#380 ‒ 种子油的争议：与其他膳食脂肪相比，它们是否具有独特的危害性？|莱恩·诺顿博士

如果你看一下置信区间，这个置信区间几乎跨越了1。

And if you look at the confidence interval, the confidence interval nearly crossed the one.

当置信区间跨越1时

And when a confidence interval crosses the one

这在原始分析中是这样的吗？

Now, was that in the original analysis?

我知道在拉姆斯登的重新分析中，结果是1.03到2.8。

I know in the Ramsden re analysis, it was 1.03 to 2.8.

这就是最初发表的结果吗？

Was that what was originally published?

我其实不记得那是不是发表过的数据。

I actually don't recall if that was what was published.

或者那可能就是原始数据。

Or maybe that was the original one.

不，抱歉，我认为那就是最初发表的结果。

No, I'm sorry, that was the original published, I believe.

是的

Yeah.

考虑到它所显示的风险，这个置信区间相对宽泛。

That confidence interval is relatively wide considering the risk it's showing.

当然，支持我观点的研究也存在一些局限性。

And so, because of course, the studies that support my contention, there's limitations on those as well.

但我觉得，如果我要挑出最关键的一点，那就是反式脂肪的纳入。

But I think the most powerful thing if I was going pick it out, it's really the inclusion of the trans fats.

这并不是研究人员的过错，因为在这些研究进行时，人们还不知道反式脂肪会有这种影响。

Which is through no fault of the researchers of their own because when these studies were done, they just didn't know that trans fats had that effect.

好的。

Okay.

所以，再次说明，我扮演的是‘植物油有害’这一立场的角色。

So, again, I'm playing the role of, I believe seed oils are bad.

因此，如果我们回顾随机对照试验，最大的争议在于，这些试验进行时，反式脂肪正是当时主流的替代脂肪，我不知道该如何调和这一点。

So, the biggest contention we would have if we go through the RCTs is because these RCTs were done at a time when trans fats were the substitution fat du jour, I don't know how we're gonna reconcile that.

归根结底，问题是：我是否认为反式脂肪没有你想象的那么有害？

Basically, it comes down to do I believe that trans fats are less problematic than you believe?

因为如果我认为反式脂肪实际上无害，那就会削弱你的论点。

Because if I believe that trans fats are actually not harmful, then that would take away your argument.

你就没有论点了。

You wouldn't have an argument.

除了反式脂肪之外，你还有其他论据吗？

Is there any other argument you've got besides trans fats on this?

我认为持续时间较短也是一个因素，因为如果我们看看那些没有被反式脂肪干扰的其他研究，其中一些持续时间更长。

I think the relatively short duration is another thing because if we talk about some of the other studies where it wasn't confounded by trans fats, some of those were longer.

让我换种方式说。

Let me put it a different way.

假设我们把反式脂肪排除在外。

Let's just say we take the trans fats out of it.

我们就这么假设。

Let's do that.

在随机阅读分析中也是如此。

And this was in the random read analysis as well.

如果我们纳入所有关于用多不饱和脂肪替代饱和脂肪的人体随机对照试验，总体效果为零。

If we include all the human randomized control trials, looking at substituting polyunsaturated fats for saturated fat, the overall effect is null.

无论哪种情况都没有明显影响。

There was no effect one way or the other.

医生。

Doctor.

您这么说，是指最大的考科蓝分析吗？

When you say that, you're referring to the largest Cochrane analyses?

有两个，对吧？

There are two, correct?

医生。

Doctor.

是的。

Yep.

拉姆斯登也做了一项分析，我认为，其中一些显示多不饱和脂肪酸有益的试验，受到了其中包含Omega-3脂肪酸的干扰。

Ramsden also did an analysis, I believe, where some of the trials that showed a benefit to PUFA were confounded by the fact that there was omega-3s included in them.

现在，我认为，如果查看关于Omega-3的文献，其实并没有很强的证据表明它们能降低严重心血管疾病终点事件的风险。

Now, I would argue that if you look at the literature on omega-3s, it's actually not super strong that they decrease hard cardiovascular disease endpoints.

但即便如此，它仍然是一个混杂变量。

But nevertheless, it's a confounding variable.

于是他把这些排除后发现，好，当我们剔除Omega-3后，风险反而上升了。

So he took those out and then showed, okay, when we take out omega-3s, we see an increased risk.

但当然，这仍然包含了反式脂肪的混杂因素。

But of course, that still includes the trans fats confounders.

所以，如果我们把这两者都包括进来，全部混在一起，总体效果就是：无论怎样，减少饱和脂肪、增加多不饱和脂肪，风险是相等的。

So, if we just include both and lump them all in together, the net overall effect is that one way or another reducing saturated fat, raising PUFA, it was equal risk.

因此，如果有人说‘植物油对人类健康具有独特危害’，即使我们把反式脂肪剔除出去，如果你允许这些混杂因素支持你的观点，那么你也必须允许我的混杂因素支持我的观点——我称之为逻辑对称性：如果我要用这个混杂变量来支持我的论点，我也必须接受你的混杂变量支持你的论点。

And so, what I would say if the statement is seed oils are uniquely deleterious to human health, even if we take the trans fats out of it, and if you're going to allow those to be included, you have to have, I call it logical symmetry, which is if I'm going to allow this confounding variable to support my point, I also have to allow your confounding variables to support your point.

否则，我们只能去找那些同时剔除这两者的研究。

Otherwise, we just gotta find the ones that take both of them out.

所以在这种情况下，总体影响仍然是无害的。

And so, in that case, the net effect is still no harm.

另一个混杂因素是什么？

What's the other confounding variable?

EPA和DHA吗？

The EPA and DHA?

没错。

Correct.

如果你想讨论其他一些研究，我会提到他们引用的另一项研究——罗丝玉米油试验。老实说，我对这项研究没什么看法，因为整个试验只有大约70人。

If you want to start talking about some of these other studies, I will touch on the one other study they cite is the Rose Corn Oil Trial, which I quite frankly, I don't know what to make of it because it was only like 70 people in the entire thing.

是的，对照组有26人，橄榄油组有26人，玉米油组有28人。

Yeah, it was 26 people in a control group, 26 people on an olive oil group and 28 people in a corn oil group.

所以，这又是一项为期两年的研究。

So again, this is a two year study.

这项研究的对象是患有严重心血管疾病的人群。

This was in people with significant cardiovascular disease.

所以对照组的人照常生活，其中有三人去世。

So the people in the control group, business as usual, three of those people died.

橄榄油组有五人去世，而玉米油组有八人去世。

Five people died in the olive oil group but eight people died in the corn oil group.

抱歉，我说的是总死亡人数。

Sorry, that was total deaths.

对。

Right.

仅供参考。

For what it's worth.

心脏相关死亡人数：对照组一人，橄榄油组三人，玉米油组六人。

Cardiac deaths, one in the control group, three in the olive oil group, six in the corn oil group.

再次说明，由于样本量非常小，这些数据的置信区间非常宽。

Now again, the confidence intervals on this were very large because the sample size was very small.

这个研究有点棘手，因为参与者是直接被分配了食用油的。

This one is a bit challenging because the participants were given their oils.

你可以把这看作是PretaMed研究的早期版本，这是一项自由生活研究，但橄榄油组的参与者会被提供橄榄油。

Could You think of this as kind of a early version of the PretaMed study where it was a free living study but the participants were given olive oil when they were in the olive oil group.

共有三个组。

There were three groups.

有一组是低脂饮食，另一组是高单不饱和脂肪饮食，通过橄榄油摄入，每周发给你一瓶橄榄油；或者通过坚果摄入，每周发给你坚果。

There was low fat and then there was high MUFA, high monounsaturated fat through olive oil and you were given a bottle of olive oil every week or through nuts and you were given the nuts every week.

所以，这是一项自由生活研究，但参与者会被提供玉米油或橄榄油。

So here, was a free living study but you were given your corn oil or you were given your olive oil.

除此之外，我们没有饮食回顾数据。

Other than that, we don't have dietary recall.

医生。

Doctor.

现在，这种益处没有被反式脂肪混淆。

Now, the benefit, not confounded by trans fats.

这里没有反式脂肪的干扰，这是一个积极因素。

There was no confounding with trans fats here, so that would be a positive.

