
Ivor 09:59 Yeah.
Dave 10:00 You see where I’m going here?
Ivor: 10:01 I see where you’re going. You essentially have a version of the study I just described, because you’re generally seeing the LDL 10 or 15 years before the death.
Dave 10:09 Right.
Ivor 10:09 Perfect.
Dave 10:10 And obviously, if… and to be sure, I haven’t done a deeper analysis on it. I want to, before I answer this fully, but I think I would have spotted certainly a large degree if you will of low LDL combined with a long follow up time. That would be very meaningful. If you have a long follow up time after having a low LDL. That speaks to that data directly.
Ivor 10:38` Yeah. And also, that 20 year study I mentioned, they also saw dramatic consistency within-person for LDL readings when they had multiple readings. So it wasn’t even the case that you got one low reading. They showed consistency and remarkably low variability per person. Their LDL was broadly consistent throughout the long period. So I think this is really tight data. And in a sense, this data overturns all of the theory that higher LDL is worse, even higher LDL over life is worse. And this resonates with all the autopsy data that shows no link between LDL and atherosclerosis level. And links with many other studies which show LDL doesn’t really show up. And yet we still have this pervasive belief that LDL is the “sine qua non” or the key thing in CVD and in disease. So it’s going to take a long time to overturn that, do you think?
Dave 11:37 Well, in… I’ll give a little more of a nuanced answer. I definitely do think that your LDL can be high for a bad reason. And I kind of go into that into the talk and I don’t want to unpack that too much, but I’m just going to say basically, this, that I think it can be a downstream result of something to the degree with which it’s a cause of something that I have a little more of a challenge toward. Because the downstream results of something, as I mentioned in the talk, the two main channels of area are metabolic dysregulation; if you’re metabolically unhealthy and therefore you have more VLDLs, right?
Ivor 12:15 Yeah.
Dave 12:16 And also, if you’re in a state of a challenge such as a disease, especially if it’s chronic disease like chronic inflammation, you also may have more VLDLs because there may be more, for example, cytokines signaling to the liver to generate more of these to help fight it, right?
Ivor 12:31 Yes.
Dave 12:32 But in both of these examples I just mentioned, you see higher remnants, which are those lipoproteins that actually aren’t LDLs, but also have cholesterol and they tend to be bloated with triglycerides. And metabolic dysregulation, you have trouble parking the triglycerides into the tissues, which is another reason why you see high triglycerides. Which is why in both cases, you pretty much can often look to triglycerides, especially low HDL. Either of those two and especially both of them already have in the literature something known as atherogenic dyslipidemia, which is the two of those together coupled finally added to a preponderance of small dense LDL particles.
13:15 If you have the flip side of that, and high LDL, I have yet to see in the data where this is associated with worse outcomes, especially with all-cause mortality. And that is a problem, the very thing you’re talking about, the context definitely matters.
Ivor 13:30 And it sounds like… I mean, I kind of found this for six or seven years now. LDL kind of rideshares on the back of much more important root causes because it tends to track. Like, higher particle count is a classic indication of metabolic syndrome or insulin resistance. So it’s going to track and rideshare. And the LDL itself… I mean, really, Professor Ken Sikaris from, I don’t know, was it Australia or New Zealand?
Dave 13:59 It’s Australia.
Ivor 13:59 Australia. Fantastic lecture recently and he showed me something I’d known from papers five years ago. But when your HbA1c is in the normal range, the LDL is at moderate levels on huge datasets. But when your HDL or your HbA1c goes into prediabetic, the LDL rises, right? So now LDL is correlating what high HbA1c.
Dave 14:22 Right.
Ivor 14:23 And later when you’ve even higher HbA1c, the LDL comes down again. And then when you get to really, really high HbA1c, the LDL goes right up again – really high. So it’s such yet another example where LDL is tracking with other much more important causes, and getting gifted with a correlation with heart disease.
Dave 14:42 And you want to know my theory for that by the way? My theory for that is because what’s actually happening is you have a higher triglyceride response, especially in like a very intensive, refined carbohydrate diet, right? It’s kind of comparable to the glucose response, where basically you have a hyperinsulinemic response that shuttles away both the glucose in the higher triglycerides. But per the energy model, what gets leftover? Higher LDL. You have higher LDL because there was more trafficking of the VLDLs leaving LDL leftover.
Ivor 15:16 So in a sense, LDL almost acts like a kind of remnant there. It’s an after-effect of much more important processes that are linked to disease.
Dave 15:25 Right. That’s my theory on that. Again, I like to say this a lot. But looking at LDL is sort of like looking at the last chapter of a long book. You don’t know the story before the last chapter so you can’t say a lot about what it means.
Ivor 15:39 And to that analogy – it is certainly not a summary chapter either.
Dave 15:43 No, it’s not.
Ivor 15:45 It’s a specialized, disjointed, separate chapter just stuck at the end of the book!
Dave 15:51 Right, right. And that’s why like I’ve always been excited to have the opportunity to look at LDL in the context of these other markers. That’s why the NHANES was a game changer for me personally. But I think I want to make it also a game changer for us in this conversation. You and I, we obviously agree on quite a lot. But there’s a lot of people who are watching us right now who don’t agree. Especially if you yourself are already a scientist and you can get access to the NHANES data because it is publicly available – and as I understand a lot of university students work from it as well – then you can carry this conversation forward with us, because unlike trading studies, we can actually trade analysis on the same dataset.
16:29 I would be interested in what anybody could come up with, with the NHANES data in a justifiable way, something that makes sense, to show that, “No, LDL is worse at higher levels for all-cause mortality.”
Ivor: 16:42 Yeah. Or even perhaps with some more analysis – and I know you have some tentative stuff – we won’t get into yet, even cardiovascular mortality. When HDL is appropriately high and triglyceride is low, we may yet see that even for cardiovascular mortality the higher LDL does not indicate any extra mortality.
Dave 17:02 Possibly.
Ivor 17:04 Excellent. So I’ll tell you what I’m going to do for this Dave, I always intend it, but I don’t get to it, I’m too busy. I will take some of your graphs in the presentation as we discussed this in this podcast, and put them in there just for people to see them because they’re quite dramatic. And, it’s just fascinating to look at them. And they’re so iconoclastic or disruptive to most of the world’s belief system. So we’ll do that.
17:28 Hopefully people will look at an NHANES database. As you say, it’s freely available. Do some analysis, come up with counterpoints, but I suspect that they won’t be able to come up with any. That’s just my thought.
Dave 17:40 We’ll see.
Ivor 17:41 We’ll see. Thanks a lot, Dave.