但我想首先我要说的是，我觉得你提到过，心血管疾病死亡人数是一、三和六。

But I think the first thing I would say is, I think you said this, the cardiovascular disease deaths were one, three and six.

是的。

Yes.

我不知道该怎么理解这个数据。

I don't know what to make of that.

这些数字太小了，很容易受到抽样误差的影响，如果你有上万名参与者，这种差异很可能被稀释掉。

Those are such small numbers that you're so prone to sampling errors where if you'd had 10,000 participants, that's very likely gonna get lost in the wash.

我们单看橄榄油组吧。

And let's just take the olive oil group.

即使是反对种子油的人，通常也承认橄榄油对心脏有益。

Even people who are anti seed oil typically acknowledge that olive oil is heart healthy.

但在橄榄油组中，心血管疾病死亡人数是对照组的三倍。

Well, you had three times the deaths from cardiovascular disease in the olive oil group compared to the control group.

所以这与我们现有的证据不符，而且样本量太小了。

And so, it doesn't fit with the evidence we have and it's so small.

就像你说的，置信区间意味着，如果我们认为1.03到2.8这个范围已经很宽了，那这个研究中的置信区间其实是0.6到37。

And like you said, the confidence intervals mean, if we wanna talk about a confidence interval being wide of 1.03 to 2.8 as being wide, the confidence interval in this, I believe was like point six to 37.

我的意思是

I mean

所以这也统计上不显著吗？

So not statistically significant either?

不是。

No.

不是。

No.

这个差距实在是太大了。

It was absolutely massive.

但这里我导师以前常说，如果你对数据足够折磨，它就会承认你希望它呈现的结果。

But and here's where my PhD advisor used to say, if you torture the data enough, it will confess what you wanted to show.

所以如果你只说多不饱和脂肪组的死亡人数是六倍，从技术上讲你是对的，但你忽略了数据中非常重要的部分。

And so if you just say, well, there was six times the number of deaths in the polyunsaturated group, you're technically correct, but you're leaving out a really big portion of the data.

而且，当我们把这些研究全部纳入荟萃分析，不剔除任何混杂变量时，我们看到了什么？

And again, when we plop all these studies into the meta analysis, don't take out any confounding variables, what do we see?

我们看到的是无显著效应。

We see a null effect.

现在，如果我们剔除反式脂肪，我记得2017年有一项荟萃分析，他们整合了所有随机对照试验，这些试验研究的是用多不饱和脂肪替代饱和脂肪，且未受反式脂肪干扰。

Now, if we take out the trans fats, there was a meta analysis I think in 2017 where they put together all the trials, there were human randomized control trials looking at substitution of polyunsaturated fats for saturated fats not confounded by trans fat.

这种替代带来了非常明确的益处。

This was a very clear benefit to substitution.

我认为死亡风险降低了大约百分之二十。

I think it was around twenty percent reduced mortality risk.

不，甚至更高。

No, it was even more.

我的优势在于可以作弊，因为我正在看原始数据——实际上，玉米油试验的危险比是迄今为止所有试验中最高的。

I mean, I have the luxury of cheating because I'm looking at the So actual the rose corn oil trial has the largest hazard ratio of any trial ever done.

这个试验。

This Of this these trial.

试验。

Trials.

是4.64。

Was four point six four.

对。

Right.

这意味着，如果你食用玉米油，风险增加了364%。

What that means is there was a three sixty four percent increase in risk if you took corn oil.

但正如你所说，这并未达到统计学显著性，因为风险比的95%置信区间是0.58到37.15。

But it did not reach anywhere near statistical significance to your point because the hazard ratio confidence interval, the 95% confidence interval was 0.58 to 37.15.

所以，如果我们仅限于这一批文献和这五项研究，我们并未获得统计学显著性，尽管确实存在风险上升的趋势。

So, if we limit ourselves to this body of literature and the one, two, three, four, five studies, we do not achieve statistical significance, though we do have a trend towards risk.

这一趋势是1.13，意味着相对风险增加了13%。

That trend is 1.13, meaning a 13% relative risk increase.

如果我们把反式脂肪也包括进来，

If we include the trans fat,

这些数据。

the props.

是的。

Yes.

所以，正如你所说，这个分析有点混乱，因为我们同时纳入了MCE、悉尼研究，还包含了红花玉米油。

So, is a bit of a messy analysis to your point because we're including the MCE, we're including Sydney, but we're also including rose corn oil.

我们只是试图让尽可能多的研究进入荟萃分析的引擎，但仍然没有达到显著性。

We're just trying to get as many bodies as possible through the engine of the meta analysis and we don't reach significance.

事实上，唯一达到统计显著性的研究是悉尼心脏研究。

In fact, the only study that reaches statistical significance is the Sydney Heart Study.

该研究显示风险增加了74%，置信区间为1.04到2.91。

That had a 74% increase with a confidence interval of 1.04 to 2.91.

所以它确实达到了显著性，但我猜你的观点会是：没错，但他们的反式脂肪含量高达25%到50%。

So, it got there, but I guess your argument is going to be yes, but they were at 25 to 50% trans fats.

所以，这确实是个难以下定论的问题，莱恩，因为一方面，这里多不饱和脂肪看起来相当不利。

So, this is a tough one to score, Lane, because on the one hand, it looks pretty bad for polyunsaturated fats here.

每次你吃它们，似乎都让你朝着错误的方向发展，除了亚油酸VA研究。

Every time you eat them, it just seems to move you in the wrong direction, except for the linoleic VA study.

嗯，有几项不同的人类随机对照试验。

Well, there's a few different human randomized control trials.

所以VA研究，这项研究有几个优点。

So the VA study So there's several strengths to this study.

告诉我们这项研究是关于什么的。

Tell us what the study was about.

这项研究是在退伍军人之家进行的，饮食是受控的，研究方为参与者提供食物，参与者人数大约为850人，平均随访时间我认为接近九年。

So it was in veterans' homes, the food intake was controlled, they provided it to the participants and it was I think around eight fifty participants, and the average follow-up I believe was just under nine years.

对于这类研究来说，这实际上已经相当长了。

So that's actually a pretty long study for this.

没有受到反式脂肪的干扰，而且他们确实显示了风险降低了大约百分之二十到三十，虽然我手头没有原始数据。

Not confounded by trans fats, and they did show, again, don't have the raw numbers in front of me, I think it was around a twenty or thirty percent reduction in risk.

不过，它并没有达到统计学显著性。

It was though it did not reach statistical significance.

它的置信区间并不宽，实际上相当狭窄，但跨越了1。

It didn't have a wide confidence interval, it was actually quite narrow but it crossed unity.

所以，它显示总体风险降低了18%，但你的置信区间是0.56到1.21。

So, it had a 18% reduction in overall risk but it was point five six to 1.21 was your confidence intervals.

但这一点与你刚才说的相反，如果相信这一点的话，他们包含了Omega-3。

But, this is one that's on the other side of what you said a moment ago, they included omega threes if we believe that.

所以问题是，Omega-3有帮助吗？还是这项研究样本量不足？

So, the question is, did the omega-3s help or was the study underpowered?

是因为这个原因才没有达到统计显著性，还是根本就无关紧要？

Is that why it didn't reach statistical significance or does it just not matter?

这就是为什么我们要做荟萃分析，因为当涉及到单个研究时，尤其是像这样终点指标比较混乱的情况，即使800人听起来很多，但当你试图分析死亡率和心血管事件这类硬终点时

I mean this is why we do meta analyses because when it comes to single studies, sometimes especially when things are messy like this with hard endpoints, even if 800 people sounds like a lot, when you're trying to get hard endpoints like mortality and cardiovascular disease events

这规模其实很小。

It's pretty small.

医生。

Doctor.

确实很小。

It's pretty small.

还有其他一些试验。

There are some other trials out there.

奥斯陆心脏健康研究，只有400人，但显示风险降低了大约47%。

The Oslo Heart Health Study, that was only 400 people, but that showed, I think, like a forty seven percent risk reduction.

同样，受到Omega-3的干扰。

Again, confounded by omega-3s.

然后，我认为最强有力的研究之一是芬兰医院研究。

Then there was probably one of the strongest studies I think is the Finnish hospital study.

原因在于它没有受到反式脂肪或Omega-3的干扰。

And the reason is not confounded by trans fats, not confounded by omega threes.

参与者每种饮食都坚持了六年，因此这是一个交叉设计。

People did each diet for six years, so it was a crossover design.

所以，我有必要简要介绍一下什么是交叉设计。

So I guess it's important for me to cover what a crossover design is briefly.

交叉设计是指让受试者接受两种治疗。

So a crossover design is when you take people and you have them do both treatments.

不过，这种方法也有缺点。

Now, there are downsides.

缺点是，什么是交叉效应？比如，你在某种饮食期间发生了心血管事件，但在此之前你刚完成另一种饮食。

The downside is, okay, what about a crossover effect where, okay, maybe you have, let's say a cardiovascular event on one of the diets, but you did the other diet before that.

那么，这个事件是源于你当前的饮食，还是之前的饮食呢？

So is it from the diet you're on now or is it from the previous diet?

研究人员通常通过让受试者以不同顺序接受两种饮食来解决这个问题。

The way researchers attempt to get around that is by essentially making the order in which they do both diets split.

也就是说，有50%的人先做一种饮食，再做另一种；另外50%的人则反过来。

So that there's fifty percent of people did one diet first and then the next diet and then they reversed it.

通常，这是通过随机分配每个受试者的饮食顺序来实现的。

Usually you do that by randomizing each individual person.

但这项研究也有其弱点。

Now here's the weakness of the study.

他们没有逐个随机分配，而是用了两家医院。

They didn't randomize individually, they had two hospitals.

一家医院先开始一种饮食，另一家医院开始另一种饮食，六年后他们互换了饮食方案。

One hospital started with one diet, one hospital did the other diet, and then they switched them after six years.

这完全说得通。

Completely makes sense.

可以理解。

Understandable.

从后勤角度来说。

Logistically.

是的。

Yep.

否则，你就要试图

Otherwise, you're trying to like

弄清楚床A的鲍勃和床B的萨莉分别在吃什么饮食。

Figure out if Bob in bed a is on this diet versus Sally in bed b.

没错。

Exactly.

我完全理解他们为什么这么做。

I understand why they did it for sure.

我想，如果你要论证这引入了大量偏差，你的论点可能是两家医院本身存在某些固有差异，使得其中一家的患者更容易患上心血管疾病。

I guess if you were going to try and make the argument that this introduced a lot of bias, the argument you would make is that perhaps there was inherent characteristics about one hospital versus the other that were more prone to people getting cardiovascular disease.

我认为这种风险可能很低，但确实可以提出这样的论点。

I would say that risk is probably pretty low, but it's an argument that can be made.

这项研究的好处是包含了1200人，但由于他们进行了交叉设计，实际上相当于2400人。

The benefits of this is it was 1,200 people, but since they crossed them over, effectively it was like 2,400 people.

交叉实验的一个优势是能够提高统计效力。

One of the benefits to a crossover experiment is you can up your power.

我希望有听众比我更精通统计学，因为这已经过去太久了，而我以前是知道答案的。

I hope somebody listening is sharper than I am in my statistics because it's been so long, and I used to know the answer to this.

我觉得这甚至比从1200人提升到2400人更令人印象深刻。

I think it's actually even more impressive than 1,200 to 2,400.

我认为这更强大，虽然不想过度使用‘功效’这个词，但交叉设计在统计上更有力，因为每个人都能成为自己的对照。

I think it's even more, not to overuse the word power, but it's even more powerful statistically when you do a crossover because every person gets to be their own control.

没错。

Exactly.

这可能是最大的优势，正如你所说，它关系到你如何进行t检验和统计比较。

And that is probably the largest benefit, which is, as you said, it's the way you do the t test and the statistical comparisons.

每个人都是自己的对照。

Each person is their own control.

因为在平行组设计中，你只是让一组人采用一种饮食，另一组人采用另一种饮食，你假设随机化过程能将可能存在于两组之间的基线特征随机分布，使它们在各组间均衡，但你并不能完全确定这一点。

Because when you're doing a parallel group design, which is where you just have one group on one diet, one group on the other diet, you are assuming that the randomization process randomly distributes the baseline characteristics that may be different between groups to where they're equally distributed amongst the groups, but you don't know that for sure.

但当你采用交叉设计时，虽然不敢说绝对保证，但你可以非常有信心地认为，固有的基线特征不再构成混杂变量。

But when you cross over, now you're guaranteed, I don't want to say guaranteed, but you are very confident that inherent baseline characteristics are now no longer a confounding variable.

而且人们在整个实验过程中的整体生活方式很可能保持相似。

And it's likely that the people's overall lifestyles are going to be similar throughout the course of the experiment.

由固有生活方式差异带来的偏倚风险也更低。

There's less risk of bias from inherent lifestyle differences as well.

有些人压力更大，有些人压力更小，诸如此类的因素。

Somebody having more stress versus less stress, those sorts of things.

在这项研究中，再次受到反式脂肪的干扰，但未受Omega-3的影响，且样本量充足，持续时间长，每种饮食长达六年，总计十二年。

And in this study, again, confounded by trans fats, not confounded by omega-3s, and pretty well powered for a long duration, six years on each diet, twelve years overall.

他们观察到心血管疾病事件和死亡风险显著降低。

And they saw a pretty significant reduction in the risk of cardiovascular disease events and mortality.

是的，治疗组的风险降低了41%。为了让大家了解实际的干预措施是什么。

Yeah, forty one percent reduction in the treatment group that was and just to give folks what the actual intervention was.

高饱和脂肪饮食组（对照组）的饱和脂肪含量为18%，多不饱和脂肪为4%，三到四左右。

So, the high saturated fat or control arm was 18% saturated fat, 4% polyunsaturated fat, three to four.

他们留了一些弹性空间。

They gave some wiggle room.

治疗组的饱和脂肪含量为14%。

The treatment group was fourteen percent.

因此，他们将多不饱和脂肪提高到14%，并将饱和脂肪减少一半至9%。

So, they increased to 14% polyunsaturated and cut saturated fat in half to 9%.

这些目标与MCE研究有点相似。

These were kind of similar targets to the MCE study.

我想他们还测量了组织中的亚油酸。

I want to say they measured linoleic acid in tissue too.

我相信他们确实这么做了。

I believe they did that.

我不确定。

I don't know.

如果我错了，我愿意认错，但我认为他们实际上检查了亚油酸的组织和血液生物标志物，以证实饮食干预确实提高了亚油酸含量。

If I'm wrong, I'll eat crow but I think that they actually looked at like tissue and blood biomarkers of linoleic acid and showed an increase in linoleic acid content to verify the dietary treatment worked.

你对这个现象最好的解释是什么？

What would be your best explanation for this?

因为再次强调，芬兰研究显示，低饱和脂肪、高多不饱和脂肪的饮食使心血管疾病风险降低了41%，而且置信区间非常窄。

Because again, the Finnish study, I mean forty one percent relative risk reduction in cardiovascular disease with that low saturated fat, high polyunsaturated fat, and the confidence intervals are really tight.

我认为，这正是为什么在这种情况下，尽管我承认一些研究中多不饱和脂肪似乎有负面影响，但我仍认为那些显示风险降低的研究整体上更可靠。

I think, again, this is why this is a situation where, okay, I'm acknowledging what seems to be a negative effect of polyunsaturated fats in some of these studies, but I'm explaining why I think that the studies that show a reduction in risk are on balance better.

再次说明，我并不是在批评这些研究人员，因为正如我们之前讨论的，我认为他们在当时的信息条件下已经尽力而为了。

Again, not making a criticism of the researchers because as we talked about, I think they did the best they could at the time with the information they had.

所以这绝对不是对研究人员的指责。

So this isn't an indictment on the researchers whatsoever.

但当你观察那些显示风险降低的研究时，它们持续时间更长，通常控制得更好，而且没有包含反式脂肪。

But when you look at these studies that showed a reduction in risk, they're longer, they're typically better controlled, and they don't have the inclusion of trans fats.

我认为这些因素是

I think those things are

最关键的驱动因素。

the most powerful movers.

因此，我们认为这就是芬兰心脏研究得出不同结论的最好解释。

So we think that that is the best explanation for why the Finnish Heart Study came up with a different answer.

医生。

Doctor.

是的。

Yes.

2017年有一项荟萃分析，我记不清第一作者的名字了，但他们基本上对那些未受反式脂肪干扰的研究进行了荟萃分析。

There was a meta analysis in 2017, and I can't remember the name of the lead author, But they basically did a meta analyses of the studies not confounded by trans fats.

其中一些研究包含了鱼油或欧米伽-3脂肪酸。

Now, included some of the ones that had fish oil in them or omega-3s rather.

但总体来看，我想说的是风险降低了大约20%。

But they showed overall, again, I want to say it was around a 20% reduction in risk.

所以，如果你看看拉姆斯登的研究，即使在多不饱和脂肪最不利的情况下，风险增加也很小，且置信区间非常宽。

So again, if you look at the Ramsden analysis, in their best case for polyunsaturated fats being bad, it's a smaller increase in risk and the confidence intervals are very wide.

让我暂时扮演一下反对者的角色。

So let me play devil's advocate for a moment.

这可能是个有点荒谬的观点，但会不会这更多反映的是饱和脂肪的危害，而不是多不饱和脂肪的益处？

And again, this is sort of maybe a silly argument, but what if this says more about saturated fats harm than polyunsaturated fats benefit?

在进行营养学研究时，这种情况确实存在。

Well, and this is a situation where when you're doing nutritional studies

你必须用其他东西来替代。

You have to substitute something.

对，对。

Right, right.

因为

Because

如果说是等热量的话。

If it's going to say isocaloric.

在我们讨论与另一位人士进行辩论时，我明确表示，我并不是说种子油完全没有任何有害影响，因为饮食中添加的油脂是热量的主要来源，而多余的热量并非无害。

One of the things when we were talking about doing this debate with the other individual, I was very clear in saying, I'm not saying that there's no deleterious effect to seed oils whatsoever, because added oils in the diet are a major source of calories and excess calories are not innocuous.

所以我们必须进行公平比较，也就是在一对一替换的情况下，哪种更有益？

So, we have to compare apples to apples, which is when you are substituting in a one to one ratio, what is more beneficial?

事实上，我认为你可以取任何食物或任何营养素，并找到它激活的正面和负面通路。

The reality is, I think you could take any food or any nutrient, and you can find both positive and negative pathways that it activates.

问题不在于某种营养素或特定食物是否激活了正面或负面通路。

The question is not whether or not a nutrient or a particular food activates positive or negative pathways.

我之所以这么说，是因为正如你我都知道的，根本不存在所谓的正面或负面通路。

I say that very broadly because, as you and I both know, there's not really such a thing as a positive or negative pathway.

但总的来说，可能会有正面和负面的结果。

But just in general, there could be positive and negative outcomes.

重要的是，总体来看，净效应是什么？

What matters is on balance, what is the net effect?

我来解释一下，我会用另一个金融例子。

And I explain this, I'll use another financial example.

我们正在讨论这些通路。

So we're talking about these pathways.

我们讨论的是机制，这些是很重要的。

We're talking about mechanisms, which are important.

如果某种结果存在，就一定有相应的机制来解释它。

If an outcome exists, there's a mechanism or mechanisms to explain it.

但这些就像单只股票。

But those are like single stocks.

而像心血管疾病事件这样的结果，就像一只共同基金。

An outcome, like a cardiovascular disease event, that is like a mutual fund.

我的意思是，我可以拿一只表现非常好的共同基金，比如年收益率达到20%。

And so what I mean by that is I could take a mutual fund that's doing really well, say, up 20% year over year.

但我可以在其中找到几只股票。

But I could find a few stocks in it.

我会说，哦，你不想投资这只股票。

I'm like, oh, you don't wanna invest in this.

看看这些。

Look at these.

它们下跌了大约50%。

They're down like 50%.

但更重要的是，是这些下跌的个股，还是整个共同基金表现优异？

But what matters more, those individual stocks that are down or the overall mutual fund is killing it.

你关心的是整个共同基金的表现。

You care about what the overall mutual fund is doing.

举个例子，吸烟会降低患帕金森病的风险。

And just an example of this, smoking decreases the risk of Parkinson's disease.

这一效应在研究文献中非常一致。

This is a very consistent effect in the research literature.

但总体而言，没有人会认为吸烟对你有益。

But on balance, don't think anybody's going to say smoking is good for you.

我们应该提醒听众，似乎是尼古丁带来了这种益处。

We should just point out to listeners, it appears to be nicotine that is causing that benefit.

我们不希望任何帕金森病患者或有患病风险的人去吸烟。

Don't want anybody with Parkinson's disease at risk for Parkinson's picking up cigarettes.

如果你真要采取什么措施，那就选择尼古丁吧。

If you're going to do anything, just choose some nicotine.

请选择非烟草来源的尼古丁，拜托了。

Choose nontobacco nicotine, please.

但或许更好的类比是，我认为阿司匹林总体上是一种抗凝剂，但它也会激活一些促凝途径。

But perhaps a better comparison would be, we know I believe aspirin is an anticoagulant overall, but it also activates some procoagulant pathways.

但其总体效果是抗凝的。

But the overall effect is anticoagulation.

所以，如果我只是想挑出来说，看，它激活了这些通路，但这并不是说不重要，总体而言，它对抗凝作用是净正面的。

So if I just wanted to pick out and say, Well, look, it activates these pathways, not that it doesn't matter, but on balance, on balance, it is a net positive for anticoagulation.

因此，当我们讨论多不饱和脂肪与饱和脂肪时，我们稍后会深入这些机制，是的，确实存在一些机制表明多不饱和脂肪可能有负面影响。

And so, when it comes to looking at polyunsaturated fats versus saturated fats, and we will get into the mechanisms of these, Yes, there are mechanisms that exist that would suggest that polyunsaturated fats can have a negative effect.

但总体来看，更多机制表明饱和脂肪才是负面因素。

But on balance, there are more mechanisms that exist that show saturated fat to be a negative.

回到你最初的问题，这是多不饱和脂肪有益，还是饱和脂肪有害的影响？

To circle back to your original question, is this a effect of polyunsaturated fats being beneficial or saturated fats being a negative?

我认为几乎不可能把这两个问题分开，因为在营养学研究中要准确分析，你关注的是替代关系。

I think it's almost impossible to disconnect those two questions because when it comes to nutrition research to do it accurately, you're looking at substitutions.

所以，如果你要减少饱和脂肪，就必须用别的东西来替代。

So, if you're gonna take saturated fat out, you've to put something in.

因此，如果我们看那些用碳水化合物替代饱和脂肪的研究，变化其实并不明显。

And so, if we look at studies where you substitute carbohydrate for saturated fat, not really much change.

我想这很可能很大程度上取决于碳水化合物的类型。

I would say that probably depends on the type of carbohydrate very much.

如果你摄入的是富含纤维的碳水化合物来源，我非常确信你会看到风险降低。

If you were doing fiber dense sources of carbohydrate, I'm pretty sure you'd see a reduction in risk.

单不饱和脂肪似乎能降低心血管疾病的风险。

Monounsaturated fats, there appears to be a reduction in risk of cardiovascular disease.

至少在队列研究中，它似乎不如多不饱和脂肪那么有效。

It appears to be not quite as powerful, at least in the cohort studies, as polyunsaturated fats.

但同样，因为我们必须用其他东西替代，那就让我们简单点说。

But again, since we have to substitute something, let's make it very basic.

即使多不饱和脂肪本身没有任何心脏保护作用，只要用它替代饱和脂肪，你依然会看到结果的改善，因此它仍然是心脏保护性的。

Even if there was no net, if polyunsaturated fats didn't do anything cardioprotective, it's still cardioprotective in that when you put them in in place of saturated fat, you have improvements in outcomes.

我认为存在一些机制可以解释为什么多不饱和脂肪具有心脏保护作用。

I would argue that there are mechanisms in place that explain why polyunsaturated fats are cardioprotective.

我想强调的是，我认为很难将这两个问题分开来看。

My point I'm making is that I think it's difficult to disconnect those two questions.

我们是否知道，关于这个问题的考科蓝分析是否混合了包含反式脂肪和不包含反式脂肪的研究？

Do we know if the Cochrane analyses on this question have blended and included the studies across those that contain trans fats and those that don't?

2017年的那个吗？

The one from 2017?

是的。

Yeah.

他们考察了所有多不饱和脂肪酸与饱和脂肪酸的研究。

They look at all the PUFA versus SFA studies.

不是。

No.

他们明确排除了受反式脂肪干扰的研究。

So they specifically excluded ones that were confounded by trans fats.

兰姆代尔的一项研究包含了几乎所有这些研究。

Ramsden had one where it included basically all of them.

总体效果为零，没有任何影响。

Net effect being null, nothing.

当他剔除受欧米伽三干扰的研究后，结果显示多不饱和脂肪酸具有负面影响。

And when he pulled out the ones that were confounded by omega threes, they showed a negative effect of polyunsaturated fats.

反式脂肪还在吗？

With trans fats still in?

但反式脂肪还在。

But with trans fats still in.

好的。

Okay.

那考科兰的没有反式脂肪？

And Cochrane had no trans fats?

没有反式脂肪。

No trans fats.

但含有omega-3。

But did have omega-3s.

但含有omega-3。

But did have omega-3s.

并且显示PUFA有轻微益处。

And had a slight benefit to PUFA.

是的。

Yeah.

嗯，还算不错。

Well, pretty decent.

你想一想，你之前说是多少来着？

Think what did you say it was?

我记得好像是20，我想应该是23%。

It was 20 I want to say it was like 23%.

哦，30%，对，一个

Oh, 30 Yeah, one

二十九或三十一

twenty nine or 31

或者差不多。

or something.

是的。

Yeah.

大概就在那个范围。

Somewhere around there.

很抱歉没有记住原始数据。

I apologize for not remembering the raw numbers.

里面有很多信息在浮动。

There's a lot floating around in there.

有没有人做过排除欧米伽-3和反式脂肪的分析？

Has anybody done the analysis of excluding omega-three, excluding trans fats?

我的意思是，你

Well, mean, you're

基本上我知道你只参考了很少的研究。

basically I know you're down down very few studies.

我觉得你只剩下大约两项研究了。

I think you're down to like two studies.

我只是想问你一个问题。

I just want to ask you a question.

抱歉打断一下。

I apologize to interrupt.

没关系。

That's okay.

当你提到Omega-3时，你指的是亚油酸，还是EPA和DHA？

When you say omega-3s, are you talking linoleic acid, or are you talking EPA DHA?

我认为他们特指的是EPA和DHA。

I believe they're specifically talking about the EPA DHA.

但我知道像α-亚麻酸这样的...

But I know like alpha linoleic

α-亚麻酸。

Alpha linoleic acid.

是的。

Yeah.

α-亚麻酸也是Omega-3的一种。

Alpha linoleic acid is the omega three as well.

α-亚麻酸。

Alpha linoleic acid.

哦，不是。

Oh, no.

那也是其中一部分。

That was part of it.

那也是其中一部分。

That was part of it as well.

因为它转化为EPA和DHA的效率很低。

Because it doesn't convert very efficiently to EPA and DHA.

所以这意味着，那些包含Omega-3的研究，是否特意补充了EPA和DHA？而这些成分当然只能从海洋来源中大量获取？

So does that mean that these studies that contain omega-3s are going out of their way to supplement with EPA and DHA, which of course you can really only get in high quantities from marine sources?

我读过的研究文献中并没有明确说明。

It wasn't clear in the research literature I read.

基本上就是说，饮食中含有一些Omega-3。

It was just basically like, yeah, there was some omega-3s in the diet.

我认为，有两项研究没有受到反式脂肪或Omega-3的干扰。

There's basically, I think two studies that were not confounded by trans fats, not confounded by omega-3s.

那就是我们之前讨论过的那项最终研究。

And that was the finished study that we talked about.

还有STARS研究，但STARS研究并没有关注硬性心血管疾病终点。

And then STARS, which STARS was not looking at hard cardiovascular disease endpoints.

这项研究关注的是斑块的进展。

This was looking at plaque progression.

所以我认为这是一项为期一年的研究，他们比较了低饱和脂肪、高多不饱和脂肪与高饱和脂肪、低多不饱和脂肪的饮食效果。

So I believe this was a one year study and they looked at other people doing low saturated fat, higher polyunsaturated fat versus higher saturated fat, lower polyunsaturated fat.

他们观察了斑块的进展。

And they looked at the progression of plaque.

所以这不是硬性结局，但我认为这项研究的优势在于，你所观察的是我们已知会导致心肌事件的病理过程，即使没有发生心肌事件，也能很好地了解实际情况。

So not hard outcome, but I would say that the strength of this is since you are looking at the progression of what we know causes these myocardial events, even if you don't have a myocardial event, you can get a really good idea of what's going on.

举个例子，我这里只是假设一下：在其他一些研究中，有人可能已经发展到接近99%的血管阻塞，但在研究观测期间并未发生心肌事件，于是就被归为风险降低了。

And so, example, and I'm just gonna make a hypothetical here, you could have somebody in one of these other studies where they got right up to the point where they got like a 99% blockage, but no myocardial event in the time where they were measuring it and they just came up as that's a risk reduction.

但事实上——

But in reality-

他们确实出现了疾病进展。

They had disease progression.

对。

Right.

所以这种方法的好处是，你仍然能得到比生物标志物更硬的终点，但又能比仅仅观察死亡率和心血管事件这类粗略的指标更细致一些。

So the benefit of this is you still get a harder endpoint than just biomarkers, but you can get a little bit more granular than just looking at, I guess the way to describe mortality and cardiovascular disease events is big blunt instruments.

那么，我们现在来多聊聊这些机制性的问题。

So, let's talk a little bit more about these mechanistic things now.

因为again，大多数这些试验都使用了最极端的工具——全因死亡率作为最终结局。

Because again, most of these trials were done with a brute force tool of like the hardest outcome is all cause mortality.

但again，这是一个很高的标准。

But again, that's a high bar.

在进行临床试验时，你得在结局指标和达到该指标的难度之间做权衡，这么说你明白吧？

When you do clinical trials, you trade off between outcome and barrier to outcome, for lack of a better term, right?

所以，如果你想使用全因死亡率，那是最终的衡量标准，但这是一个很高的门槛。

So, if you want to use all cause mortality, that is the ultimate measuring stick, but it's a high bar to clear.

然后你可以转向特定疾病的死亡率。

And then you can go disease specific mortality.

好的，我们现在来看冠心病死亡。

Okay, we're now gonna look at coronary death.

然后你可以再进一步，比如MACE（主要不良心血管事件），包括心肌梗死、中风和心血管死亡。

Then you can go one step lower and an example would be MACE major adverse cardiac events where we're going to use heart attack and stroke and cardiac death.

再往下一层，我们可以只关注胆固醇的变化，或者低密度脂蛋白胆固醇的变化。

Then you can go one step below that and say, well, we're just going to look at a change in cholesterol or we're look at a change in LDL cholesterol.

你对疾病过程的了解会越来越深入，或者说，在MACE之下，下一个层级就是疾病进展。

And you get more and more insight into disease process or I guess below MACE you would go disease progression would actually be your next thing.

然后你开始了。

Then you started.

是的。

Yeah.

让我们从宏观转向微观。

Let's go from the macro to the micro.

在那之前，你可能不同意，但我认为

Before we do that, you may disagree, I but think

现在是个讨论孟德尔随机化研究的好时机。

now would be a good time to talk about the Mendelian randomization studies.

除非你想先讨论疾病进展，或者先讨论机制。

Unless you want to do progression first, or unless you want to do the mechanism first.

好的。

Okay.

你想要

Do you want

从LDL的角度来讨论孟德尔随机化吗？

to talk about Mendelian randomization through the lens of LDL?

是的。

Yes.

好的。

Okay.

首先，我认为有必要说明为什么我现在提出这一点。

So, first off, I think it's important to point out why I'm bringing this up now.

因为就这些研究而言，我认为它们是最有力的。

Because when it comes to these studies, I think that these are the most powerful.

正如你提到的，死亡率是一个难以衡量的结果。

Because as you mentioned, mortality is a difficult outcome to get.

这是一个很高的标准。

It's a high bar to clear.

孟德尔随机化的好处在于，它本质上相当于一项终身的随机对照试验。

The benefit to Mendelian randomization is that essentially you have a lifelong randomized control trial.

孟德尔随机化利用了这样一个事实：在出生时，基因变异是随机分配的。

Now, Mendelian randomization takes advantage of the fact that at birth, genetic variants are randomly assigned.

所以，在一项研究中，如果你和我，彼得，都参与了这项研究，研究人员只会采用他们现有的随机化流程。

So, in a research study, if you and I are in a research study, Peter, the researchers are just gonna do whatever randomization process they have.

好的，彼得，你正在做这个。

Okay, Peter, you're doing this.

莱恩，你也在做这个。

Lane, you're doing this.

完全是随机的。

Completely random.

大自然实际上是出于设计这样做的，我们现在已知大约有十几种变异会影响你的低密度脂蛋白胆固醇水平。

Mother nature actually does that by design, which we now know identified, I think around a dozen variants that will essentially change your LDL cholesterol levels.

有一些变异影响低密度脂蛋白的清除，有一些影响其生成，还有一些影响你吸收的胆固醇量。

There are variants that deal with LDL clearance, LDL production, how much cholesterol you absorb.

有各种各样的变异。

There's all different kinds of variants.

这样做的好处是，你可以观察一个人一生中的低密度脂蛋白胆固醇暴露情况，并试图分析其对死亡率和心血管疾病风险的影响。

Now, the benefit to this is you can look at someone's lifelong LDL cholesterol exposure and attempt to see what is the risk of mortality and cardiovascular disease.

这非常强大，而且这些研究中的有些涉及数十万人。

That is very powerful, and these are studies where some of these have hundreds of thousands of people in them.

而且，自然对这一过程的随机化至关重要，因为如果我们观察到组间的差异，就可以合理地认为这些差异是由该遗传变异的分配所致，而非某种混杂因素。

And again, the randomization of this process by nature is so important because now we can, if we see differences between groups, we can assume it's due to that assignment of that genetic variant versus some kind of confounding variable.

不过，这里有一个注意事项，即为了使孟德尔随机化有效，必须满足几个条件。

Now, there's a caveat here, which is for a Mendelian randomization to be useful, couple things have to be true.

你所说的必须成立。

What you said has to be true.

必须存在对感兴趣变量的遗传分配。

There has to be genetic assignment to the variable of interest.

这一点在许多情况下都成立。

That's true for many things.

身高就是如此。

It's true for height.

体成分也是如此。

It's true for body composition.

许多精神疾病也是如此。

It's true for many psychiatric disorders.

我的意思是，有很多事情都受到遗传的显著影响。

I mean there's lots of things for which genetics play a significant role.

低密度脂蛋白胆固醇就是其中之一。

LDL cholesterol turns out to be one of them.

但这是人们在讨论孟德尔随机化时常常忽略的另一件重要事情：这些基因不能直接影响另一个影响目标结果的变量。

But this is the other important thing that people often forget when talking about MR, which is that those same genes cannot have a direct impact on another variable that affects the outcome of interest.

因为如果它们确实有影响，你的随机化就出问题了。

Because if they do, your randomization just got wonky.

当然。

For sure.

这就是为什么他们会进行多效性检验，我们之前也提到过。

And that's why they do tests for pleiotropy which we talk about.

当我谈到结果时，我会解释为什么这些差异几乎不可能是由多效性引起的。

And when I talk about the results, I'm gonna explain why it's very unlikely that these differences were due to pleiotropy.

但这要求我们做出一个假设。

But it does require us to make one assumption.

这看起来可能微不足道，但我在这里力求逻辑一致，这不是对饱和脂肪与多不饱和脂肪的终身测试。

And this may seem ticky tacky, but I'm trying to be logically consistent here, which is this is not a lifelong test of saturated fat versus polyunsaturated fat.

这是对用多不饱和脂肪替代饱和脂肪后，我们预期LDL胆固醇会发生什么变化的终身测试。

This is a lifelong test of what we expect to happen to LDL cholesterol by substituting polyunsaturated fats for saturated fats.

换句话说，这是对LDL因果关系的检验。

Stated another way, this is a test of LDL causality.

没错。

Correct.

由此直接得出营养对这一影响的结论，是一种跳跃。

It's a leap to then make the statement about what does nutrition do to this.

我再给你举一个偏离常规但与LDL相关、与我们主题无关的例子。

I'm gonna give you another example of this that is off the beaten path, but related to LDL, unrelated to our topic.

我相信我在书里写过这一点，尽管可能在编辑时被删掉了。

And I believe I included this in my book, although it may have got left on the editing floor.

我知道我写过，但记不清它是否最终被保留了。

I know I wrote about it, I can't remember if it made the final cut.

这最终归结为一个问题：现在我摘下辩论的帽子，单纯从孟德尔随机化角度做个补充，因为这很有趣。

And it came down to the question of, so this is me taking off my debate hat and just doing an aside on MR because it's interesting.

一些研究显示，胆固醇水平较低的人更容易患癌症。

A handful of studies showed that people with lower cholesterol were more prone to cancer.

因此，人们开始担心：如果我们为了预防心脏病而降低人们的胆固醇，会不会反而增加患癌风险？

And so, the concern became, well gosh, we shouldn't be lowering people's cholesterol in an effort to prevent heart disease if it increases the risk of cancer.

当然，平均来看情况并非如此，但总有一项研究似乎暗示了这种可能性，而你永远无法确定是否存在混杂因素，毕竟癌症太复杂了。

And of course, on average that didn't seem to be the case but there was always one study that might have suggested that and you never knew if there was a confounder because cancer.

你不可能通过随机分组来发现这一点。

You weren't randomizing to find that.

因此，孟德尔随机化可能变得有价值，前提是你相信那些调控胆固醇合成、吸收和LDL受体表达的基因，并不同时调控癌症进程。

So, is where MR became potentially valuable, provided you believe that the genes that regulate cholesterol, synthesis, absorption and LDL receptor expression don't also regulate a cancer process.

如果你相信这一点，那么孟德尔随机化研究的结果非常明确：LDL胆固醇与癌症之间没有任何关联，无论是正面还是负面。

And if you believe that, then the Mendelian randomization was quite clear that there was no relationship, neither good nor bad, between LDL cholesterol and cancer.

这意味着LDL对癌症没有因果作用，而正如你可能要讨论的，它与ASCVD则存在关联。

So, that meant that LDL had no causal role one way or the other on cancer, whereas as you'll probably talk about it, has a relationship towards ASCVD.

这真是个很好的解释，我刚开始读关于孟德尔随机化的内容时，花了好长时间才理解。

So, that's a great explanation and I think it took me a while to wrap my head around Mendelian randomization when I first started reading about it.

但正是这些研究让我改变了对低密度脂蛋白的看法。

But these really were the studies that made me change my mind on LDL.

我想指出它的优势所在。

I want to point out the strengths to this.

终身暴露——根据脂质假说，我们预期终身暴露于低密度脂蛋白会对心血管疾病和死亡风险产生线性影响。

Lifelong exposure, which based on the lipid hypothesis, we would expect to see a linear effect of lifelong exposure to LDL on the risk of cardiovascular disease and mortality.

而且你可以获得大量受试者的终身数据。

And you can get a large number of subjects for their lifetime.

而且这是随机化的，所以它不是一项队列研究。

And it's randomized, so it's not a cohort study.

这些就是它的优点。

Those are the benefits.

但缺点是，我们仍需假设：我们并没有直接测量人们摄入饱和脂肪与多不饱和脂肪的差异。

The one downside is, again, we're having to make a leap of, okay, we're not directly measuring people eating saturated fat versus polyunsaturated fat.

但我会说，这个推断其实很小，因为已有充分证据表明，增加饱和脂肪的摄入会提高LDL胆固醇，而增加多不饱和脂肪酸则会降低LDL胆固醇。

But I would say that this leap is pretty small because it is very well established that increasing saturated fat intake increases LDL cholesterol, and increasing PUFAs decrease LDL cholesterol.

提高多不饱和脂肪酸并降低饱和脂肪，能显著降低LDL胆固醇。

And raising PUFAs and lowering saturated fat significantly decrease LDL cholesterol.

我认为在大多数研究中，用多不饱和脂肪替代饱和脂肪，LDL胆固醇大约能下降15%。

I think in most studies, you get like a around a 15% change in LDL cholesterol from substituting in polyunsaturated fats for saturated fat.

现在，我要告诉你，有一篇论文发表于大约12年前。

Now, I will tell you there's a paper from, I think it's in 12.

那是第一项关于胆固醇的大规模孟德尔随机化研究。

It was the first big MR study with cholesterol.

当你看他们绘制的所有遗传变异图时，那条线几乎笔直穿过数据点——我们讨论的相关系数可能高达0.9以上，在这类研究中，这是极其显著的关联。

And if you look at the figure where they take all the genetic variants and you look at how the line goes almost straight through it, I mean we're talking about an R of probably above point nine, which in studies like this, that is an incredible association.

每降低一毫摩尔的LDL-

For every one millimole reduction in LDL-

大约相当于37毫克/分升。

Which is about 37 milligrams per deciliter.

医生。

Doctor.

是39.4毫克每分升，对。

It's 39.4 milligrams per deciliter, yeah.

心血管疾病的风险降低了百分之五十到五十五。

There was a fifty to fifty five percent risk reduction in cardiovascular disease.

再次说明，多效性论点基本站不住脚，因为无论是什么类型的基因变异都无关紧要。

Again, the pleiotropy argument is pretty much null and void because it didn't matter what kind of variant it was.

只要它增加了LDL的清除，或者减少了LDL的生成，不管它如何影响LDL胆固醇的代谢，都呈现出相同的剂量效应。

If it increased LDL clearance, if it decreased production, no matter how it affected the metabolism of LDL cholesterol, it had the same dose effect.

唯一的例外是一些CETP基因变异，这些变异在药物试验中也是如此，CETP变异基本上会提高HDL并降低LDL。

The only exception to that was some of the CETP variants, which when it came to drug trials as well, CETP variants, basically I believe they raise HDL and they lower LDL.

但关键在于，这些变异和药物通过降低LDL来减少胆固醇总量。

But, and here's the big point, the lowering of LDL with those variants and those drugs, it lowers the cholesterol mass.

但ApoB或LDL颗粒数量的降低却与之不一致。

But there's a discordant lowering of ApoB or LDL particle number.

每个LDL颗粒上都含有一个ApoB蛋白。

So every single LDL particle has an ApoB protein on it.

我们来探讨一下机制，这将对脂质假说至关重要。

And we get into the mechanisms, this is going to be very central to the lipid hypothesis.

如果你降低了胆固醇总量，但ApoB没有相应大幅下降，那你实际上只是让每个颗粒变小了。

If you lower cholesterol mass, but your ApoB doesn't drop that much, you basically have just made each particle smaller.

在这一类变异中，我们观察到风险略有降低，但这种降低基本可以归因于ApoB的微小下降。

And what we see is in that particular subset of variants, there is a small risk reduction, but it's basically explained by the small decrease in ApoB.

你并没有获得通过降低LDL胆固醇所预期的风险降低效果。

You don't get the predicted risk reduction that you would get with reducing LDL cholesterol.

但在那些同时降低LDL和相应降低ApoB的变异中，效果非常一致。

But amongst the variants that decrease LDL and correspondingly decrease ApoB, it is a very, very consistent effect.

同样，无论哪种基因变异，其风险降低效果都是一样的。

Again, regardless of the genetic variant, it has the same risk reduction.

当我们观察他汀类药物试验时，无论LDL胆固醇通过何种方式降低——不同他汀的作用机制各异——其剂量反应关系都是一致的，约为百分之二十二。

And then when we look at the statin trials, regardless of the way that LDL cholesterol gets lowered, and there's different statins that work different ways, it is the same dose response, which is about twenty two percent.

是的。

Yeah.

所以，我本来想问你，当我们综合所有孟德尔随机化研究时，汤姆·戴斯普林格有一张图，把每一个孟德尔随机化研究、每一个随机对照试验和每一个观察性流行病学研究都画在同一个图表上，用三条线表示。

So, was going to ask you, why do you think that when we look at the totality of the Mendelian randomization, and Tom Dayspring has a figure that lays every single MR study, every single RCT study and every single observational epidemiology study on the same graph with three lines.

基本上，每条线都做了线性回归。

Basically, the linear regression through each.

你认为对于以下观察结果，最好的解释是什么？

And what do you think is the best explanation for the following observation?

也就是说，当你把所有的孟德尔随机化研究都做一遍时——因为大多数听我们讲话、看我们的人，我不想用手势太多。

Which is when you do all of the MR studies and just again, so people, especially because most people are listening to us, watching us, I don't want to use my hands too much.

但在X轴上，你有LDL浓度，在Y轴上，你有死亡率或心血管死亡率。

But on the X axis, you have LDL concentration and on the Y axis, you have mortality or cardiovascular mortality.

正如你所指出的，这些线都是直线下行的。

And as you pointed out, these lines are just going straight down.

这意味着，无论通过基因还是药物，随着LDL胆固醇的降低，心血管死亡率也在下降。

Meaning as LDL cholesterol is going down, either genetically or through drugs, cardiovascular mortality is going down.

但为什么在孟德尔随机化研究中，理论上这应该是最纯粹的研究，每降低一毫摩尔或大约每降低40毫克/分升的LDL胆固醇，就能带来50%到55%的死亡率下降？

But why is it that in the MR study, which in theory would be the most pure study, every millimole or every roughly 40 milligrams per deciliter reduction in LDL cholesterol is giving you 50 to 55% reduction.

但当你用他汀类药物做同样的事情时，你仍然能获得益处，这种益处似乎没有尽头，但只有22%。

But when you do the same thing with a statin, you still get a benefit and that benefit appears non ending but it's only 22%.

这难道不说明他汀类药物是有害的吗？

Doesn't that tell us that statins are bad?

正如你所说，仍然存在风险降低。

Well again, as you mentioned, there's still a risk reduction.

是的，让我换种方式说。

Yeah, let me say it in another way.

这是否暗示，尽管他汀类药物总体有益，但它们也在造成某种伤害？

Does that suggest that while statins are net positive, they're doing something bad?

这是因为，想想人们什么时候开始服用他汀类药物。

So, this is because think about when people are getting on statins.

这么说吧，我个人的倾向是希望LDL胆固醇其实没有那么糟糕，因为我的母系家族都有高LDL的问题。

So, say this as somebody, my bias should be that I hope that LDL cholesterol is not bad because my whole mother's side of the family runs high LDL.

即使摄入低饱和脂肪、高膳食纤维的饮食，我的胆固醇水平仍然在150左右。

Even with low saturated fat diet, even with high dietary fiber intake, I run about 150.

我40岁时就开始服用他汀类药物，但那时我的血管内皮已经暴露在这一水平的LDL中长达40年了。

I started taking a statin when I was 40 years old, but my endothelium, my blood vessels have already had 40 of LDL exposure at that level.

所以，即使我从40岁起每天服用他汀类药物直到去世，仍然会有一些LDL胆固醇渗透进内皮，对动脉粥样硬化的形成产生了一定影响。

And so even if I get on a statin at age 40, every single day to the day I die, there is still some LDL cholesterol that has penetrated that endothelium and has contributed to some degree of atherosclerosis.

我的冠状动脉钙化评分做过检查，数值相当低。

I mean, I had a coronary calcium score done, it was pretty low.

总体而言。

Overall.

由于我胰岛素敏感性良好，还有其他诸多因素，我的整体风险对于我的年龄来说非常低。

My net risk was very low for my age because I have good insulin sensitivity, all those sorts of things.

这一点也很重要。

Important to point out as well.

当我们谈论LDL时，并不是说其他因素都不重要。

When we're talking about LDL, we're not saying everything else doesn't matter.

我想要说的是，这是一个独立的风险因素。

This is an independent risk factor is what I'm saying.

胰岛素敏感性、代谢健康，这些因素仍然很重要。

Insulin sensitivity, metabolic health, all those things still matter.

这些并不是相互排斥的。

These are not mutually exclusive.

但当你天生携带一种能降低LDL的基因变异时，你的整个循环系统从出生起就暴露在更低的LDL胆固醇水平下。

But when you have a genetic variant that lowers LDL from the day you are born, your entire circulatory system is exposed to less LDL cholesterol.

至于有多少胆固醇进入内膜，这取决于浓度。

And when it comes to how much gets into the intima, it is concentration dependent.

这基本上取决于你血液中直径小于70纳米的载有ApoB的颗粒数量。

And it's basically dependent on the number of particles in your bloodstream with ApoB that are under 70 nanometers.

因为任何直径小于70纳米且含有ApoB的脂蛋白都能穿透内皮并被滞留在那里。

Because any lipoprotein under 70 nanometers in diameter with an ApoB can penetrate the endothelium and get retained there.

所以，在这些孟德尔研究中，你看到如此显著的效果，仅仅是因为时间因素。

So when it comes to these Mendelian studies, the reason you see such a powerful effect is just a time effect.

因为大多数人，当他们被处方他汀类药物时，我不知道平均年龄是多少，但我猜大概和我差不多吧。

Because most people, when they get an I don't know what the average age is somebody's prescribed a statin, I'm imagining it's probably around my Yeah.

你其实只是在观察我们之前提到的那个投资类比中的差异。

You're really just looking at the difference in that investment analogy we used earlier.

四十年后，你会看到明显的差异。

Forty years, you'll see a big difference.

但如果你从出生那天就开始投资，七十年后，由于复利效应，你会看到巨大的差异。

But if you start investing from the day you were born, in seventy years, you're going see a massive difference due to compounding interest.

尽管这可能是一个比较粗糙的类比，但这些问题会随着时间累积。

And while it may be kind of a rudimentary comparison, these problems compound over time.

所以，对于LDL偏高的人，给他们开他汀类药物，你显然是在踩刹车，但那辆卡车早已开始从山坡上滑下。

And so what you're doing with somebody who has high LDL, put them on a statin, you're obviously like pumping the brakes, but that truck still started rolling down the hill.

而对于那些一生LDL都偏低的人，它根本就没启动过。

Whereas with people who have lifelong low LDL, it never really got started.

是的。

Yeah.

医生。

Doctor.

我想指出

I want to point

再强调一点。

out one more thing.

需要指出的是，他们对每个变异体都进行了回归分析——医生。

Important to point out, there was a regression they looked at each- Doctor.

每个变异体。

Each variant.

降低一毫摩尔。

One millimole reduction.

因此，无论哪种变异体，降低幅度都相同。

So regardless of the variant, same reduction.

而在药物试验和他汀类药物试验中，无论他汀类药物如何发挥作用，包括像我印象中那种旁路手术或通过抑制吸收来降低LDL的手术也是如此。

And in the drug trials, the statin trials, regardless of how the statins worked, also with surgeries like I think it was a little bypass or things that reduce LDL through absorption.

不，我认为不同的药物，只要它们降低LDL的机制不同，都能带来同样的益处。

No, I think different drugs, that are independent of the mechanism by which LDL is lowered, you'll see the same benefit.

此外，在饮食减少LDL方面，当他们观察一些这些队列研究时，关于风险也是如此。

And then also with dietary reduction of it, in terms of the risk, when they look at some of these cohort studies.

如果你想说是多效性，或者想说是其他可能的原因，但当这种效应非常一致且剂量也非常一致时，这种说法很难成立。

If you want to argue it's pleiotropy or you want to argue it's possibly something else, it's a really hard argument to make when it is a very consistent effect and the dose is also very consistent.

当我们谈论相互印证的证据时。

When we talk about converging lines of evidence.

有一个论点可以提出，但这并不一定是我们今天要讨论的论点。

There's an argument to be made, which is not necessarily the argument we're having today.

我甚至不会试图为它做最强辩护，我直接陈述它。

I won't even attempt to steel man it, I'll just state it.

这个论点是，如果你查看他汀类药物的所有文献，看到死亡率的下降，并不意味着这是因为降低了LDL。

The argument is that if you look at all the literature of statins and you see reduction in mortality, that doesn't mean that it's because it's lowering LDL.

它可能是由于其他原因。

It could be because it's doing something else.

它是在降低炎症，或者在做其他事情，而你认为，考虑到所有孟德尔随机化研究与临床试验数据的结合，这种观点很难成立。

It's lowering inflammation or it's doing something else and you're arguing that that's a tough argument to make in light of all of the MR coupled with all the clinical trial data.

如果是这样，你应该会看到剂量反应关系上的差异。

If that was the case, you would see a difference in the dose response.

你应该会看到设计相似的试验中出现不一致的结果。

You would see inconsistencies in the trials with similar designs.

我举一个有点出人意料的类比，也许能说明这一点。

I'll give a comparison that's kind of out of left field, maybe it'll make the point.

比如，我不认为未加工的红肉本身具有致癌性。

And that is, for example, I don't believe that unprocessed red meat specifically is inherently carcinogenic.

原因是，尽管在一些队列研究中它被列为致癌物，但这种效应并不总是一致的。

And the reason is, even though you see it come up as carcinogenic in some of these cohort studies, the effect isn't always consistent.

当他们控制整体饮食质量时，比如人们摄入足够的水果和蔬菜，因为如果一个人多吃某种食物，往往就会少吃另一种。

And when they control for overall diet quality, where people are eating enough fruits and vegetables, because again, if people eat more of one thing, they tend to eat less of another.

当你控制了整体饮食质量的变量后，红肉与癌症之间就不再有明显的关联了。

When you control for some of the overall diet quality variables, you don't really see a consistent association of red meat with cancer.

现在，我可能是错的，但我对此并不信服。

Now, I could be wrong, but I'm just not convinced by it.

但当谈到膳食纤维对心血管疾病和癌症的影响时，研究文献中存在明显的剂量反应关系，且结果非常一致。

But when it comes to dietary fibers effect on cardiovascular disease and cancer, there's a dose response and it is very consistent in the research literature.

事实上，我几乎不知道有任何研究在考察膳食纤维对降低心血管疾病、癌症和死亡风险的作用时，没有显示出益处。

In fact, I'm not really aware of hardly any study looking at dietary fiber and reducing the risk of cardiovascular disease, cancer, and mortality that doesn't show a benefit.

我的意思是，如果把所有研究的森林图放在一起，所有的结果都会指向保护作用。

I mean, if there's a forest plot of all the studies out there, everything is going be the protection side.

当你看到这种一致性时，即使你可以说，可能是其他因素，也可能是其他因素。

When you have that consistency, even though you could argue, well, it could be other things, it could be other things.

我想，如果你非要提出‘你也是’式的反驳，那我们几乎永远无法真正说某件事导致了另一件事。

I guess if you want to make the whataboutism argument, it's hard for us to ever like actually say something causes something else.

我的意思是，也许不是吸烟导致了肺癌，因为我们无法对吸烟进行随机对照试验，因为这不道德，但我们之所以坚信，是因为剂量效应和结果的一致性。

I mean, maybe it's not the smoking that causes lung cancer because we can't really do randomized control trials on smoking because it wouldn't be ethical, but we feel very strongly because of the dose effect and because of the consistency of the results.

我明白这个论点是可以提出的，但我想说，那你觉得它到底在起什么作用呢？

So, I get that argument that can be made, but I guess I would say, well, then what do you think it's doing?

这非常一致的效果该如何解释？

What would explain this very consistent effect?

通常最后都会变成一种说法：你并不能确定。

It typically just ends up being an argument of, Well, you don't know for sure.

这让我想起在研究生院时，我做了一场关于膳食蛋白质来源中亮氨酸含量的报告，我们用亮氨酸含量不同的各种膳食蛋白质来源做了一项剂量反应实验。

And it reminds me of, in graduate school, I was giving a talk on leucine content of dietary protein sources, and we did a dose response experiment with different dietary protein sources that varied in the leucine content.

结果几乎显示出这些蛋白质来源中的亮氨酸含量、其提升血浆亮氨酸能力以及对蛋白质合成的影响三者之间存在近乎完美的关联。

And pretty much showed almost a perfect association between the amount of leucine in those protein sources, the ability of those protein sources to increase plasma leucine, and the effect on protein synthesis.

当时在场的其他科学家对我说：‘但你不能这么说。’

And I had people that were other scientists in the audience say, Well, but you can't say that.

这些其他的蛋白质来源还含有不同的其他氨基酸。

These other protein sources, they have different other amino acids in them.

这不仅仅是亮氨酸的问题。

It's not just leucine.

还有其他因素在变化。

There's other things that are changing.